HK1246857B - Partially synchronized multilateration or trilateration method and system for positional finding using rf - Google Patents

Description

Partially synchronized multilateration or trilateration method and system for position finding using RF

This application is a divisional application of an application with a filing date of June 26, 2017, application number 201580071160.8, and invention name “Partially synchronized multilateration or trilateration method and system for position finding using RF”.

Cross-reference to related applications

This application claims the benefit of U.S. Provisional Application No. 62/068,537, filed October 24, 2014, entitled “PARTIALLY SYNCHRONIZED MULTILATERATION/TRILATERATION METHOD AND SYSTEM FOR POSITIONAL FINDING USING RF.”

Technical Field

Embodiments of the present invention relate to wireless communication and wireless network systems and systems for radio frequency (RF) based identification, tracking and location of objects, including RTLS (Real Time Location Services) and LTE-based location services.

Background Art

RF-based identification and location-finding systems for determining the relative or geographic location of objects are generally used to track single objects or groups of objects, as well as for tracking individuals. Conventional location-finding systems have been used for location determination in open, outdoor environments. RF-based global positioning systems (GPS)/global navigation satellite systems (GNSS) and assisted GPS/GNSS are commonly used. However, conventional location-finding systems have some inaccuracies when locating objects in closed (i.e., indoor) environments and outdoors.

Cellular wireless communication systems offer various methods for locating the position of user equipment (UE) indoors and in environments less suitable for GPS. The most accurate methods are positioning technologies based on multilateration/trilateration methods. For example, Release 9 of the LTE (Long Term Evolution) standard specifies DL-OTDOA (Downlink Observed Time Difference of Arrival), and Release 11 specifies U-TDOA (Uplink Time Difference of Arrival), which are derivatives of multilateration/trilateration methods.

Since time synchronization errors affect positioning accuracy, a fundamental requirement for multilateration/trilateration-based systems is that the systems are fully and precisely time-synchronized to a single common reference time. Cellular, DL-OTDOA, and U-TDOA positioning methods also require that the transmissions from multiple antennas in the case of DL-OTDOA, or multiple receivers in the case of U-TDOA, be time-synchronized.

In addition, in multilateration/trilateration methods, the target position (e.g., UE position) can be determined relative to the antennas (i.e., UE position versus antenna position). Therefore, inaccuracies in the antenna position database (transmit antennas for DL-OTDOA and receive antennas for U-TDOA) may lead to UE positioning errors.

LTE standards, Releases 9 and 11, do not specify the accuracy of time synchronization for positioning purposes, leaving this to wireless cellular service providers. On the other hand, these standards do not provide limits on ranging accuracy. For example, when using a 10 MHz ranging signal bandwidth, the requirements for DL-OTDOA are 50 meters at 67% reliability and for U-TDOA are 100 meters at 67% reliability.

The above limitations are the result of a combination of range measurement errors and errors caused by a lack of precise synchronization (e.g., time synchronization errors). Based on the relevant LTE test specification (3GPP TS 36.133 Version 10.1.0 Release 10) and other documents, it is possible to estimate the time synchronization error, assuming that the synchronization error is uniformly distributed. One such estimate amounts to 200ns (100ns peak-to-peak). It should be noted that Voice over LTE (VoLTE) functionality also requires cellular network synchronization down to 150 nanoseconds (75ns peak-to-peak), assuming that the synchronization error is uniformly distributed. Therefore, as a next step, the time synchronization accuracy of the LTE network can be assumed to be within 150ns.

Regarding distance positioning accuracy, FCC NG911 mandates 50-meter and 100-meter positioning accuracy requirements. However, for the location-based services (LBS) market, indoor positioning requirements are even more stringent—3 meters at 67% reliability. Consequently, the ranging and positioning errors caused by a 150-ns time synchronization error (43-ns standard deviation) are significantly greater than a 3-meter ranging error (10-ns standard deviation).

While time synchronization of cellular networks may be sufficient to comply with mandatory FCC NG E911 emergency location requirements, this synchronization accuracy falls short of the needs of users of LBS or RTLS systems, who require significantly more accurate location. Therefore, there is a need in the art to mitigate location errors caused by: 1) the lack of accurate time synchronization of cellular/wireless networks for LBS and RTLS purposes; and 2) inaccuracies in cellular/wireless antenna location databases.

Summary of the Invention

The present invention relates to methods and systems for radio frequency (RF)-based identification, tracking and positioning of objects, including real-time location service (RTLS) systems that substantially eliminate one or more disadvantages associated with existing systems. The methods and systems can use receivers and/or transmitters that are partially synchronized (in time). According to an embodiment, RF-based tracking and positioning is implemented in a cellular network, but can also be implemented in any wireless system and RTLS environment. The proposed system can use software to implement digital signal processing and software-defined radio technology (SDR). Digital signal processing (DSP) can also be used.

One approach described herein utilizes clusters of receivers and/or transmitters, with precise time synchronization within each cluster, while inter-cluster time synchronization may be less precise or not required at all. Embodiments of the present invention can be used in all wireless systems/networks and encompass simplex, half-duplex, and full-duplex modes of operation. The following embodiments operate in wireless networks employing various modulation types, including OFDM modulation and/or its derivatives. Thus, the following embodiments operate in LTE networks and are also applicable to other wireless systems/networks.

As described in one embodiment, implementing RF-based tracking and positioning over a 3GPP LTE cellular network can significantly benefit from a cluster of receivers and/or transmitters that are precisely synchronized (in time).The proposed system can use software and/or hardware implemented digital signal processing.

The following description may set forth additional features and advantages of the present invention, and some of these features and advantages may be apparent from the description, or may be learned by practicing the embodiments. The advantages of the embodiments may be realized and attained by the structure particularly pointed out in the written description and the claims hereof as well as the accompanying drawings.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are intended to provide further explanation of the claimed embodiments.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are included to provide a further understanding of the embodiments and are incorporated in and constitute a part of this specification, illustrate the embodiments and together with the description serve to explain the principles of the embodiments. In the drawings:

1 and 1A illustrate narrow bandwidth ranging signal frequency components according to an embodiment;

FIG2 illustrates exemplary wide bandwidth ranging signal frequency components;

3A , 3B and 3C illustrate block diagrams of a master unit and a slave unit of an RF mobile tracking and positioning system according to an embodiment;

FIG4 illustrates an embodiment of synthesizing a wideband baseband ranging signal;

FIG5 illustrates the elimination of signal precursors by cancellation according to an embodiment;

FIG6 illustrates precursor cancellation with fewer carriers according to an embodiment;

FIG7 illustrates an embodiment of a one-way transfer function phase;

FIG8 illustrates an embodiment of a positioning method;

FIG9 illustrates LTE reference signal mapping;

FIG10 illustrates an embodiment of an enhanced cell ID+RTT positioning technique;

FIG11 illustrates an embodiment of the OTDOA positioning technology;

FIG12 illustrates the operation of a time observation unit (TMO) installed at an operator's eNB facility according to an embodiment;

FIG13 illustrates an embodiment of a wireless network positioning device diagram;

FIG14 illustrates an embodiment of a wireless network positioning downlink ecosystem for enterprise applications;

FIG15 illustrates an embodiment of a wireless network positioning downlink ecosystem for network wide applications;

FIG16 illustrates an embodiment of a wireless network positioning uplink ecosystem for enterprise applications;

FIG17 illustrates an embodiment of a wireless network positioning uplink ecosystem for network-wide applications;

FIG18 illustrates an embodiment of a UL-TDOA environment that may include one or more DAS and/or femto/small cell antennas;

FIG. 19 illustrates an embodiment of a UL-TDOA similar to that of FIG. 18 that may include one or more cell towers that can be used in place of DAS base stations and/or femto/small cells;

FIG20 illustrates an embodiment of cell-level positioning;

FIG21 illustrates an embodiment of cell and sector ID positioning;

FIG22 illustrates an embodiment of E-CID plus AoA positioning;

FIG23 illustrates an embodiment of AoA positioning;

FIG24 illustrates an embodiment of TDOA with wide yet close distances between receive antennas;

FIG25 illustrates an embodiment of a three-sector deployment;

FIG26 illustrates an embodiment of antenna port mapping;

FIG27 illustrates an embodiment of LTE Release 11 U-TDOA positioning technology;

FIG28 illustrates an embodiment of a high-level block diagram of a multi-channel location management unit (LMU);

FIG29 illustrates an embodiment of DL-OTDOA technology in a wireless/cellular network with a positioning server;

FIG30 illustrates an embodiment of U-TDOA technology in a wireless/cellular network with a location server;

FIG31 illustrates an embodiment depicting a rack-mount housing;

FIG32 illustrates an embodiment of a high-level block diagram of multiple single-channel LMUs clustered (consolidated) in a rack-mount enclosure;

FIG. 33 illustrates an embodiment of a high-level block diagram of multiple small cells with integrated LMUs (one-to-one antenna connection/mapping) clustered (integrated) in a rack-mounted housing; and

FIG34 illustrates an embodiment of a high-level block diagram of LMU and DAS integration.

FIG35 illustrates an embodiment of a high-level block diagram of LMU and WiFi infrastructure integration.

DETAILED DESCRIPTION

Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings.

Embodiments of the present invention relate to a method and system for RF-based identification, tracking, and location (including RTLS) of objects. According to an embodiment, the method and system utilize narrow-bandwidth ranging signals. The embodiment operates in the VHF band, but can also be used in the HF, LF, and VLF bands, as well as the UHF band and higher frequencies. It utilizes a multipath mitigation processor. The use of a multipath mitigation processor improves the accuracy of the tracking and location performed by the system.

Embodiments include small, very portable base units that allow users to track, locate, and monitor multiple people and objects. Each unit has its own ID. Each unit broadcasts an RF signal with its ID, and each unit is capable of sending back a return signal that may contain its ID as well as voice, data, and additional information. Each unit processes the return signals from the other units and, depending on triangulation or trilateral measurement and/or other methods used, continuously determines its relative and/or actual location. Preferred embodiments can also be easily integrated with products such as GPS devices, smart phones, two-way radios, and PDAs. The resulting product can have all the functions of a separate device while utilizing the existing display, sensors (such as altimeter, GPS, accelerometer, and compass) and processing capacity of its host computer. For example, a GPS device with the device technology described herein may be able to provide the user's location on a map and the location of other members of a mapping group.

As integrated circuit technology improves, the size of a preferred embodiment of an FPGA based implementation is between approximately 2 x 4 x 1 inches and 2 x 2 x 0.5 inches or less. Depending on the frequency used, the antenna may be integrated into the device or extend through the device housing. An ASIC (application specific integrated circuit) based version of the device may be able to incorporate the functionality of the FPGA and most other electronic components into the unit or accessory device. An ASIC based standalone version of the product may produce a device size of 1 x 0.5 x 0.5 inches or less. The antenna size may be determined by the frequency used and portions of the antenna may be integrated into the housing. The ASIC based embodiment is designed to be integrated into the product and may consist of just the chipset. There should not be any significant physical size difference between the main control unit or the accessory unit.

The device can process multipath mitigation algorithms using standard system components (off-the-shelf components) operating in multiple frequency ranges (bands). Software for digital signal processing and software-defined radios can be used. Signal processing software combined with minimal hardware allows for the integration of radios that transmit and receive software-defined waveforms.

U.S. Patent No. 7,561,048 discloses a narrow bandwidth ranging signal system whereby the narrow bandwidth ranging signal is designed to accommodate small bandwidth channels, such as voice channels that are only a few kilohertz wide (although some of these small bandwidth channels can extend to tens of kilohertz). This is in contrast to conventional position finding systems that use channels ranging from hundreds of kilohertz to tens of megahertz.

The advantages of this narrow-bandwidth ranging signal system are as follows: 1) At lower operating frequencies/bands, the ranging signal bandwidth of conventional location-finding systems exceeds the carrier (operating) frequency. Therefore, such systems cannot be deployed at LF/VLF and other lower frequency bands (including HF). Unlike conventional location-finding systems, the narrow-bandwidth ranging signal system described in U.S. Patent No. 7,561,048 can be successfully deployed at LF, VLF, and other frequency bands because its ranging signal bandwidth is much lower than the carrier frequency. 2) At the lower end of the RF spectrum (some VLF, LF, HF, and VHF bands), for example, up to the UHF band, conventional location-finding systems cannot be used because the FCC strictly limits the allowed channel bandwidth (12 to 25 kHz), which makes the use of conventional ranging signals impossible. Unlike conventional location-finding systems, the ranging signal bandwidth of the narrow-bandwidth ranging signal system is fully compatible with FCC regulations and other international spectrum regulatory agencies. Furthermore, 3) it is well known (see "MRI: The Basics," by Ray H. Hashemi and Mayiam G. Bradley, 2003) that, regardless of the operating frequency/band, narrow-bandwidth signals inherently have a higher signal-to-noise ratio (SNR) than wide-bandwidth signals. This increases the operating range of the narrow-bandwidth ranging signal location-finding system, regardless of its operating frequency/band, including the UHF band.

Therefore, unlike traditional location-finding systems, narrow-bandwidth ranging signal location-finding systems can be deployed at the lower end of the RF spectrum (e.g., VHF and lower frequency bands, down to the LF/VLF bands, where multipath is less pronounced). Furthermore, narrow-bandwidth ranging location-finding systems can also be deployed in the UHF band and higher, thereby improving the ranging signal noise ratio (SNR) and thus increasing the operating range of the location-finding system.

To minimize multipath (e.g., RF energy reflections), it is desirable to operate in the VLF/LF bands. However, at these frequencies, the efficiency of portable/mobile antennas is minimal (on the order of 0.1% or less due to the small antenna length (size) relative to the RF wave length). Additionally, at these lower frequencies, noise levels from natural and artificial sources are much higher than at higher frequencies/bands (e.g., VHF). These two phenomena can together limit the applicability of a location-finding system, e.g., its operating range and/or mobility/portability. Therefore, for some applications where operating range and/or mobility/portability are important, higher RF frequencies/bands, such as HF, VHF, UHF, and UWB, can be used.

At the VHF and UHF bands, the noise levels from natural and artificial sources are significantly lower than at the VLF, LF, and HF bands; and at VHF and HF frequencies, multipath phenomena (e.g., RF energy reflections) are less severe than at UHF and higher frequencies. In addition, at VHF, antenna efficiency is significantly better than at HF and lower frequencies, and at VHF, RF penetration is much better than at UHF. Therefore, the VHF band offers a good compromise for mobile/portable applications. On the other hand, in some specific cases, such as GPS where VHF frequencies (or lower frequencies) cannot penetrate the ionosphere (or become deflected/refracted), UHF can be a good choice. However, in any case (and all cases/applications), a narrow-bandwidth ranging signal system can have advantages over traditional wide-bandwidth ranging signal location-finding systems.

The actual application may determine the exact specifications (e.g., power, emissions, bandwidth, and operating frequency/band). Narrowband ranging allows users to receive a license or receive a license exemption, or use unlicensed bands as set forth in the FCC, because narrowband ranging allows operation over many different bandwidths/frequencies, including the most stringent narrow bandwidths set forth in the FCC and meeting the corresponding technical requirements of the appropriate section: 6.25kHz, 11.25kHz, 12.5kHz, 25kHz, and 50kHz. Therefore, multiple FCC sections and exemptions within such sections may be applicable. The primary FCC regulations that apply are: 47 CFR Part 90 - Private Land Mobile Radio Service, 47 CFR Part 94 - Personal Radio Services, and 47 CFR Part 15 - Radio Frequency Devices. (By comparison, wideband signals in this context are from hundreds of kHz up to 10-20MHz.)

Typically, for Part 90 and Part 94, VHF implementations allow users to operate devices up to 100 mW under certain exemptions (one example is low-power radio services). For some applications, the allowable transmit power at the VHF band is between 2 and 5 watts. For 900 MHz (UHF band), it is 1 W. At frequencies between 160 KHz and 190 KHz (LF band), the allowable transmit power is 1 watt.

Narrowband ranging can comply with many, if not all, different spectrum allowable ranges and allow for precise ranging while still meeting the most stringent regulatory requirements. This is not only valid for the FCC, but also for other international organizations that regulate spectrum usage around the world, including Europe, Japan, and South Korea.

Here is a list of common frequencies used, along with typical power consumption and the distance at which an accessory device can communicate with another reader in a real-world environment (see "Indoor Propagation and Wavelength," Dan Dobkin, WJ Communications, V 1.4 7/10/02):

915 MHz 100 mW 150 feet

2.4 GHz 100 mW 100 feet

5.6 GHz 100 mW 75 feet

The proposed system operates at VHF frequencies and employs a proprietary method to transmit and process RF signals. More precisely, it uses DSP technology and software-defined radio (SDR) to overcome the limitations of narrow bandwidth requirements at VHF frequencies.

Operating at a lower frequency (VHF) reduces dispersion and provides better wall penetration. The end result is a roughly tenfold increase in range relative to commonly used frequencies. For example, comparing the measured range of the prototype with the measured range of the RFID technology listed above:

216 MHz 100 mw 700 feet

Utilizing narrowband ranging technology, the range of commonly used frequencies and typical power consumption, as well as the distance at which an accessory device can communicate with another reader in a real-world environment, can be significantly increased:

Battery power consumption depends on the device design, transmit power, and duty cycle (e.g., the time interval between two consecutive distance (position) measurements). In many applications, the duty cycle is large, ranging from 10x to 1000x. In applications with a large duty cycle (e.g., 100x), an FPGA version transmitting 100 mW of power can have an operational life of approximately three weeks. An ASIC-based version can be expected to have an operational life of 10x. Furthermore, ASICs inherently have lower noise levels. Therefore, the operating range of an ASIC-based version can also be increased by approximately 40%.

Those skilled in the art will appreciate that embodiments do not compromise the long operating range of the system but significantly improve location-finding accuracy in RF-challenged environments (eg, buildings, urban corridors, etc.).

Typically, tracking and positioning systems employ tracking-positioning-navigation methods. These methods include time of arrival (TOA), time difference of arrival (DTOA), and a combination of TOA and DTOA. Time of arrival (TOA) as a range measurement technique is generally described in U.S. Patent No. 5,525,967. TOA/DTOA-based systems measure the direct line of sight (DLOS) time of flight of an RF ranging signal, e.g., time delay, which is then converted into a range.

In the case of RF reflections (e.g., multipath), multiple copies of the RF ranging signal with various delay times are superimposed on the DLOS RF ranging signal. Tracking and positioning systems using narrow-bandwidth ranging signals cannot distinguish the DLOS signal from reflected signals without multipath mitigation. As a result, these reflected signals cause errors in the estimated DLOS time of flight of the ranging signal, which in turn affects range estimation accuracy.

Embodiments advantageously utilize a multipath mitigation processor to separate DLOS signals from reflected signals. Consequently, embodiments significantly reduce errors in the estimated DLOS time-of-flight of ranging signals. The proposed multipath mitigation method can be used across all RF bands. It can also be used in wide-bandwidth ranging signal location-finding systems. Furthermore, it can support various modulation/demodulation techniques, including spread spectrum techniques such as DSS (direct spread spectrum) and FH (frequency hopping).

In addition, to further improve the accuracy of the method, noise reduction methods can be applied. These noise reduction methods may include (but are not limited to) coherent summation, incoherent summation, matched filtering, time diversity techniques, etc. By applying post-processing techniques such as maximum likelihood estimation (e.g., Viterbi algorithm) and minimum variance estimation (Kalman filtering), the residual multipath interference error can be further reduced.

Embodiments can be used in systems with simplex, half-duplex, and full-duplex operating modes. Full-duplex operation is very demanding in terms of complexity, cost, and arithmetic operations associated with the RF transceiver, which limits the operating range of the system in portable/mobile device implementations. In half-duplex operating mode, a reader (often referred to as a "master") and a slave device (sometimes also referred to as a "slave" or "target") are controlled by a protocol that allows either the master or the slave to transmit at any given time.

Alternating between transmission and reception allows a single frequency to be used for distance measurement. This arrangement reduces system cost and complexity compared to a full-duplex system. Simplex operation is conceptually simpler but requires stricter synchronization of events between the master and target units, including the start of the ranging signal sequence.

In an embodiment of the present invention, a narrow-bandwidth ranging signal multipath mitigation processor does not increase the ranging signal bandwidth. Advantageously, it utilizes different frequency components to allow propagation of the narrow-bandwidth ranging signal. Further ranging signal processing can be performed in the frequency domain by employing super-resolution spectrum estimation algorithms (MUSIC, rootMUSIC, ESPRIT) and/or statistical algorithms such as RELAX, or in the time domain by combining a synthetic ranging signal with a relatively large bandwidth and applying further processing to this signal. The different frequency components of the narrow-bandwidth ranging signal can be pseudo-randomly selected, can be contiguous or spaced apart in frequency, and can have uniform and/or non-uniform spacing in frequency.

The embodiment extends the multipath mitigation technique. The signal model for narrowband ranging is a complex exponential (as described elsewhere in this document) whose frequency is proportional to the delay defined by the range plus a similar term whose delay is defined by the time delay associated with multipath. The model is independent of the actual implementation of the signal structure (e.g., stepped frequency, linear frequency modulation, etc.).

The frequency separation between the direct path and multipath is nominally extremely small, and ordinary frequency-domain processing is insufficient to estimate the direct path range. For example, a stepped-frequency ranging signal with a 100 kHz step rate on 5 MHz at a range of 30 meters (100.07 nanoseconds of delay) produces a frequency of 0.062875 rad/s. Multipath reflections with a path length of 35 meters will produce a frequency of 0.073355. The separation is 0.0104792. The frequency resolution of a 50-sample observable has an inherent resolution of 0.12566 Hz. Therefore, it is impossible to use traditional frequency estimation techniques to separate the direct path from the reflected path and accurately estimate the direct path range.

To overcome this limitation, embodiments utilize a unique combination of subspace decomposition high-resolution spectral estimation methods and implementations of multimodal cluster analysis. Subspace decomposition techniques rely on partitioning the estimated covariance matrix of the observed data into two orthogonal subspaces: a noise subspace and a signal subspace. The theory behind subspace decomposition is that the projection of an observable onto the noise subspace consists of noise, and the projection of an observable onto the signal subspace consists of signal.

The super-resolution spectrum estimation algorithm and the RELAX algorithm are capable of distinguishing closely spaced frequencies (sine waves) in the spectrum even in the presence of noise. The frequencies do not need to be harmonically related, and unlike the digital Fourier transform (DFT), the signal model does not introduce any simulated cycles. For a given bandwidth, these algorithms provide significantly higher resolution than the Fourier transform. Therefore, direct line of sight (DLOS) can be reliably distinguished from other multipaths (MPs) with high accuracy. Similarly, applying the thresholding method described later to an artificially generated synthetic wide-bandwidth ranging signal makes it possible to reliably distinguish DLOS from other paths with high accuracy.

According to an embodiment, digital signal processing (DSP) can be employed by a multipath mitigation processor to reliably distinguish DLOS paths from other MP paths. Various super-resolution algorithms/techniques exist in spectrum analysis (spectrum estimation) technology. Examples include subspace-based methods such as the Multiple Signal Characterization (MUSIC) algorithm or the root-MUSIC algorithm, the Signal Parameter Estimation via Rotationally Invariant Technique (ESPRIT) algorithm, the Pisarenko Harmonic Decomposition (PHD) algorithm, and the RELAX algorithm.

In all the above super-resolution algorithms, the incoming (ie, received) signal is modeled as a linear combination of the complex exponential of the frequency and its complex amplitude. In the case of multipath, the received signal may be as follows:

where β×e ^i2πf×t is the transmitted signal, f is the operating frequency, L is the number of multipath components, and α and τ are the complex attenuation and propagation delay of the _Kth path, respectively. The multipath components are indexed so that propagation delays are considered in ascending order. Therefore, in this model, _τ represents the propagation delay of the DLOS path. Obviously, the value of _τ is of greatest interest, as it is the minimum of all _τ . The phase _θ is normally assumed to be random from one measurement cycle to another, with a uniform probability density function U(0,2π). Therefore, we assume _α =const (i.e., a constant value).

The parameters α _K and τ _K are random time-varying functions that reflect the movement of people and equipment in and around the building. However, since their rate of change is extremely slow compared to the measurement time interval, these parameters can be considered as time-varying random variables within a given measurement cycle.

All of these parameters are frequency dependent because they are related to radio signal characteristics (e.g., transmission and reflection coefficients). However, in embodiments, the operating frequency changes very little. Therefore, it can be assumed that the above parameters are frequency independent.

Equation (1) can be expressed in the frequency domain as:

Where: A(f) is the complex amplitude of the received signal, (2π×τ _K ) is the artificial “frequency” to be estimated by the super-resolution algorithm, and the operating frequency f is the independent variable; α _K is the Kth path amplitude.

In equation (2), (2π×τ _K ) and the subsequent super-resolution estimation of the τ _K value are based on continuous frequencies. In practice, there are a finite number of measurements. Therefore, the variable f may not be a continuous variable, but a discrete variable. Therefore, the complex amplitude A(f) can be calculated as follows:

where η is the discrete complex amplitude estimate (i.e., measurement) at the discrete frequency _fn .

In equation (3), can be interpreted as the amplitude and phase of a sinusoidal signal with frequency _fn after it has propagated through the multipath channel. It should be noted that all super-resolution algorithms based on spectrum estimation require complex input data (ie complex amplitudes).

In some cases, it is possible to convert the real signal data (e.g., ) into a complex signal (e.g., an analytical signal). For example, such a conversion can be accomplished using a Hilbert transform or other methods. However, at short distances, the value τ0 is extremely small, resulting in an extremely low (2π×τ _K ) "frequency."

These low "frequencies" pose a problem in the case of Hilbert transform (or other method) implementations. In addition, if amplitude values are to be used (e.g., ), the number of frequencies to be estimated may include not only (2π×τ _K ) "frequencies," but also combinations thereof. As a general rule, increasing the number of unknown frequencies affects the accuracy of the super-resolution algorithm. Therefore, reliable and accurate separation of DLOS paths from other multipath (MP) paths requires complex amplitude estimation.

The following is a description of the method and multipath mitigation processor operation during the task of obtaining complex amplitudes in the presence of multipath. It should be noted that while the description focuses on half-duplex operation, it can be easily extended to full-duplex operation. Simplex operation is a subset of half-duplex mode, but will require additional event synchronization.

In half-duplex operation, a reader (often called a "master") and an accessory (also called a "slave" or "target") can be controlled by a protocol that allows either the master or the slave to transmit at any given time. In this mode of operation, the accessory (target) acts as a transponder. The accessory receives ranging signals from the reader (master), stores them in memory, and then, after a certain time (delay), transmits them back to the master.

Examples of ranging signals are shown in FIG1 and FIG1A . The exemplary ranging signal utilizes contiguous frequency components. Other waveforms, including pseudo-random waveforms, that are spaced apart or orthogonal in frequency and/or time, may also be used, as long as the ranging signal bandwidth remains narrow. In FIG1 , the duration T _f of each frequency component may be long enough to achieve the narrow bandwidth property of the ranging signal.

FIG2 shows another variation of the ranging signal with different frequency components. It includes multiple frequencies ( _f1 , _f2 , _f3 , _f4 , _fn ) transmitted over a long period of time to form narrowband individual frequencies. This type of signal is more efficient, but it occupies a wider bandwidth, and wide-bandwidth ranging signals affect signal-to-noise ratio (SNR), which in turn reduces the operating range. Furthermore, such wide-bandwidth ranging signals may violate FCC requirements for VHF bands or lower frequency bands. However, in some applications, such wide-bandwidth ranging signals allow for easier integration into existing signal and transmission protocols. Furthermore, such signals reduce tracking-positioning time.

These multi-frequency (f ₁ , f ₂ , f ₃ , f ₄ , f _n ) bursts may also be contiguous and/or pseudo-random, spaced apart or orthogonal in frequency and/or time, and so on.

Compared to wideband ranging, the narrowband ranging mode can produce the accuracy of instantaneous wideband ranging while increasing the range over which this accuracy can be achieved. This performance is achieved because, at a fixed transmit power, the SNR at the receiver of the narrowband ranging signal (in the appropriate signal bandwidth) is greater than the SNR at the receiver of the wideband ranging signal. The SNR gain is approximately the ratio of the total bandwidth of the wideband ranging signal to the bandwidth of each channel of the narrowband ranging signal. This provides a good compromise when very fast ranging is not required (for example, for stationary and slow-moving targets, such as people walking or running).

The master and slave devices are identical and can operate in either master or transponder mode. All devices include a data/remote control communication channel. The devices can exchange information, and the master device can remotely control the slave devices. In this example, depicted in FIG1 , during operation of the master device (i.e., reader), the multipath mitigation processor initiates a ranging signal to the slave device, and after some delay, the master/reader receives a repeated ranging signal from the slave device.

The multipath mitigation processor of the master device then compares the received ranging signal with the ranging signal originally sent from the master device and determines an estimate in the form of the amplitude and phase of each frequency component _fn . It should be noted that in equation (3), it is defined for a one-way ranging signal trip. In an embodiment, the ranging signal makes a round trip. In other words, it travels in both directions: from the master device/reader to the target/slave device and from the target/slave device back to the master device/reader. Therefore, the complex amplitude of this round-trip signal received back by the master device can be calculated as follows:

and

Many techniques exist for estimating complex amplitude and phase values, including, for example, matched filtering. According to an embodiment, complex amplitude determination is based on values derived from RSSI (Received Signal Strength Indicator) values at the master and/or slave device receivers. Phase values are derived by comparing the phase of the returned baseband ranging signal received by the reader/master device with the phase of the original (i.e., transmitted by the reader/master device) baseband ranging signal. Furthermore, because the master and slave devices have independent clock systems, detailed explanation of device operation is enhanced by analyzing the impact of clock accuracy on phase estimation errors. As described above, one-way amplitude values can be obtained directly from the target/slave device. However, one-way phase values cannot be directly measured.

In an embodiment, the ranging baseband signal is the same as the ranging baseband signal depicted in FIG1 . However, for simplicity, the ranging baseband signal is shown here to consist of two frequency components, each containing multiple cycles of a cosine or sine wave of a different frequency: _F1 and _F2 . Note that _F1 = _f1 and _F2 = _f2 . The number of cycles in the first frequency component is L and the number of cycles in the second frequency component is P. Note that L may or may not be equal to P, since for _Tf = constant, each frequency component can have a different number of cycles. Furthermore, there is no time gap between each frequency component, and both _F1 and _F2 begin at an initial phase equal to zero.

Figures 3A, 3B, and 3C depict block diagrams of a master or slave unit (accessory) of an RF mobile tracking and location system. _FOSC refers to the frequency of the device's system clock (crystal oscillator 20 in Figure 3A). All frequencies generated within the device are generated by this system clock crystal oscillator. The following definitions are used: M is the master device (unit); AM is the accessory (target) device (unit). Accessory devices operate in transponder mode and are referred to as transponder (AM) units.

In a preferred embodiment, the device consists of an RF front-end and an RF back-end, a baseband, and a multipath mitigation processor. The RF back-end, baseband, and multipath mitigation processor are implemented in FPGA 150 (see Figures 3B and 3C). The system clock generator 20 (see Figure 3A) oscillates at: F _OSC = 20 MHz or ω _OSC = 2π × 20 × 10 ^6. This is an ideal frequency because the system clock frequency is not always equal to 20 MHz in actual devices:

It should be noted that

It should be noted that F _OSC frequencies other than 20 MHz can be used without any impact on system performance.

The electronic components of the two devices (master and slave) are identical, and the different operating modes are software programmable. Blocks 155 through 180 (see FIG. 3B ) generate a baseband ranging signal in digital format via the master device's FPGA 150. This signal consists of two frequency components, each containing multiple cycles of a cosine or sine wave at a different frequency. At the start, t=0, the master device's FPGA 150 ( FIG. 3B ) outputs the digital baseband ranging signal to its upconverter 50 via I/Q DACs 120 and 125. The FPGA 150 starts at frequency _F1 and, after time _T1, begins generating frequency _F2 for a duration of _T2 .

Because the frequency of the crystal oscillator may ^differ from 20 MHz, the actual frequencies generated by the FPGA may ^be _F1γM and _F2γM . Furthermore, time _T1 may be _T1βM ^and _T2 may _{be T2βM. It is also assumed that T1, T2, F1, and F2 are such that F1γM*T1βM=F1T1 and F2γM*T2βM=F2T2 , where F1T1} ^and _{F2T2 are integers .} ^This _means ^that _{the initial} ^phases _{of F1} ^and _{F2 are equal to zero .}

Since all frequencies are clocked from the system crystal oscillator 20, the baseband I/Q DACs 120 and 125 of the master device output as follows: where K _F1 and K _F2 are constant coefficients. Similarly, the output frequencies TX_LO and RX_LO from the frequency synthesizer 25 (the LO signals for mixers 50 and 85) can be expressed by constant coefficients. These constant coefficients are the same for the master device (M) and the transponder (AM) - the difference is the clock frequency of the system crystal oscillator 20 of each device.

The master (M) and transponder (AM) operate in half-duplex mode. The master's RF front-end uses a quadrature upconverter (i.e., mixer) 50 to upconvert the baseband ranging signal generated by the multipath mitigation processor and transmits this upconverted signal. After transmitting the baseband signal, the master switches from TX mode to RX mode using the RF front-end TX/RX switch 15. The transponder receives and downconverts the received signal back using its RF front-end mixer 85 (generating the first IF) and ADC 140 (generating the second IF).

This second IF signal is then digitally filtered in the transponder RF backend processor using a digital filter 190 and further down-converted to a baseband ranging signal using an RF backend quadrature mixer 200, digital I/Q filters 210 and 230, a digital quadrature oscillator 220, and a summer 270. This baseband ranging signal is stored in the transponder's memory 170 using a RAM data bus controller 195 and control logic 180.

The transponder then switches from RX to TX mode using the RF front-end switch 15 and begins retransmitting the stored baseband signal after a certain delay t _RTX . It should be noted that the delay is measured in the AM (transponder) system clock. Thus, the master receives the transponder transmission and downconverts the received signal back to baseband using its RF back-end quadrature mixer 200 , digital I and Q filters 210 and 230 , and digital quadrature oscillator 220 (see FIG3C ).

The master device then calculates the phase difference between _F1 and _F2 in the received (ie, recovered) baseband signal using the multipath mitigation processor arctangent block 250 and the phase comparison block 255. The amplitude value is derived from the RF backend RSSI block 240.

To improve the accuracy of the estimation, it is always necessary to improve the SNR of the amplitude estimate from block 240 and the phase difference estimate from block 255. In a preferred embodiment, the multipath mitigation processor calculates the amplitude and phase difference estimates for many time instances over the duration ( _Tf ) of the ranging signal frequency component. When averaged, these values can improve the SNR. The SNR improvement can be approximately proportional to N, where N is the number of instances when the amplitude and phase difference values are obtained (i.e., determined).

Another method of SNR improvement is to determine amplitude and phase difference values by applying a matched filtering technique over a period of time. Yet another method would be to estimate the phase and amplitude of the received (i.e., repeated) baseband ranging signal frequency components by sampling the received (i.e., repeated) baseband ranging signal frequency components with respect to the original (i.e., transmitted by the master/reader) baseband ranging signal frequency components in I/Q format and integrating over a period _T≤Tf . The integration has the effect of averaging multiple instances of the amplitude and phase in the I/Q format. The phase and amplitude values can then be converted from the I/Q format to the sum format.

Let us assume that at t=0, the master baseband processor (both in FPGA 150) begins a baseband ranging sequence under the control of the master's multipath processor.

Where T _f ≥ _{T 1} β ^M .

The phase at the output of the DACs 120 and 125 of the master device is as follows:

It should be noted that DACs 120 and 125 have internal propagation delays that are not dependent on the system clock.

Similarly, transmitter circuit components 15, 30, 40, and 50 may introduce additional delays that are not dependent on the system clock.

Therefore, the phase of the RF signal transmitted by the master device can be calculated as follows:

The RF signal from the master device (M) experiences a phase shift that depends on the multipath phenomenon between the master device and the slave devices.

The value depends on the transmit frequencies, for example, _F1 and _F2 . A transponder (AM) receiver cannot resolve every path due to the limited (i.e., narrow) bandwidth of the RF portion of the receiver. Therefore, after a certain time, for example, after 1 microsecond (equivalent to about 300 meters of flight), when all reflected signals have reached the receiver antenna, the following equation applies:

The output (e.g., first IF) in the AM (transponder) receiver at the first down-converter element 85, i.e., the phase of the signal is as follows:

It should be noted that the propagation delay in the receiver RF section (elements 15 and 60 to 85) does not depend on the system clock. After passing through the RF front-end filters and amplifiers (elements 95 to 110 and 125), the first IF signal is sampled by the RF back-end ADC 140. Assume that ADC 140 is downsampling the input signal (e.g., the first IF). Therefore, the ADC also acts as a downconverter to generate the second IF. The first IF filter, amplifier, and ADC add propagation delay time. At the ADC output (second IF):

In FPGA 150, the second IF signal (from the ADC output) is filtered by RF back-end digital filter 190 and further down-converted back to a baseband ranging signal by the third down-converter (i.e., quadrature mixer 200, digital filters 230 and 210, and digital quadrature oscillator 220), summed in summer 270, and stored in memory 170. At the output of the third down-converter (i.e., the quadrature mixer):

It should be noted that the propagation delay in the FIR section 190 does not depend on the system clock.

After the RX->TX delay, the baseband ranging signal stored (in memory 170) from the master (M) is retransmitted. Note that the RX->TX delay

By the time the signal from the transponder reaches the receiver antenna of the master (M), the RF signal from the transponder (AM) undergoes an additional phase shift that depends on the multipath. As discussed above, this phase shift occurs after a certain period of time when all reflected signals have reached the receiver antenna of the master:

In the master receiver, the signal from the transponder goes through the same down-conversion process as in the transponder receiver. The result is the recovered baseband ranging signal originally sent by the master.

For the first frequency component F ₁ :

For the second frequency component F2:

Substitution:

where _{TD_M-AM} is the propagation delay through the master (M) and transponder (AM) circuits.

where: is the LO phase shift from the mixer of the master (M) and transponder (AM) containing the ADC at time t=0.

Furthermore: _{KSYN_TX} = _{KSYN_RX_1} + _KADC + _{KSYN_RX_2}

The first frequency component F1:

2×10 ^-6 ＜t＜T ₁ β ^M +T _{D_M-AM} ;

The first frequency component F1 continues:

2×10 ^-6 ＜t＜T ₁ β ^M +T _{D_M-AM} ;

The second frequency component F2:

t＞T ₁ β ^M +T _{D_M-AM} +2×10 ^-6

The second frequency component F2, continues:

t＞T ₁ β ^M +T _{D_M-AM} +2×10 ^-6

Further substitution:

where α is a constant.

The final phase equation is:

2×10 ^-6 ＜t＜T ₁ β ^M +T _{D_M-AM} ;

t＞T ₁ β ^M +T _{D_M-AM} +2×10 ^-6 (5)

According to equation (5):

Where i＝2,3,4……………; and equal to

For example, the difference between time instances t1 and t2

2×10 ^-6 ＜t ₁ ＜T ₁ β ^M +T _{D_M-AM} ; t ₂ ＞T ₁ β ^M +T _{D_M-AM} +2×10 ^-6

To find the difference, we need to know T _{D — M-AM} : T _{D — M-AM} = T _{LB — M} β ^M + T _{LB — AM} β ^AM + t _RTX β ^AM ;

Where _{TLB_M} and _{TLB_AM} are the propagation delays through the master (M) and transponder (AM) TX and RX circuits measured by placing the devices in loopback mode. Note that the master and transponder devices can automatically measure _{TLB_M} and _{TLB_AM} ; and we also know the _tRTX value.

From the above formula and the value of t _RTX , T _{D — M-AM} can be determined, and thus for given t ₁ and t ₂ , the value can be found as follows:

2×10 ^-6 ＜t ₁ ＜T ₁ β ^M +T _{D_M-AM} ; t ₂ ＝t ₁ +T ₁ β ^M

2×10 ^-6 ＜t ₁ ＜T ₁ β ^M +T _{D_M-AM} ; t ₂ ＝t ₁ +T ₁ β ^M

2×10 ^-6 <t ₁ <T ₁ β ^M +T _{D_M-AM} ; t ₂ =t ₁ +T ₁ β ^M ; (6)

Or assuming β ^M = β ^AM = 1:

2×10 ^-6 <t ₁ <T ₁ +T _{D_M-AM} ; t ₂ =t ₁ +T ₁ ; (6A)

According to equation (6), it can be inferred that at the operating frequency, the complex amplitude value of the ranging signal can be obtained by processing the returned baseband ranging signal.

The initial phase value can be assumed to be equal to zero because the subspace algorithm is insensitive to constant phase shifts. If necessary, the value (phase initial value) can be found by determining the TOA (time of arrival) using the narrow bandwidth ranging signal method described in U.S. Patent No. 7,561,048, which is incorporated herein by reference in its entirety. This method estimates the ranging signal round-trip delay, which is equal to 2×T _FLT β ^M and the value can be found from the following equation:

or:

In a preferred embodiment, the returned baseband ranging signal phase value is calculated by the multipath processor's inverse tangent block 250. To improve the SNR, the multipath mitigation processor phase comparison block 255 uses equation (6A) to calculate for many instances n (n = 2, 3, 4 ...) and then averages them to improve the SNR. Note that: 2 x ^10-6 < _tn < _Tf + _TDM-AM ; _tm = _t1 + _Tf .

As can be seen from Equations 5 and 6, the recovered (i.e., received) baseband ranging signal has the same frequency as the original baseband signal transmitted by the master. Therefore, even though the master (M) and transponder (AM) system clocks may differ, there is no frequency translation. Because the baseband signal is composed of several frequency components, each consisting of multiple cycles of a sine wave, it is possible to estimate the phase and amplitude of the received ranging signal by sampling the frequencies of the individual components of the received baseband signal with the corresponding individual frequency components of the original (i.e., transmitted by the master) baseband signal and integrating the resulting signals over a period T ≤ T _f .

This operation produces complex amplitude values for the received ranging signal in I/Q format. Note that each individual frequency component of the baseband signal sent by the master must be shifted in time by T _{D — M-AM} . The integration operation has the effect of averaging multiple instances of amplitude and phase (e.g., improving SNR). Note that the phase and amplitude values can be converted from I/Q format to sum format.

This method of sampling, integration over a period of _T≤Tf , and subsequent conversion from I/Q format to Sum format can be implemented in the phase comparison block 255 in Figure 3C. Thus, depending on the design and implementation of block 255, either the method of the preferred embodiment based on equation (5) or the alternative method described in this section can be used.

Although the ranging signal bandwidth is narrow, the frequency difference _fn - _f1 can be relatively large, for example, on the order of several megahertz. Therefore, the receiver bandwidth must be kept wide enough to pass all _f1 : _fn ranging signal frequency components. This wide receiver bandwidth can affect the signal-to-noise ratio (SNR). To reduce the effective receiver bandwidth and improve the SNR, the RF backend processor in FPGA 150 can filter the received ranging signal baseband frequency components using digital narrow-bandwidth filters tuned for each individual frequency component of the received baseband ranging signal. However, these numerous digital filters (the number of filters equals the number of individual frequency components, n) place an additional burden on FPGA resources, increasing their cost, size, and power consumption.

In a preferred embodiment, only two narrow-bandwidth digital filters are used: one filter is always tuned for the _f1 frequency component, while the other filter can be tuned for all other frequency components _f2 : _fn . Multiple instances of the ranging signal are sent by the master device. Each instance consists of two frequencies: _f1 : _f2 ; _f1 : _f3 ...; _f1 : _fi ...; _f1 : _fn . Similar strategies are possible.

Note that it is also possible to keep the baseband ranging signal component to two (or even one), thereby generating the rest of the frequency components by adjusting the frequency synthesizer (e.g., changing K _SYN ). It may be necessary to use direct digital synthesis (DDS) technology to generate the LO signals for the up-converter and down-converter mixers. For high VHF band frequencies, this will place an undue burden on the transceiver/FPGA hardware. However, for lower frequencies, this may be an advantageous approach. An analog frequency synthesizer can also be used, but it may take extra time to stabilize after changing the frequency. In addition, in the case of an analog synthesizer, two measurements at the same frequency will have to be made to offset the phase offset that may occur after changing the frequency of the analog synthesizer.

The actual _{TD_M-AM} used in the above equation is measured in terms of both the master (M) and transponder (AM) system clocks, e.g., _{TLB_AM} and _tRTX are counted in terms of the transponder (AM) clock, and _{TLB_M} is counted in terms of the master (M) clock. However, when calculating, both _{TLB_AM} and _tRTX are measured (counted) in terms of the master (M) clock. This introduces an error:

The phase estimation error (7) affects the accuracy. Therefore, it may be necessary to minimize this error. If β ^M =β ^AM , in other words, all master and transponder (slave) system clocks are synchronized, then the contribution from the t _RTX time is eliminated.

In a preferred embodiment, the master device and transponder units (devices) are capable of synchronizing clocks with any device. For example, the master device can serve as a reference. Clock synchronization is accomplished using a remote control communication channel, whereby the frequency of the temperature-compensated crystal oscillator (TCXO) 20 is adjusted under the control of the FPGA 150. The frequency difference is measured at the output of the master device's summer 270 while the selected transponder device transmits a carrier signal.

The master then sends a command to the transponder to increase/decrease the TCXO frequency. This process can be repeated several times to achieve greater accuracy by minimizing the frequency at the output of summer 270. It should be noted that ideally, the frequency at the output of summer 270 should be equal to zero. An alternative approach is to measure the frequency difference and correct the estimated phase without adjusting the transponder's TCXO frequency.

While β ^M - β ^AM can be significantly reduced, phase estimation errors exist when β ^M ≠ 1. In this case, the error tolerance depends on the long-term stability of the clock generator of the reference device (usually the master device (M)). In addition, the clock synchronization process can take a considerable amount of time, especially when there are a large number of units in the field. During the synchronization process, the tracking-positioning system may become partially or completely inoperable, which negatively affects system readiness and performance. In this case, the above-mentioned method, which does not require TCXO frequency adjustment of the transponder, is preferred.

Commercially available (off-the-shelf) TCXO components are highly accurate and stable. Specifically, TCXO components used in commercial GPS applications are very precise. Using these devices, the phase error affecting positioning accuracy can be reduced to less than one meter without the need for frequent clock synchronization.

After the narrow-bandwidth ranging signal multipath mitigation processor obtains the complex amplitude of the returned narrow-bandwidth ranging signal, further processing (i.e., execution of the super-resolution algorithm) is performed in a software-based component as part of the multipath mitigation processor. This software component can be implemented in the host CPU of the master device (reader) and/or in a microprocessor (not shown) embedded in FPGA 150. In a preferred embodiment, the software component of the multipath mitigation algorithm is executed by the host CPU of the master device.

The super-resolution algorithm produces an estimate of the (2π×τ _K ) "frequency," ie, the value of τ _K. In a final step, the multipath mitigation processor selects the τ with the minimum value (ie, the DLOS delay time).

In some cases where the narrow bandwidth requirement of the ranging signal is slightly relaxed, the DLOS path can be separated from the MP path by using a continuous (in time) chirp. In a preferred embodiment, this continuous chirp is linear frequency modulation (LFM). Of course, other chirp waveforms can also be used.

Let's assume that a chirp with bandwidth B and duration T is transmitted under the control of a multipath mitigation processor. This gives a chirp rate of radians per second. Multiple chirps are transmitted and received back. Note that the chirp signal is generated digitally, with each chirp starting at the same phase.

In the multipath processor, each received single chirp is aligned so that the returned chirp is from the middle of the region of interest.

The linear frequency modulation pulse waveform equation is:

s( ^t )=exp(i(ω ₀ t+βt ² )), where ω ₀ is the initial frequency for 0<t<T.

For a single delay round trip τ, ie no multipath, the returned signal (cirp) is s(t-τ).

The multipath mitigation processor then "de-ramps" s(t-τ) by performing complex conjugate mixing with the originally transmitted chirp. The resulting signal is a complex sine wave:

f _τ (t)＝exp(-ω ₀ τ)exp(-2iβτt)exp(iβτ ² ), (8)

where exp(-iw ₀ τ _k ) is the amplitude and 2βτ is the frequency and 0≤t≤T. Note that the last term is the phase and it is negligible.

In the case of multipath, the de-ramped composite signal is composed of multiple complex sine waves:

where L is the number of ranging signal paths, including the DLOS path, and 0≤t≤T.

Multiple chirps are transmitted and processed. Each chirp is individually handled/processed as described above. The multipath mitigation processor then combines the results of the individual chirp processing:

where N is the number of chirps, ρ = T + t _dead ; t _dead is the dead time between two consecutive chirps; and 2βτ _k is the artificial delay “frequency.” Again, the lowest “frequency” is of most interest, corresponding to the DLOS path delay.

In equation (10), it can be viewed as the N samples of the sum of the complex sinusoids at the following times:

0≤t _α ≤T; t ₁ =t _α +ρ; t ₂ =t _α +2ρ....; t _m-1 =t _α +(N-1)ρ; m∈0:m-1;

Therefore, the number of samples can be a multiple of N, for example, αN; α=1, 2, .....

According to equation (10), the multipath mitigation processor generates αN complex amplitude samples in the time domain that are used in further processing (i.e., executing the super-resolution algorithm). This further processing is implemented in a software component that is part of the multipath mitigation processor. This software component can be executed by the master device (reader) host CPU and/or by a microprocessor (not shown) embedded in FPGA 150, or both. In a preferred embodiment, the multipath mitigation algorithm software is executed by the master device host CPU.

The super-resolution algorithm produces an estimate of the 2βτ _k "frequency", ie, the value of τ _K. In a final step, the multipath mitigation processor selects the τ with the minimum value (ie, the DLOS delay time).

A special processing method called "thresholding technique" can be explained as an alternative to super-resolution algorithms. In other words, it is used to improve the reliability and accuracy of distinguishing DLOS paths from MP paths using artificially generated synthetic wide bandwidth ranging signals.

The frequency domain baseband ranging signal shown in Figures 1 and 1A can be converted into a time domain baseband signal s(t):

It is easy to verify that s(t) is periodic with a period of 1/Δt, and for any integer k, s(k/Δt) = 2N + 1, which is the peak value of the signal, where n = N in Figures 1 and 1A.

Figure 4 shows two cycles of s(t) for the case of N = 11 and Δf = 250 kHz. The signal appears as a train of pulses of height 2N + 1 = 23 separated by 1/Δf = 4 microseconds. Between the pulses is a sine wave with varying amplitude and 2N zeros. The signal's wide bandwidth can be attributed to the narrowness of the high pulses. It can also be seen that the bandwidth extends from zero frequency to NΔf = 2.75 MHz.

The basic idea behind the threshold method used in the preferred embodiment is to improve the reliability and accuracy of artificially generated, synthetic, wide-bandwidth ranging measurements when distinguishing DLOS paths from other MP paths. The threshold method detects when the leading edge of a wideband pulse arrives at the receiver. Due to filtering in the transmitter and receiver, the leading edge does not rise instantaneously, but rather rises above the noise at a steadily increasing rate. The TOA of the leading edge is measured by detecting when the leading edge crosses a predetermined threshold, T.

A smaller threshold is desirable because it is crossed earlier, and the error delay τ between the true start of the pulse and the crossing is smaller. Therefore, any pulse replicas that arrive due to multipath have no effect if the replica's start has a delay greater than τ. However, the presence of noise imposes a limit on how small the threshold T can be. One way to reduce the delay τ is to use the derivative of the received pulse rather than the pulse itself, because the derivative rises faster. The second-order derivative has an even faster rise. Higher-order derivatives could be used, but in practice they can raise the noise level to unacceptable values, so thresholding the second-order derivative is used.

Although the 2.75 MHz wide signal depicted in FIG4 has an extremely wide bandwidth, it is not suitable for the measurement range of the above-described method. The method requires that each transmitted pulse has a zero signal precursor. However, it is possible to achieve this by modifying the signal so that the sine wave between pulses is substantially canceled. In a preferred embodiment, this is achieved by constructing a waveform that closely approximates the signal over a selected interval between high pulses and then subtracting it from the original signal.

The technique can be illustrated by applying it to the signal in Figure 1. The two black dots shown on the waveform are the endpoints of an interval I centered between the first two pulses. The left and right endpoints of interval I that have been experimentally determined to provide the best results are:

An attempt was made to generate a function g(t) that substantially cancels the signal s(t) over this interval, but without causing much harm outside of the interval. Since expression (11) indicates that s(t) is a sine sinπ(2N+1)Δft modulated by 1/sinπΔft, a function h(t) that closely approximates 1/sinπΔft over interval I is first found, and then g(t) is formed as the product:

g(t)＝h(t)sinπ(2N+1)Δft (13)

h(t) is generated by the sum of:

in

φ ₀ (t)≡1,φ _k (t)＝sinkπΔft for k＝1,2,...,M (15)

and the coefficients a _k are chosen to minimize the least square error on interval I

The solution is easily obtained by taking the partial derivative of J with respect to a _k and setting it equal to 0. The result is a linear system of M + 1 equations.

It can be solved for a _k , where

Then,

Using the definition of the function φ _k (t) given by (12)

Subtracting g(t) from s(t) yields a function r(t) that will substantially eliminate s(t) on interval I. As indicated in the Appendix, a suitable choice of the upper limit M for the summation in equation (20) is M = 2N + 1. Using this value and the results from the Appendix,

in

c＝-a ₀ (22)

From equation (17), we see that a total of 2N+3 frequencies (including the zero-frequency DC term) are required to obtain the desired signal r(t). FIG5 shows the desired signal r(t) for the original signal s(t) shown in FIG1 , where N=11. In this case, the construction of r(t) requires 25 carriers (including the DC term b ₀ ).

The important properties of r(t) constructed above are as follows:

1. The lowest frequency is 0 Hz and the highest frequency is (2N+1)Δf Hz, as seen from (14). Therefore, the total bandwidth is (2N+1)Δf Hz.

2. All carriers are cosine functions (including DC) spaced Δf apart, except one carrier which is a sine function at frequency.

3. While the original signal s(t) has a period of 1/Δf, r(t) has a period of 2/Δf. The first half of each cycle of r(t), which is a complete cycle of s(t), contains the canceled portion of the signal, and the second half of the cycle of r(t) is a large oscillation fragment. Therefore, the cancellation of the precursor can occur in every other cycle of s(t).

This occurs because the cancellation function g(t) actually strengthens s(t) in every other cycle of s(t). One reason for this may be that g(t) reverses its polarity at each peak of s(t), while s(t) does not. An example of a method for increasing the processing gain by 3 dB by making every cycle of s(t) contain a cancelled portion is described below.

4. The length of the cancelled portion of s(t) is about 80-90% of 1/Δf. Therefore, Δf needs to be small enough so that this length is long enough to cancel any residual signal from the previous non-zero portion of r(t) due to multipath.

5. Immediately following each zero portion of r(t) is the first period of the oscillatory portion. In a preferred embodiment, in the TOA measurement method described above, the first half of this period is used to measure TOA, specifically, the beginning of its rise. It is interesting to note that the peak of this first half-period (which can be called the main peak) is slightly larger than the corresponding peak of s(t) at approximately the same point in time. The width of this first half-period is roughly inversely proportional to NΔf.

6. Substantial processing gains can be achieved by:

(a) Using repetition of the signal r(t), since r(t) is periodic with a period of 2/Δf, an additional 3 dB of processing gain is possible by the method described later.

(b) Narrowband filtering: Because each of the 2N+3 carriers is a narrowband signal, the occupied bandwidth of the signal is much smaller than the bandwidth of a wideband signal that extends across the entire allocated frequency band.

For the signal r(t) shown in FIG5 , where N=11 and Δf=250 kHz, the length of the cancelled portion of s(t) is approximately 3.7 microseconds or 1,110 meters. This is more than sufficient to cancel any residual signal from the previously non-zero portion of r(t) due to multipath. The main peak has a value of approximately 35, and the maximum magnitude in the precursor (i.e., cancellation) region is approximately 0.02, which is 65 dB lower than the main peak. This is desirable for achieving good performance using the TOA measurement thresholding technique described above.

The use of fewer carriers is depicted in FIG6 , which illustrates a signal generated using Δf = 850 kHz, N = 3, and M = 2N + 1 = 7, for a total of 2N + 3 = 9 carriers. In this case, the signal has a period of microseconds, compared to the signal in FIG5 , which has a period of 8 microseconds. Since this example has more periods per unit time, more processing gain can be expected.

However, since fewer carriers are used, the amplitude of the main peak is about 1/3 as large as before, which tends to offset the expected additional processing gain. In addition, the length of the zero-signal precursor segment is shorter, about 0.8 microseconds or 240 meters. This should still be sufficient to cancel any residual signal caused by multipath in the previously non-zero portion of r(t). It should be noted that the total bandwidth of (2N+1)Δf=5.95MHz is about the same as before, and the width of half a cycle of the main peak is also about the same. Since fewer carriers are used, there should be some additional processing gain when each carrier is narrow-band filtered at the receiver. In addition, the maximum magnitude in the precursor (i.e., cancellation) region is now about 75dB lower than the main peak, an improvement of 10dB from the previous example.

Transmission at RF Frequency: So far, for simplicity, r(t) has been described as a baseband signal. However, it can be converted to RF, transmitted, received, and then reconstructed as a baseband signal at the receiver. To illustrate, consider what happens to one of the frequency components ω _k in the baseband signal r(t) traveling through one of the multipath propagation paths with index j (using degrees/second frequency for simplicity of notation):

b _k cos ω _k t (in transmitter at baseband)

b _k cos(ω+ω _k )t (transformed by frequency ω until RF)

a _j b _k cos[(ω+ω _k )(t-τ _j )+φ _j ] (at the receiver antenna)

a _j b _k cos[ω _k (t-τ _j )+φ _j +θ] (converted to baseband via frequency-ω) (23)

It is assumed here that the transmitter and receiver are frequency synchronized. The parameter _bk is the kth coefficient in expression (21) for r(t). The parameters _τj and _φj are the path delay and phase shift (due to the dielectric properties of the reflector) for the jth propagation path, respectively. The parameter θ is the phase shift that occurs during downconversion to baseband in the receiver. A similar sequence of functions can exist for the sinusoidal components of equation (21).

It is important to note that as long as the zero-signal precursor in r(t) has a length sufficiently greater than the maximum effective propagation delay, the final baseband signal in equation (20) can still have a zero-signal precursor. Of course, when all frequency components (index k) on all paths (index .j) are combined, the baseband signal at the receiver can be a distorted version of r(t), including all phase shifts.

Figures 1 and 1A illustrate sequential carrier transmission and signal reconstruction. Assuming the transmitter and receiver are time and frequency synchronized, the 2N+3 transmit carriers do not necessarily need to be transmitted simultaneously. For example, consider the transmission of the signal whose baseband representation is shown in Figures 1A and 6.

In FIG6 , N=3, and it is assumed that each of the nine frequency components is transmitted sequentially for 1 millisecond. The start and end times of each frequency transmission are known at the receiver, so it can sequentially start and end its reception of each frequency component at those corresponding times. Since the signal propagation time is very short compared to 1 millisecond (it may typically be less than a few microseconds in the intended application), a small portion of each received frequency component should be ignored and can be easily canceled by the receiver.

The entire process of receiving the nine frequency components can be repeated in additional received 9-millisecond blocks to increase processing gain. In one second of total receive time, there will be approximately 111 such 9-millisecond blocks available for processing gain. Additionally, within each block, there will be additional processing gain available from the main peak.

It is worth noting that, in general, signal reconstruction can be done in a very economical way and inherently allows for all possible processing gains. For each of the 2N+3 receive frequencies:

1. Measure the phase and amplitude of each 1 millisecond reception of the frequency to form a series of stored vectors (phasors) corresponding to the frequency.

2. Average the frequency storage vectors.

3. Finally, use the 2N+3 vector averages of the 2N+3 frequencies to reconstruct 1 period of the baseband signal with duration 2/Δf, and use the reconstruction to estimate the signal TOA.

This method is not limited to 1 millisecond transmissions, and the length of the transmissions can be increased or decreased. However, the total time of all transmissions should be short enough to stop any action of the receiver or transmitter.

Obtaining cancellation on alternate half-cycles of r(t): By simply reversing the polarity of the cancellation function g(t), cancellation between the peaks of s(t) is possible, where r(t) was previously oscillating. However, in one embodiment, to obtain cancellation between all peaks of s(t), the function g(t) and its polarity-reversed version must be applied at the receiver, and this involves coefficient weighting at the receiver.

Weighting of coefficients at the receiver: The coefficients _bk in equation (21) are used in the construction of r(t) at the transmitter, and can be introduced at the receiver instead, if necessary. This is easily seen by considering a series of signals in equation (20), where the final signal is the same when _bk is introduced at the end rather than at the beginning. Ignoring noise, the values are as follows:

cosω _k t (in transmitter at baseband)

cos(ω+ω _k )t (converted by frequency ω until RF)

a _j cos[(ω+ω _k )(t-τ _j )+φ _j ] (at the receiver antenna)

a _j cos[ω _k (t-τ _j )+φ _j +θ] (converted to baseband via frequency-ω)

a _j b _k cos[ω _k (t-τ _j )+φ _j +θ] (weighted by coefficient b _k at baseband) (24)

The transmitter can then transmit all frequencies with the same amplitude, which simplifies its design. Note that this approach also weights the noise at each frequency, and its contribution should be taken into account. Also note that the weighting coefficients should be applied at the receiver to achieve the opposite polarity of g(t) and thus obtain twice the usable main peak.

Adjustment of Δf to the center frequency in the channel: To meet FCC requirements, channelized transmission with constant channel spacing may be required at VHF and lower frequencies. In channelized transmit bands with constant channel spacing that are small compared to the total allocated frequency band (i.e., for VHF and lower frequency bands), small adjustments to Δf, if necessary, allow all transmit frequencies to be centered in the channel without substantially changing performance from the original design values. In the two examples of baseband signals previously presented, all frequency components are multiples of Δf/2, so if the channel spacing is divided by Δf/2, the lowest RF transmit frequency can be centered in one channel, with all other frequencies falling in the channel center.

In some radio frequency (RF) based identification, the tracking and location system has both the master unit and the slave unit perform voice, data, and control communication functions in addition to the distance measurement function. Similarly, in a preferred embodiment, both the master unit and the slave unit also perform voice, data, and control communication functions in addition to the distance measurement function.

According to a preferred embodiment, a number of sophisticated signal processing techniques, including multipath mitigation, are performed on the ranging signals. However, these techniques may not be applicable to voice, data, and control signals. Consequently, the operating range of the proposed system (and other existing systems) may be limited not by its ability to reliably and accurately measure distance, but by being out of range during voice and/or data and/or control communications.

In other radio frequency (RF)-based identification systems, the distance measurement function of the tracking and location system is separated from the voice, data, and control communication functions. In these systems, separate RF transceivers are used to perform voice, data, and control communication functions. The disadvantage of this approach is the increased cost, complexity, and size of the system.

To avoid these drawbacks, in a preferred embodiment, several individual frequency components of a narrow-bandwidth ranging signal, or baseband narrow-bandwidth ranging signal, are modulated with the same data/control signal and, in the case of voice, with digitized voice packet data. At the receiver, the individual frequency component with the highest signal strength is demodulated, and the reliability of the resulting information can be further improved by performing "voting" or other signal processing techniques that exploit information redundancy.

This approach allows avoiding the "null" phenomenon, in which incoming RF signals from multiple paths destructively combine with the DLOS path and each other, thereby significantly reducing the received signal strength and its associated SNR. Furthermore, such an approach allows finding a set of frequencies at which incoming signals from multiple paths constructively combine with the DLOS path and each other, thereby increasing the received signal strength and its associated SNR.

As mentioned previously, super-resolution algorithms based on spectrum estimation generally use the same model: a linear combination of the complex exponential of the frequency and its complex amplitude. This complex amplitude is given by Equation 3 above.

All super-resolution algorithms based on spectrum estimation require prior knowledge of the number of complex exponentials, i.e., the number of multipath paths. This number of complex exponentials is called the model size and is determined by the number of multipath components L, as shown in Equations 1 to 3. However, when estimating path delays (i.e., for RF tracking-positioning applications), this information is not available. This adds another dimension to the spectrum estimation process via super-resolution algorithms, i.e., the model size estimation.

It has been demonstrated (Kei Sakaguchi et al., "Influence of the Model Order Estimation Error in the ESPRIT Based High Resolution Techniques") that underestimating the model size affects the accuracy of frequency estimation, while overestimating the model size can lead to the generation of spurious frequencies, i.e., nonexistent frequencies. Existing methods for model size estimation, such as AIC (Akaikes Information Criterion) and MDL (Minimum Description Length), are highly sensitive to correlation (complex exponential) between signals. However, this is often the case with RF multipath. Even after applying a forward-backward smoothing algorithm, for example, residual correlation may still be present.

In the Sakaguchi paper, it is recommended to use an overestimated model and distinguish between the actual frequency (signal) and the pseudo-frequency (signal) by estimating their power (amplitude) and then rejecting signals with very low power. Although this method is an improvement over existing methods, it is not guaranteed. The present inventors implemented the method of Kei Sakaguchi et al. and simulated more complex cases with larger model sizes. It was observed that in some cases, the amplitude of the pseudo-signal can be very close to the amplitude of the actual signal.

All super-resolution algorithms based on spectrum estimation work by dividing the incoming signal complex amplitude data into two subspaces: the noise subspace and the signal subspace. If these subspaces are defined (separated) appropriately, then the model size is equal to the signal subspace size (dimension).

In one embodiment, model size estimation is done using the "F" statistic. For example, for the ESPRIT algorithm, a singular value decomposition of the estimate of the covariance matrix (with forward/backward correlation smoothing) is performed in ascending order. Afterwards, a division is performed, dividing the (n+1) eigenvalue by the nth eigenvalue. This ratio is the "F" random variable. The worst case is an "F" random variable with (1,1) degrees of freedom. The 95% confidence interval for the "F" random variable with (1,1) degrees of freedom is 161. Setting this value as a threshold determines the model size. It should also be noted that for the noise subspace, the eigenvalue represents an estimate of the noise power.

This method of applying the "F" statistic to the ratio of the eigenvalues is a more accurate way to estimate model size. It should be noted that the other degrees of freedom in the "F" statistic can also be used for threshold calculation and therefore for model size estimation.

Nevertheless, in some cases, due to imperfect real-world measurements, two or more very close (in time) signals can be simplified into a single signal. Therefore, the above method may underestimate the number of signals, that is, the model size. Since underestimating the model size will reduce the accuracy of the frequency estimate, it should be cautious to increase the model size by adding a certain number. This number can be determined experimentally and/or based on simulations. However, when the signals are not close together, the model size may be overestimated.

In such cases, spurious (i.e., non-existent) frequencies may appear. As mentioned earlier, using signal amplitude for spurious signal detection is not always effective because in some cases spurious signals are observed to have amplitudes very close to the actual signal amplitude. Therefore, in addition to amplitude discrimination, filters can be implemented to improve the probability of spurious frequency elimination.

The frequencies estimated by the super-resolution algorithm are simulated frequencies (Equation 2). In reality, these frequencies are the individual path delays of the multipath environment. Therefore, negative frequencies should not exist, and all negative frequencies generated by the super-resolution algorithm are spurious frequencies to be rejected.

Alternatively, methods other than super-resolution can be used to estimate the DLOS range from the complex amplitude values obtained during the measurement. Although these methods have lower accuracy, they can establish a range for distinguishing delays, i.e., frequencies. For example, the ratio

In the Δf interval, where the signal amplitude is close to the maximum, i.e., avoiding nulls, a DLOS delay range is provided. Although the actual DLOS delay can be up to a factor of two larger or smaller, this defines a range that helps suppress spurious results.

In an embodiment, the ranging signal makes a round trip. In other words, it travels in both directions: from the master/reader to the target/slave and from the target/slave back to the master/reader.

The master device transmits a tone: α×e ^−jωt , where ω is the operating frequency in the operating band and α is the tone signal amplitude.

At the target's receiver, the received signal (one-way) is as follows:

Where: N is the number of signal paths in a multipath environment; K0 and _τ0 are the amplitude and flight time of the DLOS signal; | _K0 |=1, _K0 >0, | _Km≠0 |≤1 and Km _≠0 can be positive or negative.

S _one-way (t)＝α×e ^-jωt ×A(ω)×e ^-jθ(ω) (26)

where: is the one-way multipath RF channel transfer function in the frequency domain; and A(ω)≥0.

The target device retransmits the received signal:

S _retransmit (t)＝α×e ^-jωt ×A(ω)×e ^-jθ(ω) (27)

At the master receiver, the round-trip signal is:

or:

S _{round_trip} (t)＝α×e ^-jωt ×A ² (ω)×e ^-j2θ(ω) (28)

On the other hand, according to equations (26) and (28):

where: is the round-trip multipath RF channel transfer function in the frequency domain.

According to Equation 29, a round-trip multipath channel has a larger number of paths than a unidirectional channel multipath because in addition to the τ ₀ ÷ τ _N path delay, the expression also includes combinations of these path delays, such as: τ ₀ + τ ₁ , τ ₀ + τ ₂ , τ ₁ + τ ₂ , τ ₁ + τ ₃ , etc.

These combinations significantly increase the number of signals (complex exponential). Therefore, the probability of very dense (in time) signals may also increase and may lead to an underestimate of the effective model size. Therefore, it is desirable to obtain a unidirectional multipath RF channel transfer function.

In a preferred embodiment, the one-way amplitude value is directly available from the target/slave device. However, the one-way phase value cannot be measured directly. It is possible to determine the one-way phase from the round-trip phase measurement observation:

and

However, for each value of ω, there are two values of phase α(ω) such that

e ^j2α(ω) =e ^jβ(ω)

A detailed description of resolving this ambiguity is presented below. If the different frequency components of the ranging signal are close to each other, then for the most part, the one-way phase can be found by dividing the round-trip phase by two. Exceptions may include regions near the "null," where the phase can undergo significant changes even with small frequency steps. It should be noted that the "null" phenomenon is where incoming RF signals from multiple paths destructively combine with the DLOS path and each other, thus significantly reducing the received signal strength and its associated SNR.

Assume h(t) is the one-way impulse response of the communication channel. The corresponding transfer function in the frequency domain is

where A(ω)≥0 is the magnitude and α(ω) is the phase of the transfer function. If a one-way impulse response is retransmitted back through the same channel it was received, then the resulting two-way transfer function is

G(ω)＝B(ω)e ^jβ(ω) =H ² (ω)＝A ² (ω)e ^j2α(ω) (31)

where B(ω)≥0. Assume that the two-way transfer function G(ω) is known for all ω in some open frequency interval (ω ₁ ,ω ₂ ). Is it possible to determine the one-way transfer function H(ω) bounded on (ω ₁ ,ω ₂ ) that produces G(ω)?

Since the magnitude of the bidirectional transfer function is the square of the single vector value, it is clear that

However, when trying to recover the phase of the one-way transfer function from observations of G(ω), the situation is more subtle. For each value of ω, there are two values of the phase α(ω) such that

e ^j2α(ω) =e ^jβ(ω) (33)

A large number of different solutions may be generated by independently selecting one of the two possible phase values for each different frequency ω.

The following theorem, assuming that any one-way transfer function is continuous at all frequencies, helps resolve this situation.

Theorem 1: Assume I is an open interval of frequency ω that contains no zeros of the two-way transfer function G(ω) = B(ω)e ^jβ(ω) . Assume is a continuous function on I with β(ω) = 2γ(ω). Then J(ω) and -J(ω) are one-way transfer functions on I that generate G(ω), and there are no other one-way transfer functions.

Proof: One solution of the one-way transfer function is that the function is continuous on I, because it is differentiable on I, and where β(ω) = 2α(ω). Since G(ω) ≠ 0 on I, H(ω) and J(ω) are non-zero on I. Then,

Since H(ω) and J(ω) are continuous and non-zero on I, their ratio is continuous on I, and therefore the right-hand side of (34) is continuous on I. The condition β(ω) = 2α(ω) = 2γ(ω) implies that for each ω∈I, α(ω) - γ(ω) is either 0 or π. However, α(ω) - γ(ω) cannot switch between these two values without causing a discontinuity on the right-hand side of (34). Therefore, for all ω∈I, α(ω) - γ(ω) = 0, or for all ω∈I, α(ω) - γ(ω) = π. In the first case, we have J(ω) = H(ω), and in the second case, we have J(ω) = -H(ω).

This theorem proves that to obtain a one-way solution G(ω) = B(ω)e ^jβ(ω) on any open interval I that does not contain zeros of the transfer function, we form a function to choose values of γ(ω) such that β(ω) = 2γ(ω) so that J(ω) is continuous. This is always possible because a solution with this property, H(ω), is known to exist.

The alternative procedure for arriving at a one-way solution is based on the following theorem:

Theorem 2: Assume that H(ω)=A(ω)e ^jα(ω) is a one-way transfer function, and assume that I is an open interval of frequency ω that does not contain zeros of H(ω). In one embodiment, the phase function α(ω) of H(ω) must be continuous on I.

Proof : Assume ω ₀ to be a frequency in interval I. In Figure 7 , the complex values H(ω ₀ ) have been plotted as points in the complex plane, and by assumption, H(ω ₀ ) ≠ 0. Assume ε > 0 to be an arbitrarily small real number, and consider the two angles of index ε shown in Figure 7 , and the circle centered at H(ω ₀ ) and tangent to the two rays OA and OB. By assumption, H(ω) is continuous for all ω. Therefore, if ω is sufficiently close to ω ₀ , then the complex value H(ω) can lie in the circle, and it can be seen that |α(ω) - α(ω ₀ )| < ε. Since ε > 0 is arbitrarily chosen, we conclude that as ω → ω ₀ , α(ω) → α(ω ₀ ), such that the phase function α(ω) is continuous at ω ₀ .

Theorem 3: Assume I is an open interval of frequency ω that does not contain zeros of the two-way transfer function G(ω)=B(ω)e ^jβ(ω) . Assume is a function on I where β(ω)=2γ(ω) and γ(ω) is continuous on I. Then J(ω) and -J(ω) are one-way transfer functions that generate G(ω) on I, and there are no other one-way transfer functions.

Proof : The proof is similar to that of Theorem 1. We know that a solution of the one-way transfer function is the function where β(ω) = 2α(ω). Since G(ω) ≠ 0 on I, H(ω) and J(ω) are non-zero on I. Then,

By assumption, γ(ω) is continuous on I, and by Theorem 2, α(ω) is also continuous on I. Therefore, α(ω)-γ(ω) is continuous on I. The condition β(ω)=2α(ω)=2γ(ω) implies that for every ω∈I, α(ω)-γ(ω) is either 0 or π. However, α(ω)-γ(ω) cannot switch between these two values without becoming discontinuous on I. Therefore, for all ω∈I, α(ω)-γ(ω)=0, or for all ω∈I, α(ω)-γ(ω)=π. In the first case, we have J(ω)=H(ω), and in the second case, J(ω)=-H(ω).

Theorem 3 tells us that to obtain a one-way solution on any open interval I that does not contain zeros of the transfer function G(ω) = B(ω)e ^jβ(ω) , we simply form a function to choose values of γ(ω) such that β(ω) = 2γ(ω) so that the phase function γ(ω) is continuous. This is always possible because a solution with this property, namely H(ω), is known to exist.

While the above theorems show how to reconstruct the two one-way transfer functions that produce the two-way function G(ω), they are useful on frequency intervals I that do not contain zeros of G(ω). In principle, G(ω) can be observed on frequency intervals (ω ₁ ,ω ₂ ) that may contain zeros. The following is a possible way to circumvent this problem, assuming that there are a finite number of zeros of G(ω) in (ω ₁ ,ω ₂ ) and that the one-way transfer function has derivatives of all orders on (ω ₁ ,ω ₂ ), not all of which are zero at any given frequency ω:

Assume that H(ω) is a one-way function that produces G(ω) on the interval (ω ₁ ,ω ₂ ), and assume that G(ω) has at least one zero on (ω ₁ ,ω ₂ ). The zeros of G(ω) can separate (ω ₁ ,ω ₂ ) into a finite number of adjacent open frequency intervals J ₁ ,J ₂ ,...,J _n . On each such interval, the solution H(ω) or -H(ω) can be found using Theorem 1 or Theorem 3. We need to "stitch these solutions together" so that the stitched solution is either H(ω) or -H(ω) across all (ω ₁ ,ω ₂ ). To do this, we need to know how to pair solutions in two adjacent subintervals so that we do not stitch from H(ω) to -H(ω) or from -H(ω) to H(ω) when moving from one subinterval to the next.

We illustrate the stitching procedure starting with two adjacent open subintervals, J ₁ and J _2. These subintervals can be contiguous at the frequency ω ₁ , which is a zero of G(ω) (of course, ω ₁ is not contained in either subinterval). By our above assumptions about the properties of one-way transfer functions, there must exist a minimum positive integer n such that H ⁽ⁿ⁾ (ω ₁ ) ≠ 0, where the superscript (n) denotes the nth-order derivative. Then, depending on whether our solution in J ₁ is H(ω) or -H(ω), the restriction on the nth-order derivative of our one-way solution in J ₁ as ω → ω ₁ from the left can be H ⁽ⁿ⁾ (ω ₁ ) or -H ⁽ⁿ⁾ (ω ₁ ). Similarly, depending on whether our solution in J ₂ is H(ω) or -H(ω), the restriction on the nth-order derivative of our one-way solution in J ₂ as ω → ω ₁ from the right can be H ⁽ⁿ⁾ (ω ₁ ) or -H ⁽ⁿ⁾ (ω ₁ ). Since H ⁽ⁿ⁾ ( _ω1 )≠0, the two constraints can be equal if the solutions in _J1 and _J2 are both H(ω) or both -H(ω). If the left and right constraints are not equal, then we invert the solution in subinterval _J2 . Otherwise we do not do so.

After inverting the solution (if necessary) in subinterval J ₂ , we perform the same procedure for subintervals J ₂ and J ₃ , inverting the solution (if necessary) in subinterval J _3. Continuing in this manner, we eventually establish a complete solution on the interval (ω ₁ ,ω ₂ ).

It would be desirable not to require the higher order derivatives of H(ω) in the above reconstruction procedure, since they are difficult to compute accurately in the presence of noise. This problem is unlikely to occur, since at any zero of G(ω), it seems likely that the first order derivative of H(ω) may be non-zero, and if not, then it is likely that the second order derivative may be non-zero.

In a practical approach, the bidirectional transfer function G(ω) may be measured at discrete frequencies that must be close enough to enable reasonably accurate calculation of the derivative of G(ω) about zero.

For RF-based distance measurement, it may be necessary to resolve an unknown number of extremely dense, overlapping, and noisy echoes of a ranging signal with an a priori known shape. Assuming the ranging signal is narrowband, in the frequency domain, this RF phenomenon can be described (modeled) as the sum of multiple sinusoids, each of which is based on a multipath component and each with a complex attenuation and propagation delay of the path.

The Fourier transform of the above sum can express this multipath model in the time domain. By exchanging the roles of the time and frequency variables in this time domain expression, this multipath model can be transformed into a harmonic signal spectrum, where the propagation delays of the paths are transformed into harmonic signals.

Ultra-high resolution spectrum estimation methods are designed to identify closely spaced frequencies in a spectrum and are used to estimate individual frequencies of multiple harmonic signals, such as path delays. Therefore, path delays can be accurately estimated.

Super-resolution spectrum estimation exploits the eigenstructure and inherent properties of the covariance matrix of the baseband ranging signal samples to provide a solution for the underlying estimate of individual frequencies (e.g., path delays). One of the eigenstructure properties is that the eigenvalues can be combined and thus partitioned into orthogonal noise and signal eigenvectors, also known as subspaces. Another eigenstructure property is the rotationally invariant signal subspace property.

Subspace decomposition techniques (MUSIC, rootMUSIC, ESPRIT, etc.) rely on splitting the estimated covariance matrix of the observed data into two orthogonal subspaces: the noise subspace and the signal subspace. The theory behind subspace decomposition methods is that the projection of the observable onto the noise subspace consists of the noise, and the projection of the observable onto the signal subspace consists of the signal.

Spectral estimation methods assume that the signal is narrowband and the number of harmonic signals is also known, that is, the size of the signal subspace needs to be known. The size of the signal subspace is called the model size. In general, any details of it cannot be known and can change rapidly with changes in the environment (particularly, indoors). One of the most difficult and detailed issues when applying any subspace decomposition algorithm is the size of the signal subspace, which can be regarded as the number of frequency components present, and it is the number of multipath reflections plus direct paths. Due to real-world measurement defects, there may always be errors in the model size estimation, which may in turn lead to the loss of accuracy of frequency estimation (i.e., distance).

To improve distance measurement accuracy, one embodiment includes six features that advance the state of the art in subspace decomposition high-resolution estimation methods. Included is an algorithm that combines two or more estimated individual frequencies using different eigenstructure properties, which further reduces ambiguity in delay path determination.

Root Music finds the individual frequencies that minimize the projected energy when the observable is projected onto the noise subspace. The Esprit algorithm determines the individual frequencies based on a rotation operator. In many ways, this operation is the conjugate of Music, in that it finds the frequencies that maximize the projected energy when the observable is projected onto the signal subspace.

Model size can be used in both algorithms, and in practice, in complex signal environments such as those seen in indoor ranging, the model sizes that provide the best performance for Music and Esprit are generally not equal, for reasons that will be discussed below.

For Music, it is preferred to identify the basis elements of the decomposition as "signal eigenvalues" (Type I error). This minimizes the amount of signal energy projected onto the noise subspace and improves accuracy. For Esprit, in contrast, it is preferred to identify the basis elements of the decomposition as "noise eigenvalues." This is also a Type I error. This minimizes the effect of noise on the energy projected onto the signal subspace. Therefore, the model size for Music can generally be slightly larger than the model size for Esprit.

Secondly, in complex signal environments, there are times when it is difficult to estimate the model size with sufficient statistical reliability, given strong reflections and the possibility that the direct path is actually much weaker than some of the multipath reflections. This problem can be solved by estimating a "base" model size for both Music and Esprit, and processing the observable data using Music and Esprit within a model size window defined by the base model size for each. This results in multiple measurement values for each measurement.

The first feature of an embodiment is the use of the F statistic to estimate the model size (see above). The second feature is the use of different Type I error probabilities in the F statistic for Music and Esprit. This implements the Type I error difference between Music and Esprit as discussed above. The third feature is the use of a base model size and window to maximize the probability of detecting the direct path.

Due to the potentially rapidly changing physical and electronic environment, not every measurement will provide a stable answer. This is addressed by using cluster analysis on multiple measurements to provide a stable range estimate. A fourth feature of an embodiment is the use of multiple measurements.

Because multiple signals are present, the probability distribution of multiple answers generated by multiple measurements (each measurement using multiple model sizes from both the Music and Esprit implementations) can be multimodal. Traditional cluster analysis may be insufficient for this application. The fifth feature is the development of multimodal cluster analysis to estimate the direct range and equivalent range of the reflected multipath components. The sixth feature is the analysis of the statistics of the range estimates provided by the cluster analysis (range and standard deviation, and sorting out those estimates that are statistically identical). This results in a more accurate range estimate.

The above method can also be used in a wide-bandwidth ranging signal location finding system.

For the derivation of r(t) in the threshold method, starting from expression (20), we have

The trigonometric constant is used

Except for a ₀ , the coefficient a _k is zero for even k. The reason for this is that on the interval I, the function 1/sinπΔft that we are trying to approximate by h(t) is even around the center of I, but for even k, k≠0, the basis function sinkπΔft is odd around the center of I and is therefore orthogonal to 1/sinπΔft on I. Therefore, we can make the substitution k=2n+1 and assume M is an odd positive integer. In fact, we can assume M=2N+1. This choice has been experimentally determined to provide a good amount of cancellation of oscillations in the interval I.

Now we make the substitution k=N-n in the first sum and k=N+n+1 in the second sum, and we get

Subtracting g(t) from s(t) yields

Now assume

where k = 1, 2, ..., N

where k = N + 1, N + 2, ..., 2N + 1

c＝-a ₀ (A5)

Then (A4) can be written as

Embodiments of the present invention relate to a method for positioning in wireless communications and other wireless networks that substantially obviates one or more of the shortcomings of the related art. By utilizing the multipath mitigation processes, techniques, and algorithms described in U.S. Patent No. 7,872,583, embodiments of the present invention advantageously improve the accuracy of tracking and positioning functions in various types of wireless networks. These wireless networks include wireless personal area networks (WPGANs) such as ZigBee and Bluetooth, wireless local area networks (WLANs) such as WiFi and UWB, wireless metropolitan area networks (WMANs), which are typically composed of multiple WLANs, with WiMax being a prominent example, wireless wide area networks (WANs) such as the White Space TV band, and mobile device networks (MDNs), which are typically used to transmit voice and data. MDNs are typically based on the Global System for Mobile Communications (GSM) and Personal Communications Service (PCS) standards. More recently, MDNs are based on the Long Term Evolution (LTE) standard. These wireless networks typically consist of a combination of devices, including base stations, desktops, tablets, and laptops, handhelds, smartphones, actuators, specialized accessories, sensors, and other communication and data devices (all of which are generally referred to as "wireless network devices").

Existing location and positioning solutions use a variety of technologies and networks, including GPS, AGPS, cell tower triangulation, and Wi-Fi. Some of these methods, including RF fingerprinting, RSSI, and TDOA, are used to derive this location information. While acceptable for current E911 requirements, existing positioning and ranging methods lack the reliability and accuracy required to support upcoming E911 requirements and LBS and/or RTLS applications, particularly in indoor and urban environments.

The method described in U.S. Patent No. 7,872,583 significantly improves the ability to accurately locate and track directional devices within a single wireless network or a combination of wireless networks. The embodiments are significant improvements over existing implementations of tracking and positioning methods used by wireless networks, which use enhanced cell ID numbers and OTDOA (Observed Time Difference of Arrival), including DL-OTDOA (Downlink OTDOA), U-TDOA, UL-TDOA, etc.

Cell ID positioning technology allows the location of a user (UE, user equipment) to be estimated with sector-specific coverage accuracy. Therefore, the achievable accuracy depends on the cell (base station) segmentation scheme and antenna beamwidth. To improve accuracy, enhanced Cell ID technology adds RTT (round-trip time) measurements from the eNB. Note that RTT here constitutes the difference between the transmission of a downlink DPCH (Dedicated Physical Channel) (DPDCH)/DPCCH (Dedicated Physical Data Channel/Dedicated Physical Control Channel) frame and the start of the corresponding uplink physical frame. In this example, the aforementioned frame serves as a ranging signal. Based on information about the length of this signal's propagation from the eNB to the UE, the distance to the eNB can be calculated (see Figure 10).

In the Observed Time Difference of Arrival (OTDOA) technique, the arrival times of signals from neighboring base stations (eNBs) are calculated. Once signals from three base stations are received, the UE position can be estimated in the handset (UE-based approach) or in the network (NT-based UE-assisted approach). The measured signal is the CPICH (Common Pilot Channel). The propagation time of the signal is correlated with a locally generated replica. The peak of the correlation indicates the observed propagation time of the measured signal. The time difference of arrival between the two base stations determines a hyperbola. At least three reference points are required to define the two hyperbolas. The UE's position is at the intersection of these two hyperbolas (see Figure 11).

Downlink Idle Period (IPDL) is a further OTDOA enhancement. The OTDOA-IPDL technique is based on the same measurements as the conventional OTDOA time measurements obtained during the idle period, where the serving eNB stops its transmission and allows UEs within the coverage area of this cell to hear the pilot from the distant eNB. The serving eNB provides the idle period in continuous or burst mode. In continuous mode, one idle period can be inserted in each downlink physical frame (10ms). In burst mode, the idle period occurs in a pseudo-random manner. Further improvements are achieved through time-aligned IPDL (TA-IPDL). Time alignment forms a common idle period during which each base station can stop its transmission or transmit a common pilot. Pilot signal measurements can occur in the idle period. There are several other techniques that can further improve the DL OTDOA-IPDL method, such as cumulative virtual blanking, UTDOA (uplink TDOA), etc. All of these techniques can improve the ability to hear other (non-serving) eNBs.

A significant drawback of OTDOA-based techniques is that, for this approach to work, the base station timing relationships must be known or measured (synchronized). For asynchronous UMTS networks, the 3GPP standard provides recommendations for how this timing can be recovered. However, network operators have not implemented such solutions. Consequently, an alternative approach has been proposed, using RTT measurements instead of CPICH signal measurements (see U.S. Patent Publication No. 20080285505, John Carlson et al., “SYSTEM AND METHOD FOR NETWORK TIMING RECOVERY IN COMMUNICATIONS NETWORKS”).

All of the aforementioned methods/techniques are based on time-of-arrival and/or time-difference-of-arrival measurements of terrestrial signals (RTT, CPICH, etc.). A problem with these measurements is that they are severely affected by multipath. This significantly reduces the positioning/tracking accuracy of these methods/techniques (see Jakub Marek Borkowski: "Performance of Cell ID+RTT Hybrid Positioning Method for UMTS").

One multipath mitigation technique uses detection/measurement values from multiple additional eNBs or radio base stations (RBSs). The minimum is three, but for multipath mitigation, the required number of RBSs is at least six to eight (see "METHOD AND ARRANGEMENT FOR DL-OTDOA (DOWNLINK OBSERVED TIME DIFFERENCE OF ARRIVAL) POSITIONING IN A LTE (LONG TERM EVOLUTION) WIRELESS COMMUNICATIONS SYSTEM," WO/2010/104436). However, the probability of a UE hearing this large number of eNBs is much lower than the probability of hearing three eNBs. This is because, with a large number of RBSs (eNBs), some may be far away from the UE, and the received signals from these RBSs may fall below the UE's receive sensitivity level or have a low signal-to-noise ratio (SNR).

In the case of RF reflections (e.g., multipath), multiple copies of the RF ranging signal with various delay times are superimposed on the DLOS (Direct Link) signal. Because the CPICH, uplink DPCCH/DPDCH, and other signals used in various cell ID and OTDOA methods/techniques (including RTT measurements) have limited bandwidth, it is impossible to distinguish between the DLOS signal and the reflected signals without appropriate multipath processing/mitigation. Without this multipath processing, these reflected signals can cause errors in the estimated Time Difference of Arrival (TDOA) and Time of Arrival (TOA) measurements (including RTT measurements).

For example, the 3G TS 25.515 v.3.0.0 (199-10) standard defines RTT as "the difference between the transmission of a downlink DPCH frame (signal) and the start (first apparent path) reception of the corresponding uplink DPCCH/DPDCH frame (signal) from the UE". The standard does not define what constitutes this "first apparent path". The standard goes on to note that "the definition of the first apparent path requires further refinement". For example, in a congested multipath environment, it is common for the DLOS signal (which is the first apparent path) to be severely attenuated (10 dB to 20 dB) relative to one or more reflected signals. If the "first apparent path" is determined by measuring signal strength, it may be one of the reflected signals instead of the DLOS signal. This can lead to erroneous TOA/DTOA/RTT measurements and a loss of positioning accuracy.

In previous wireless network generations, positioning accuracy was also affected by the low sampling rates of the frames (signals) used by positioning methods—RTT, CPCIH, and other signals. Current third-generation and subsequent wireless network generations have much higher sampling rates. Therefore, in these networks, the real impact on positioning accuracy comes from terrestrial RF propagation phenomena (multipath).

Embodiments may be used in all wireless networks that utilize reference and/or pilot signals and/or synchronization signals, including simplex, half-duplex, and full-duplex operating modes. For example, embodiments may operate in wireless networks that utilize OFDM modulation and/or its derivatives. Thus, embodiments may operate in LTE networks.

It is also applicable to other wireless networks, including WiMax, WiFi, and White Space. Other wireless networks that do not use reference and/or pilot or synchronization signals may employ one or more of the following types of alternative modulation embodiments as described in U.S. Patent No. 7,872,583: 1) where a portion of the frame is dedicated to a ranging signal/ranging signal element as described in U.S. Patent No. 7,872,583; 2) where the ranging signal element (U.S. Patent No. 7,872,583) is embedded in the transmit/receive signal frame; and 3) where the ranging signal element (described in U.S. Patent No. 7,872,583) is embedded with data.

These alternative embodiments employ the multipath mitigation processor and multipath mitigation techniques/algorithms described in US Patent No. 7,872,583 and may be used in all modes of operation: simplex, half-duplex, and full-duplex.

It is also possible that multiple wireless networks can utilize the preferred and/or alternative embodiments simultaneously. By way of example, a smartphone may have Bluetooth, WiFi, GSM, and LTE capabilities, with the ability to operate on multiple networks simultaneously. Depending on application needs and/or network availability, different wireless networks may be utilized to provide positioning information.

The proposed embodiment methods and systems utilize wireless network reference/pilot and/or synchronization signals. Furthermore, reference/pilot/synchronization signal measurements may be combined with RTT (round trip time) measurements or system timing. According to an embodiment, RF-based tracking and positioning is implemented on a 3GPP LTE cellular network, but may also be implemented on other wireless networks using various signaling technologies, such as WiMax, Wi-Fi, LTE, and sensor networks. The exemplary and alternative embodiments described above utilize multipath mitigation methods/techniques and algorithms as described in U.S. Patent No. 7,872,583. The proposed system may utilize software-implemented digital signal processing.

The system of the embodiments utilizes user equipment (UE), such as mobile phones or smartphones, hardware/software, and base station (Node B)/enhanced base station (eNB) hardware/software. A base station typically consists of a transmitter and receiver connected to an antenna via a feeder in a cabin or cabinet. These base stations include microcells, picocells, macrocells, umbrella cells, cell towers, routers, and femtocells. Consequently, there can be little or no added cost to the UE device and the overall system. Simultaneously, positioning accuracy can be significantly improved.

Improved accuracy comes from multipath mitigation provided by embodiments of the present invention and U.S. Patent No. 7,872,583. Embodiments utilize multipath mitigation algorithms, network reference/pilot and/or synchronization signals, and network nodes (eNBs). These may be supplemented with RTT (round trip time) measurements. The multipath mitigation algorithms are implemented in the UE and/or the base station (eNB), or in both the UE and the eNB.

Embodiments advantageously utilize a multipath mitigation processor/algorithm (see U.S. Patent No. 7,872,583) that allows separation of the DLOS signal and the reflected signal even when the DLOS signal is significantly attenuated (10 dB to 20 dB lower) relative to one or more reflected signals. Thus, embodiments significantly reduce errors in estimating the ranging signal DLOS time of flight and, therefore, in TOA, RTT, and DTOA measurements. The proposed multipath mitigation and DLOS discrimination (identification) method can be used across all RF bands and wireless systems/networks. Furthermore, it can support a variety of modulation/demodulation techniques, including spread spectrum techniques such as DSS (direct spread spectrum) and FH (frequency hopping).

In addition, to further improve the accuracy of the method, noise reduction methods can be applied. These noise reduction methods may include (but are not limited to) coherent summation, incoherent summation, matched filtering, time diversity techniques, etc. By applying post-processing techniques such as maximum likelihood estimation (e.g., Viterbi algorithm) and minimum variance estimation (Kalman filtering), the residual multipath interference error can be further reduced.

In embodiments of the present invention, the multipath mitigation processor and multipath mitigation techniques/algorithms do not alter RTT, CPCIH, and other signals and/or frames. Embodiments of the present invention utilize wireless network reference, pilot, and/or synchronization signals for obtaining channel response/estimation. The present invention utilizes channel estimation statistics generated by the UE and/or eNB (see Iwamatsu et al., "APPARATUS FOR ESTIMATING PROPAGATION PATH CHARACTERISTICS," US 2003/008156; US 7167456 B2).

LTE networks use specific (non-data) reference/pilot and/or synchronization signals (known signals) that are transmitted in every downlink and uplink subframe and may span the entire cell bandwidth. For simplicity, from now on, we will refer to reference/pilot and synchronization signals as reference signals. Figure 9 shows an example of LTE reference signals (these signals are interspersed among LTE resource elements). According to Figure 2, a reference signal (symbol) can be transmitted every six subcarriers. In addition, the reference signals (symbols) are staggered in both time and frequency. In total, the reference signal may cover every third subcarrier.

These reference signals are used by the UE for initial cell search, downlink signal strength measurement, scheduling, and handover. Included among the reference signals are UE-specific reference signals for channel estimation (response determination) for coherent demodulation. In addition to UE-specific reference signals, other reference signals can also be used for channel estimation purposes (see Chen et al., U.S. Patent Publication No. 2010/0091826 A1).

LTE uses OFDM (Orthogonal Frequency Division Multiplexing) modulation. In LTE, ISI (Inter-Symbol Interference) caused by multipath is addressed by inserting a cyclic prefix (CP) at the beginning of each OFDM symbol. The CP provides sufficient delay so that delayed reflections from the previous OFDM symbol fade away before arriving at the next OFDM symbol.

An OFDM symbol consists of multiple very closely spaced subcarriers. Within an OFDM symbol, time-staggered copies of the current symbol (caused by multipath) cause inter-carrier interference (ICI). In LTE, ICI is handled (mitigated) by determining the multipath channel response and correcting for it in the receiver.

In LTE, the multipath channel response (estimate) is calculated in the receiver based on the subcarriers that carry reference symbols. Interpolation is used to estimate the channel response with respect to the remaining subcarriers. The channel response is calculated (estimated) in the form of channel amplitude and phase. Once the channel response is determined (via periodic transmission of a known reference signal), the channel distortion caused by multipath is suppressed by applying amplitude and phase shifts on a subcarrier-by-subcarrier basis (see Jim Zyren, "Overview of the 3GPP Long Term Evolution Physical Layer," white paper).

LTE multipath mitigation is designed to remove ISI (by inserting a cyclic prefix) and ICI, rather than separating the DLOS signal from the reflected signal. For example, the time-interleaved copies of the current symbol cause each modulated subcarrier signal to be spread out in time, thus causing ICI. Correcting the multipath channel response using the LTE techniques described above can shrink the modulated subcarrier signal in time, but this type of correction does not guarantee that the resulting modulated subcarrier signal (within the OFDM symbol) is the DLOS signal. If the DLOS modulated subcarrier signal is significantly attenuated relative to the delayed reflected signal, the resulting output signal may be the delayed reflected signal and the DLOS signal may be lost.

In LTE-compatible receivers, further signal processing involves a DFT (digital Fourier transform). As is well known, the DFT technique can resolve (remove) delayed copies of a signal that are greater than or equal to a time inversely proportional to the signal and/or channel bandwidth. While the accuracy of this method may be adequate for efficient data transfer, it is not accurate enough for precise distance measurements in congested multipath environments. For example, to achieve an accuracy of thirty meters, the signal and receiver channel bandwidths should be greater than or equal to ten megahertz (1/10 MHz = 100 ns). For even better accuracy, the signal and receiver channel bandwidths should be wider: one hundred megahertz for three meters.

However, the CPICH, uplink DPCCH/DPDCH, and other signals used in various cell ID and OTDOA methods/techniques (including RTT measurements), as well as the LTE receive signal subcarriers, have bandwidths significantly below 10 MHz. Consequently, currently employed methods/techniques (in LTE) can produce positioning errors in the 100-meter range.

To overcome the above limitations, embodiments use a unique combination of subspace decomposition high-resolution spectrum estimation methods and implementations of multi-mode cluster analysis. This analysis and related multipath mitigation methods/techniques and algorithms described in U.S. Patent No. 7,872,583 allow reliable and accurate separation of DLOS paths from other reflected signal paths.

Compared to the methods/techniques used in LTE, this method/technique and algorithm (U.S. Patent No. 7,872,583) provides a 20x to 50x improvement in the accuracy of distance measurements in crowded multipath environments by reliably and accurately separating the DLOS path from other multipath (MP) paths.

The methods/techniques and algorithms described in U.S. Patent No. 7,872,583 require ranging signal complex amplitude estimation. Therefore, LTE reference signals and other reference signals (including pilot and/or synchronization signals) used for channel estimation (response determination) can also be understood as ranging signals in the methods/techniques and algorithms described in U.S. Patent No. 7,872,583. In this case, the ranging signal complex amplitude is the channel response calculated (estimated) by the LTE receiver in the form of amplitude and phase. In other words, the channel response statistics calculated (estimated) by the LTE receiver can provide the complex amplitude information required by the methods/techniques and algorithms described in U.S. Patent No. 7,872,583.

In an ideal open-space RF propagation environment without multipath, the phase variation of the received signal (ranging signal), e.g., the channel response phase, can be directly proportional to the signal's frequency (a straight line). The RF signal flight time (propagation delay) in such an environment can be directly calculated from the phase versus frequency dependence by calculating the first-order derivative of the phase versus frequency dependence. The result can be a propagation delay constant.

In an ideal world, the absolute value of the phase at the initial (or any) frequency is unimportant because the derivative is unaffected by the absolute value of the phase.

In a crowded multipath environment, the received signal phase versus frequency is a complex curve (not a straight line), and the first-order derivative does not provide information that can be used to accurately separate the DLOS path from other reflected signal paths. This is why the multipath mitigation processor, method/technique, and algorithm described in U.S. Patent No. 7,872,583 are used.

If the phase and frequency synchronization (phase coherence) achieved in a given wireless network/system is excellent, then the multipath mitigation processor and methods/techniques and algorithms described in U.S. Patent No. 7,872,583 can accurately separate the DLOS path from other reflected signal paths and determine the length (time of flight) of this DLOS path.

In this phase coherent network/system, no additional measurements are required. In other words, one-way ranging (simplex ranging) can be achieved.

However, if the degree of synchronization (phase coherence) achieved in a given wireless network/system is not precise enough, then in a crowded multipath environment, the received signal phase and amplitude variations versus frequency may be very similar to measurements made at two or more different locations (distances). This phenomenon may lead to ambiguity in the determination of the DLOS range (time of flight) of the received signal.

In order to resolve this ambiguity, it may be necessary to know the actual (absolute) phase value for at least one frequency.

However, the frequency dependence of the amplitude and phase calculated by the LTE receiver does not include the actual phase value, because all amplitude and phase values are calculated based on the downlink/uplink reference signal (e.g., relative to each other). Therefore, the amplitude and phase of the channel response calculated (estimated) by the LTE receiver requires the actual phase value at at least one frequency (subcarrier frequency).

In LTE, this actual phase value may be determined from one or more RTT measurements, TOA measurements; or

Determined from the timestamps of one or more received reference signals, provided that 1) the timestamps of the signals transmitted by the eNB are also known at the receiver (or vice versa), 2) the receiver and eNB clocks are well synchronized in time, and/or 3) by using multilateration techniques.

All of the above methods provide time-of-flight values of one or more reference signals. Based on the time-of-flight values and the frequencies of these reference signals, the actual phase values at one or more frequencies can be calculated.

Embodiments of the present invention achieve extremely accurate DLOS range determination/positioning in congested multipath environments by combining the multipath mitigation processors, methods/techniques, and algorithms described in U.S. Patent No. 7,872,583 with: 1) amplitude and phase versus frequency dependencies computed by an LTE UE and/or eNB receiver; or 2) a combination of amplitude and phase versus frequency dependencies computed by an LTE UE and/or eNB receiver with actual phase values of one or more frequencies obtained via RTT and/or TOA; and/or timestamp measurements.

In these cases, the actual phase value is affected by multipath. However, this does not affect the performance of the method/technique and algorithm described in US Patent No. 7,872,583.

In LTE, measurements of RTT/TOA/TDOA/OTDOA (including DL-OTDOA, U-TDOA, UL-TDOA, etc.) can be performed with a resolution of 5 meters. RTT measurements are performed during a dedicated connection. Therefore, there may be multiple simultaneous measurements when the UE is in handover state and when the UE periodically collects measurements and feeds them back to the UE, where DPCH frames are exchanged between the UE and different networks (base stations). Similar to RTT, TOA measurements provide the flight time (propagation delay) of the signal, but TOA measurements cannot be performed simultaneously (Jakub Marek Borkowski: "Performance of Cell ID+RTT Hybrid Positioning Method for UMTS").

To locate a UE on a plane, the DLOS distances to at least three eNBs must be determined. To locate a UE in three-dimensional space, a minimum of four DLOS distances to four eNBs must be determined (assuming at least one eNB is not on the same plane).

FIG1 shows an example of a UE positioning method.

In the case of very perfect synchronization, no RTT measurement is required.

If the degree of synchronization is not precise enough, methods such as OTDOA, Cell ID+RTT, for example AOA (angle of arrival) and combinations thereof with other methods may be used for UE positioning.

The accuracy of the Cell ID+RTT tracking positioning method is affected by multipath (RTT measurement) and the eNB (base station) antenna beamwidth. The base station antenna beamwidth is between 33 degrees and 65 degrees. These wide beamwidths lead to positioning errors of 50 to 150 meters in urban areas (Jakub Marek Borkowski: "Performance of Cell ID+RTT Hybrid Positioning Method for UMTS"). Considering that the average error of current LTE RTT distance measurements in a congested multipath environment is approximately 100 meters, the overall expected average positioning error of the LTE Cell ID+RTT method currently used is approximately 150 meters.

One embodiment is a UE positioning method based on an AOA method, whereby one or more reference signals from the UE are used for UE positioning purposes. This relates to an AOA determination device positioning for determining DLOS AOA. The device can be co-located with the base station and/or installed at one or more locations independent of the base station location. The coordinates of these locations are generally known. No changes are required on the UE side.

This device contains a small antenna array and is based on a variation of the same multipath mitigation processor, methods/techniques and algorithms described in US Patent No. 7,872,583. This one possible embodiment has the advantage of precisely determining the AOA of the (very narrow beamwidth) DLOS RF energy from the UE unit.

In another option, this added device can be a receiver-only device. Therefore, its size/weight and cost are very low.

The combination of embodiments in which accurate DLOS distance measurements are obtained and embodiments in which accurate DLOS AOA determination can be made can greatly improve the accuracy of the Cell ID+RTT tracking-positioning method - by a factor of 10 or more. Another advantage of this method is that the UE position can be determined at any time using a single cell tower (without requiring the UE to be placed in soft handover mode). Because accurate positioning can be obtained using a single cell tower, there is no need to synchronize multiple cell towers. Another option for determining DLOS AOA is to use existing eNB antenna arrays and eNB equipment. This option can further reduce the cost of the implementation of the improved Cell ID+RTT method. However, because the eNB antennas are not designed for positioning applications, the positioning accuracy can be reduced. In addition, network operators may be reluctant to implement the required changes (software/hardware) in the base stations.

In LTE (Evolved Universal Terrestrial Radio Access (E-UTRA); Physical Channels and Modulation; 3GPP TS 36.211 Release 9 Specification), Positioning Reference Signals (PRS) were added. These signals are used by UEs for DL-OTDA (Downlink OTDOA) positioning. Furthermore, this Release 9 mandates eNB synchronization. This removes the final hurdle for the OTDOA method (see paragraph 274 above). PRS improves the ability of UEs to hear multiple eNBs. It should be noted that Release 9 does not specify eNB synchronization accuracy (some proposals include 100ns).

U-TDOA/UL-TDOA is in the research stage and will be standardized in the 2011 version.

The DL-OTDOA method (Release 9) is described in detail in US Patent No. US 2011/0124347A1 ("Method and Apparatus for UE Positioning in LTE Networks," Chen et al.). Release 9 DL-OTDOA suffers from multipath issues. Some multipath mitigation can be achieved through increased PRS signal bandwidth. However, the tradeoff is increased scheduling complexity and longer time between UE positions. Furthermore, for networks with limited operating bandwidth (e.g., 10 MHz), the best possible accuracy is 100 meters (see Chen, Table 1).

The above figures are the best possible case. Other cases, especially when the DLOS signal strength is significantly lower (10 to 20 dB) compared to the reflected signal strength, result in significantly larger positioning/ranging errors (2 to 4 times).

The embodiments described herein allow for up to 50x improvement in ranging/positioning accuracy for a given signal bandwidth compared to the performance achieved by the Release 9 DL-OTDOA method and the UL-PRS method of Chen et al. described in the Background section. Thus, applying embodiments of the methods described herein to Release 9 PRS processing can reduce positioning errors to 3 meters or better in 95% of all possible cases. Additionally, this accuracy gain can reduce scheduling complexity and the time between UE position fixes.

Using the embodiments described herein, further improvements to the OTDOA method are possible. For example, ranging to a serving cell can be determined based on signals from other serving cells, thereby improving the hearability of neighboring cells and reducing scheduling complexity, including the time between UE position fixes.

The embodiments also improve the accuracy performance of Chen et al.'s U-TDOA method and UL-TDOA (described in the Background section) by up to 50 times. Applying the embodiments to Chen's UL-TDOA variant reduces positioning error to 3 meters or better in 95% of all possible cases. Furthermore, this accuracy gain reduces scheduling complexity and the time between UE positioning.

Similarly, using embodiments of the present invention, the accuracy of Chen's UL-TDOA method can be improved by up to 50 times. Therefore, applying embodiments of the present invention to Chen's U-TDOA variant can reduce positioning error to 3 meters or better in 95% of all possible cases. Furthermore, this accuracy gain can further reduce scheduling complexity and the time between UE positioning.

The DL-TDOA and U-TDOA/UL-TDOA methods described above rely on one-way measurements (ranging). Embodiments of the present invention, and indeed all other ranging techniques, require that the PRS and/or other signals used in the one-way ranging process be frequency and phase coherent. OFDM-based systems, such as LTE, are frequency coherent. However, the UE unit and the eNB are not synchronized in phase or time to within a few nanoseconds by a common source (such as UTC), for example, due to the presence of random phase adders.

To avoid the impact of phase coherence on ranging accuracy, embodiments of the multipath processor calculate the differential phase between ranging signals (e.g., reference signal, individual components (subcarriers)). This eliminates the need for random phase term adders.

As identified above in the discussion of Chen et al., applying the embodiments described herein results in significantly improved accuracy in indoor environments compared to the performance achieved by Chen et al. For example, according to Chen et al., DL-OTDOA and/or U-TDOA/UL-TDOA are primarily intended for outdoor environments, and DL-OTDOA and U-TDOA techniques may not perform well indoors (buildings, campuses, etc.). Several reasons are noted (see Chen, #161-164), including the fact that distributed antenna systems (DAS) are typically employed indoors, whereby each antenna does not have a unique ID.]

The embodiments described below operate with wireless networks employing OFDM modulation and/or its derivatives and reference/pilot/and/or synchronization signals. Thus, the embodiments described below operate with LTE networks and are also applicable to other wireless systems and other wireless networks, including other types of modulation, with or without reference/pilot/and/or synchronization signals.

The methods described herein are also applicable to other wireless networks, including WiMax, WiFi, and White Space. Other wireless networks that do not use reference/pilot and/or synchronization signals may employ one or more of the following types of alternative modulation embodiments, as described in U.S. Patent No. 7,872,583: 1) where a portion of the frame is dedicated to the ranging signal/ranging signal element; 2) where the ranging signal element is embedded in the transmit/receive signal frame; and 3) where the ranging signal element is embedded with data.

Embodiments of the multipath mitigation range estimation algorithm described herein (also described in US Pat. Nos. 7,969,311 and 8,305,215) operate by providing an estimate of range as a set consisting of the direct path (DLOS) of the signal plus multipath reflections.

The LTE DAS system generates multiple copies of the same signal seen by a mobile receiver (UE) at various time offsets. The delays are used to uniquely determine the geometric relationship between the antennas and the mobile receiver. The signal seen by the receiver is similar to that seen in a multipath environment, except for the main "multipath" component caused by the sum of the offset signals from multiple DAS antennas.

The signal set seen by the receiver is the same type of signal set that the embodiment design utilizes, except that in this case, the dominant multipath component is not traditional multipath. The multipath mitigation processor (algorithm) of the present invention is able to determine the DLOS and the attenuation and propagation delay of each path, such as reflections (see Equations 1 through 3 and the associated description). While multipath may exist due to a disparate RF channel (environment), the dominant multipath components in this signal set are associated with transmissions from multiple antennas. An embodiment of the multipath algorithm of the present invention can estimate these multipath components, isolate the range of the DAS antenna from the receiver, and provide range data to a position processor (implemented in software). Depending on the antenna placement geometry, this solution can provide both X, Y and X, Y, Z position coordinates.

Therefore, embodiments of the present invention do not require any hardware and/or new network signal additions. In addition, positioning accuracy can be significantly improved by: 1) mitigating multipath; and 2) in the case of active DAS, the lower limit of positioning error can be significantly reduced, for example, from approximately 50 meters to approximately 3 meters.

It is assumed that the position (location) of each antenna of the DAS is known.In an embodiment, the signal propagation delay for each antenna (or relative to other antennas) must also be determined (known).

For active DAS systems, signal propagation delay can be automatically determined using loopback technology, whereby a known signal is sent back and forth and the round-trip time is measured. This loopback technology also eliminates changes (drift) in signal propagation delay over temperature, time, etc.

The use of multiple macrocells and associated antennas, picocells and microcells further improves resolution by providing additional reference points.

The above-described embodiment of individual range estimates in the signal set of multiple replicas from multiple antennas can be further improved by modifying the signal transmission structure in two ways. The first is to time-division multiplex the transmissions from each antenna. The second approach is to frequency-division multiplex each of the antennas. Using both time and frequency division multiplexing can further improve the ranging and positioning accuracy of the system. Another approach is to increase the propagation delay of each antenna. The delay value is chosen to be large enough to exceed the delay spread in the specific DAS environment (channel), but less than the cyclic prefix (CP) length so that the multipath caused by the additional delay does not cause ISI (inter-symbol interference).

The addition of a unique ID or unique identifier for each antenna increases the efficiency of the resulting solution. For example, it does not require the processor to estimate all ranges starting from the signal from each antenna.

In one embodiment utilizing the LTE downlink, one or more reference signal subcarriers (including pilot and/or synchronization signal subcarriers) are used to determine subcarrier phase and amplitude, which are in turn applied to a multipath processor to perform multipath interference suppression and use multilateration and position consistency algorithms to generate range-based position observables and positioning estimates to remove outliers during editing.

Another embodiment takes advantage of the fact that LTE uplink signaling also includes reference signals to the base mobile device, which also contain reference subcarriers. In fact, there is more than one mode that contains these subcarriers from the full sounding mode used by the network to assign the frequency band to the uplink device to a mode where the reference subcarriers are used to generate a channel impulse response to assist in demodulation of the uplink signal, etc. In addition, similar to the DL PRS added in Release 9, additional UL reference signals may be added in upcoming and future standard releases. In this embodiment, the uplink signal is processed by multiple base units (eNBs) using the same range to phase, multipath mitigation processing to produce range-dependent observables. In this embodiment, a position consistency algorithm such as that established by the multilateration algorithm is used to compile the wild point observables and produce a position estimate.

In another embodiment, the relevant reference subcarrier(s) (including pilot and/or synchronization) for both the LTE downlink and LTE uplink are collected, range-to-phase mapping is applied, multipath mitigation is applied, and range-related observables are estimated. This data is then fused in such a way that a more stable set of observables for position is provided using multilateration and position consistency algorithms. The advantage is redundancy for improved accuracy due to the two different frequency bands for downlink and uplink, or improved system coherence in the case of TDD (time division duplex).

In a DAS (distributed antenna system) environment where multiple antennas transmit the same downlink signal from a microcell, the position consistency algorithm is extended to isolate the range of the DAS antenna from observables generated from reference signal (including pilot and/or synchronization) subcarriers through multipath mitigation processing, and obtain a position estimate based on the ranges of multiple DAS transmitters (antennas).

In a DAS system (environment), obtaining an accurate position estimate is possible if the signal paths from individual antennas can be resolved with high accuracy, where the path error is a fraction of the distance between the antennas (accuracy of 10 meters or better). Because all existing techniques/methods are unable to provide this accuracy in a crowded multipath environment (where signals from multiple DAS antennas may appear as induced crowded multipath), existing techniques/methods are unable to utilize the aforementioned extensions to the position consistency algorithm and this positioning method/technique in a DAS environment.

The InvisiTrack multipath mitigation method and system for object identification and location finding described in U.S. Patent No. 7,872,583 is applied to range-to-signal phase mapping, multipath interference mitigation, and processing to generate range-based position observables utilizing LTE downlink, uplink, and/or both (downlink and uplink), one or more reference signal subcarriers, and use multilateration and position consistency to produce a position estimate.

In all the above embodiments, a trilateration positioning algorithm may also be used.

DL-OTDOA positioning is specified in LTE Release 9: Evolved Universal Terrestrial Radio Access (E-UTRA); Physical Channels and Modulation; 3GPP TS 36.211 Release 9 Specification. However, it has not yet been implemented by wireless operators (carriers). Meanwhile, downlink positioning can be implemented within current (e.g., unmodified) LTE network environments by using existing physical layer measurement operations.

In LTE, the UE and eNB are required to perform physical layer measurements of radio characteristics. The measurement definitions are specified in 3GPP TS 36.214. These measurements are performed periodically and reported to higher layers for various purposes, including intra-frame and inter-frequency handovers, inter-frame radio access technology (Inter-RAT) handovers, timing measurements, and other purposes to support RRM (Radio Resource Management).

For example, RSRP (reference signal received power) is the average of the power of all resource elements carrying cell-specific reference signals over the entire bandwidth.

Another example is the RSRQ (Reference Signal Received Quality) measurement which provides additional information (RSRQ combines signal strength as well as interference level).

The LTE network provides the UE with a list of eNB neighbors (to the serving eNB). Based on network knowledge, the (serving) eNodeB provides the UE with the identifiers of the neighboring eNBs. The UE then measures the signal quality of the neighbors it can receive. The UE feeds the results back to the eNodeB. Note that the UE also measures the signal quality of the serving eNB.

For the purposes of this specification, RSRP is defined as the linear average over the power contributions (in watts) of resource elements carrying cell-specific reference signals within the considered measurement frequency bandwidth. The measurement bandwidth used by a UE to determine RSRP is determined by the UE implementation and must meet the corresponding measurement accuracy requirements.

This bandwidth is quite large, given the measurement bandwidth accuracy requirements. The cell-specific reference signals used in RSRP measurements can be further processed to determine the phase and amplitude of these reference signal subcarriers. These subcarrier phases and amplitudes are then applied to a multipath processor for multipath interference mitigation and to generate range-based position observables. Other reference signals used in RSRP measurements, such as the Secondary Synchronization Signal (SSS), may also be used.

Afterwards, based on the range observables from three or more cells, the position fix can be estimated using multilateration and position consistency algorithms.

As mentioned previously, while there are several reasons for RF fingerprinting database instability, one of the main reasons is multipath (RF signatures are extremely sensitive to multipath). Therefore, the positioning accuracy of RF fingerprinting methods/techniques is severely affected by multipath dynamics - changes with time, environment (e.g., weather), human and/or object movement, including vertical uncertainty: >100% variation depending on device Z height and/or antenna orientation (see Tsung-Han Lin et al., "Microscopic Examination of an RSSI-Signature-Based Indoor Localization System").

Embodiments of the present invention can significantly improve RF fingerprinting positioning accuracy due to the ability to find and characterize each individual path (including significantly attenuated DLOS) (multipath processor). Therefore, RF fingerprinting decisions about positioning can be supplemented with real-time multipath distribution information.

As mentioned above, positioning may require synchronization of position references in time. In wireless networks, these position references may include access points, macro/micro/pico and femto cells, as well as so-called small cells (eNBs). However, wireless operators have not implemented the synchronization accuracy required for precise positioning. For example, in the case of LTE, the standard does not require any time synchronization between eNBs for FDD (Frequency Division Duplex) networks. For LTE TDD (Time Division Duplex), this time synchronization accuracy limit is +/- 1.5 microseconds. This equates to a positioning uncertainty of 400+ meters. Although not required, LTE FDD networks are also synchronized but use even larger limits (than 1.5 microseconds).

Wireless LTE operators use GPS/GNSS signals to synchronize eNBs in frequency and time. It should be noted that LTE eNBs must maintain very precise carrier frequencies: 0.05 ppm for macro/micro cells and slightly less accurate (0.1-0.25 ppm) for other cell types. GPS/GNSS signals can also allow for desired (positioning) time synchronization accuracy better than 10 nanoseconds. However, network operators and network equipment manufacturers are attempting to reduce the costs associated with GPS/GNSS units by adopting NTP (Network Time Protocol) and/or PTP (Precision Time Protocol) (e.g., IEEE 1588v2 PTP) to support packet-based/, for example, Internet/Ethernet time synchronization.

Synchronization based on IP networks has the potential to meet the minimum frequency timing requirements, but lacks the GPS/GNSS precision required for positioning.

The method described herein is based on GPS/GNSS signals and signals generated by eNBs and/or access points, or other wireless network devices. It can also be based on IP network synchronization signals and protocols and signals generated by eNBs and/or access points, or other wireless network devices. This method is also applicable to other wireless networks, including WiMax, WiFi, and White Space.

The eNB signal is received by a Time Observation Unit (TMO) installed at the operator’s eNB facility (Figure 12). The TMO also contains an external synchronization source input.

The eNB signal is processed by the TMO and time stamped using a clock synchronized with an external synchronization source input.

The external synchronization source can come from GPS/GNSS and/or Internet/Ethernet, such as PTP or NTP.

The processed signal with a timestamp, such as the LTE frame start (it can be other signals, especially in other networks), also containing the eNB (cell) location and/or cell ID, is sent via the Internet/Ethernet backhaul to the central TMO server, which forms, maintains and updates a database of all eNBs.

The UE and/or eNB participating in the ranging and obtaining positioning process will consult the TMO server and the server can return the time synchronization offsets between the participating eNBs. These time synchronization offsets can be used by the UE and/or eNB participating in the obtaining positioning process to adjust the positioning.

Alternatively, when the UE and/or eNB participating in the ranging process can also supply the obtained ranging information to the TMO server, the TMO server can perform the positioning calculation and adjustment. The TMO server can then return the precise (adjusted) position (position).

If more than one cell eNB equipment is co-located, a single TMO can process and time-stamp the signals from all eNBs.

RTT (Round Trip Time) measurement (ranging) can be used for positioning. The disadvantage is that RTT ranging is subject to multipath, which has a strong impact on positioning accuracy.

On the other hand, in general and in the case of LTE in particular, for eNBs, RTT positioning does not require position reference synchronization (in time).

At the same time, when operating with pilot references and/or other signals of a wireless network, the multipath mitigation processor, methods/techniques, and algorithms described in U.S. Patent No. 7,872,583 can determine the channel response to the RTT signal, for example, determining the multipath channel through which the RTT signal traveled. This allows correction of the RTT measurement so that the actual DLOS time can be determined.

Knowing the DLOS time, it is possible to obtain a position fix using trilateration and/or similar positioning methods without the need for eNB or position reference synchronization in time.

Even with the TMO and TMO server in place, integrating InvisiTrack's technology may require changes in macro/micro/pico and small cells and/or UEs (handsets). While these changes can be limited to SW/FW (software/firmware), they represent a significant effort to retrofit existing infrastructure. Furthermore, in some cases, network operators and/or UE/handset manufacturers/suppliers resist device modifications. As used herein, UE refers to wireless network user equipment.

This SW/FW change can be avoided entirely if the TMO and TMO server functionality is expanded to support InvisiTrack positioning technology. In other words, another embodiment described below operates with wireless network signals but does not require any modifications to the wireless network equipment/infrastructure. Thus, the embodiment described below operates with LTE networks, but is also applicable to other wireless systems/networks, including Wi-Fi.

Essentially, this embodiment forms a parallel wireless positioning infrastructure that uses wireless network signals to obtain positioning.

Similar to the TMO and TMO server, InvisiTrack's location infrastructure can be composed of one or more wireless network signal acquisition units (NSAUs) and one or more location server units (LSUs). The LSUs collect data from the NSAUs and analyze it to determine range and location, converting it into a table of phone/UE IDs and locations at a given moment. The LSUs interface with the wireless network via a network API.

Multiple of these units can be deployed at various locations in a large infrastructure. If the NSAUs have coherent timing, all the results can be used, which provides better accuracy.

Coherent timing may be derived from a GPS clock and/or other stable clock sources.

The NSAU communicates with the LSU via a LAN (Local Area Network), a Metropolitan Area Network (MAN) and/or the Internet.

In some devices/instances, the NSAU and LSU may be combined/integrated into a single unit.

To support positioning services using LTE or other wireless networks, transmitters need to synchronize clocks and events to within tight tolerances. This is typically accomplished by locking onto the GPS 1PPS signal. This allows timing synchronization within a local area to within 3 nanoseconds of 1-sigma.

However, there are many instances where this type of synchronization is not practical. Embodiments of the present invention provide time offset estimates between downlink transmitters and tracking of time offsets to provide delay compensation values for the positioning process, allowing the positioning process to proceed as if the transmitters were clock- and event-synchronized. This is accomplished by having advance knowledge of the transmitting antennas (required for any positioning service) and a receiver with known a priori antenna positions. This receiver, called a synchronization unit, collects data from all downlink transmitters and, assuming knowledge of their positions, calculates offset timing from a preselected base antenna. The system tracks these offsets using a tracking algorithm that compensates for clock drift in the downlink transmitters. It should be noted that the process of deriving pseudoranges from the received data can utilize the InvisiTrack multipath mitigation algorithm (described in U.S. Patent No. 7,872,583). Therefore, synchronization is unaffected by multipath.

These offsets are used by the positioning processor (Location Server, LSU) to properly calibrate the data from each downlink transmitter so that it appears to have been generated by synchronized transmitters. The timing accuracy is comparable to the best 1-PPS tracking and can support 3 meters of positioning accuracy (1-sigma).

For optimal performance, the synchronization receiver and/or receiver antenna can be positioned based on the optimal GDOP. In large facilities, multiple synchronization receivers can be utilized to provide an equivalent 3 nanosecond 1-sigma synchronization offset across the entire network. By utilizing synchronization receivers, synchronization of the downlink transmitter is no longer required.

The sync receiver unit can be a separate unit that communicates with the NSAU and/or LSU. Alternatively, the sync receiver can be integrated with the NSAU.

An exemplary wireless network positioning device diagram is depicted in FIG. 13 .

An embodiment of a fully autonomous system (no customer network investment) utilizing LTE signals operates in the following modes:

1. Uplink Mode - uses wireless network uplink (UL) signals for positioning purposes (Figures 16 and 17)

2. Downlink mode - uses wireless network downlink (DL) signals for positioning purposes (Figures 14 and 15).

3. Bidirectional mode - uses both UL and DL signals for positioning.

In uplink mode, multiple antennas are connected to one or more NSAUs. These antenna positions are independent of the wireless network antennas; the NSAU antenna positions are selected to minimize GDOP (Geometric Dilution of Precision).

Network RF signals from UE/handset devices are collected by the NSAU antenna and processed by the NSAU to produce time-stamped samples of the processed network RF signals at time intervals suitable for capturing one or more instances of all signals of interest.

Optionally, the NSAU may also receive, process and time stamp samples of the downlink signal to obtain additional information, such as for determining UE/phone ID, etc.

Based on the captured time-stamped samples, the UE/handset device identification number (ID) and the time-stamped wireless network signal of interest associated with each UE/handset ID can be determined (obtained). This operation can be performed by the NSAU or by the LSU.

The NSAU may periodically supply data to the LSU. If unplanned data is required for one or more UE/handset IDs, the LSU may request additional data.

For UL mode operation, no changes/modifications may be required in the wireless network infrastructure and/or existing UEs/handsets.

In downlink (DL) mode, a UE that supports InvisiTrack may be required. In addition, if a phone is used to obtain positioning, the phone FW will have to be modified.

In some cases, the operator may make baseband signals available from the BBU (baseband unit). In such cases, the NSAU may also be capable of processing these available baseband wireless network signals instead of RF wireless network signals.

In DL mode, there is no need to associate the UE/handset ID with one or more wireless network signals, as these signals can be processed in the UE/handset, or the UE/handset can periodically generate time-stamped samples of the processed network RF signals and send these to the LSU; and the LSU can send the results back to the UE/handset.

In DL mode, the NSAU processes and timestamps the processed RF or baseband (when available) wireless network signals. Based on the captured, time-stamped samples, the wireless network signal DL frame start associated with the network antenna can be determined (obtained) and the difference (offset) between these frame starts can be calculated. This operation can be performed by the NSAU or by the LSU. The frame start offset for the network antenna can be stored on the LSU.

In DL mode, if the device can process/determine its own positioning using InvisiTrack technology, the frame start offset of the network antenna can be sent from the LSU to the UE/phone device. Otherwise, when the UE/handset device can periodically send time-stamped samples of the processed network RF signal to the LSU, the LSU can determine the device's position and send positioning data back to the device.

In DL mode, the wireless network RF signal can come from one or more wireless network antennas. To avoid the impact of multipath on the accuracy of the results, the RF signal should be found from the antenna or the antenna connection to the wireless network device.

The bidirectional mode covers determining positioning based on both UL and DL operations. This allows further improvement in positioning accuracy.

Some enterprise setups use one or more BBUs feeding one or more remote radio heads (RRHs), where each RRH in turn feeds multiple antennas with the same ID. In such environments, determining the DL mode frame start offset for the network antennas may not be necessary, depending on the wireless network configuration. This includes single BBU setups as well as multiple BBUs, where each BBU's antennas are assigned to a zone and the coverage of adjacent zones overlaps.

On the other hand, configurations whereby antennas fed from multiple BBUs are staggered in the same zone may require determining the DL mode frame start offsets of the network antennas.

In DL operation mode in a DAS environment, multiple antennas can share the same ID.

In an embodiment of the present invention, a position consistency algorithm is extended/developed to isolate the range of the DAS antenna from observables generated by multipath mitigation based on reference signal (including pilot and/or synchronization) subcarriers, and obtain a position estimate based on the ranges of multiple DAS transmitters (antennas).

However, these consistency algorithms have a limit on the number of antennas that transmit the same ID. It is possible to reduce the number of antennas that transmit the same ID by

1. For a given coverage area, stagger the antennas fed from different sectors of a sectorized BBU (a BBU can support up to six sectors)

2. For a given coverage area, interleave antennas fed from different sectors of a sectorized BBU and antennas fed from different BBUs

3. Add a propagation delay element to each antenna. The delay value is chosen to be large enough to exceed the delay spread in the specific DAS environment (channel), but smaller than the cyclic prefix (CP) length so that multipath caused by the extra delay does not cause ISI (inter-symbol interference). Adding a unique delay ID to one or more antennas can further reduce the number of antennas transmitting the same ID.

In an embodiment, an autonomous system can be provided without customer network investment. In this embodiment, the system can operate on frequency bands other than the LTE band. For example, the ISM (Industrial Scientific Medical) band and/or the White Space band can be used where LTE services are not available.

The embodiment can also be integrated with macro/micro/pico/femto stations and/or UE (mobile phone) devices. Although the integration may require customer network investment, it can reduce cost overhead and significantly improve TCO (total cost of ownership).

As mentioned above, PRS can be used by UEs for downlink observed time difference of arrival (DL-OTDOA) positioning. Regarding synchronization with neighboring base stations (eNBs), 3GPP TS 36.305 (Part 2: Functional Specification for User Equipment (UE) Positioning in E-UTRAN) specifies the transmission timing to the UE relative to the eNode B serving a candidate cell (e.g., a neighboring cell). 3GPP TS 36.305 also specifies the physical cell ID (PCI) and global cell ID (GCI) of the candidate cell for measurement purposes.

According to 3GPP TS 36.305, this information is delivered from an E-MLC (Enhanced Serving Mobile Location Centre) server. It should be noted that TS 36.305 does not specify the timing accuracy described above.

In addition, 3GPP TS 36.305 specifies that the UE shall return downlink measurement values to the E-MLC, which include Reference Signal Time Difference (RSTD) measurement values.

RSTD is a measurement made between a pair of eNBs (see TS 36.214 Evolved Universal Terrestrial Radio Access (E-UTRA); Physical Layer Measurements; Release 9). The measurement is defined as the relative timing difference between a subframe received from neighboring cell j and the corresponding subframe of serving cell i. These measurements are made using positioning reference signals. The results are fed back to a positioning server, which calculates the position.

In an embodiment, a hybrid approach may be defined to accommodate both the newly introduced PRS and existing reference signals. In other words, the hybrid approach may use/operate PRS, use/operate other reference signals (e.g., cell or node-specific reference signals (CRS)), or use/operate both signal types.

This hybrid approach offers the advantage of allowing network operators to dynamically select an operating mode depending on environmental or network parameters. For example, PRS has better audibility than CRS, but may result in up to a 7% reduction in data throughput. CRS signals, on the other hand, do not cause any throughput reduction. Furthermore, CRS signals are backward compatible with all previous LTE releases (e.g., Rel-8 and lower). Thus, the hybrid approach provides network operators with the ability to strike a balance between audibility, throughput, and compatibility.

In Long Term Evolution (LTE) implementations, LTE downlink baseband signals (generated by cells or wireless nodes and referred to herein as "nodes") are typically combined into downlink frames. A receiver for detecting and receiving such signals can detect downlink frames from multiple cells or nodes (two or more). Each downlink frame contains multiple CRSs or reference signals. Within a downlink (DL) frame, these reference signals have predetermined positions in time and frequency, e.g., there is a deterministic time offset between the start of a frame and each CRS in a given frame.

In addition, each CRS can be modulated with a unique code. The modulation and code are also predetermined. The CRS modulation is the same for all nodes, but the code (seed) is determined by the node's ID number (identification number).

Therefore, by knowing the node ID, it is possible to estimate the process position of the frame start time of each frame from each node (cell) in the spectrum of the reference signal. In order to do this, it may first be necessary to determine the frame start time or frame start of all DL signals from different nodes. For example, in an embodiment, by correlating the received DL baseband signal with a known copy of the code-modulated CRS (generated internally by the detector and/or multipath mitigation processor), it is possible to find all CRS sequences or other reference signals from various nodes, and find the rough position frame start of all observable nodes through this information. In an embodiment, the detector can also demodulate/decode the CRS and then correlate the demodulated/decoded CRS with the baseband subcarriers assigned to the CRS.

At the same time, in an embodiment, the CRS can also be used as a ranging signal by the multipath mitigation processor. Therefore, in addition to finding the coarse frame start, the detector's correlation processing can also isolate the CRS from other signals in the frame (e.g., payload) using the code used to modulate those signals. These isolated CRS and the associated frame start are then passed to the multipath mitigation processor for ranging.

A similar approach can be used in uplink mode, whereby the timing offset between different node receivers can be determined.

In a downlink embodiment, a system for tracking and locating one or more wireless network devices in communication with a network includes a user equipment receiver configured to receive multiple signals from two or more nodes in communication with the network, the multiple signals being modulated with a code determined by an identification of each of the two or more nodes transmitting the multiple signals, the user equipment receiver comprising: a detector configured to detect and isolate a reference signal from the multiple signals based on the identification; and a processor configured to track and locate the one or more wireless network devices using the reference signal as a ranging signal from each node.

In an embodiment, wherein the plurality of signals from each of the two or more nodes are combined into a frame comprising a reference signal, and wherein the detector is further configured to estimate a process position of a start of the frame from each node.

In an embodiment, wherein the detector is further configured to estimate the process position by correlating the reference signal with a known replica of such reference signal.

In an embodiment, wherein the detector is further configured to isolate the reference signal from any other signals in the frame, and wherein the detector is further configured to isolate the reference signal for each of the two or more nodes.

In an embodiment, wherein the processor is at least one multipath mitigation processor, and wherein the multipath mitigation processor is configured to receive the process position and isolation reference signals and estimate relative arrival times of ranging signals from each node.

In an embodiment, the processor is at least one multipath mitigation processor.

In an embodiment, wherein a plurality of signals from each of two or more nodes are in a frame, wherein the detector is further configured to estimate a process position of a start of the frame from each node, wherein the detector is configured to isolate a reference signal from any other signals in the frame, wherein the detector is further configured to isolate the reference signal for each of the two or more nodes, wherein the detector is configured to pass the process position and the isolated reference signal of each node to a multipath mitigation processor, and wherein the multipath mitigation processor is configured to receive the process position and the isolated reference signal and estimate a relative time of arrival of ranging signals from each node.

In an embodiment, the system further includes an uplink embodiment, wherein the node receiver is configured to receive a device signal from one or more wireless network devices, the device signal being modulated with a device code determined by a device identification of each of the one or more wireless network devices transmitting the device signal, the node receiver comprising: a device detector configured to detect and isolate a device reference signal from the device signal based on the device identification; and a second processor configured to use the device reference signal as a ranging signal from each wireless network device to track and locate the one or more wireless network devices.

In an embodiment, a system for tracking and locating one or more wireless network devices in communication with a network includes: a user equipment receiver configured to receive multiple signals from two or more nodes in communication with the network, the multiple signals being modulated with a code determined by an identification of each of the two or more nodes transmitting the multiple signals; and a processor configured to detect and isolate a reference signal from the multiple signals based on the identification, and to use the reference signal as a ranging signal from each node to track and locate the one or more wireless network devices.

In an embodiment, wherein the plurality of signals from each of the two or more nodes are combined into a frame comprising a reference signal, and wherein the processor is further configured to estimate a process location of a start of the frame from each node.

In an embodiment, wherein the processor is further configured to estimate the process position by correlating the reference signal with a known replica of such reference signal.

In an embodiment, wherein the processor is further configured to estimate a relative time of arrival of the ranging signal from each node based on the process position and the isolated reference signal.

In an embodiment, wherein the processor is further configured to isolate the reference signal from any other signals in the frame, and wherein the processor is further configured to isolate the reference signal for each of the two or more nodes.

In an embodiment, wherein a plurality of signals from each of two or more nodes are in a frame, wherein the processor is further configured to estimate a process position of the start of the frame from each node by correlating a reference signal with a known replica of the reference signal, wherein the processor is further configured to isolate the reference signal from any other signals in the frame and to isolate the reference signal for each of the two or more nodes, wherein the processor is further configured to estimate a relative time of arrival of ranging signals from each node based on the process position and the isolated reference signal.

In an embodiment, a system for tracking and locating one or more wireless network devices communicating with a network includes: a detector configured to receive multiple signals from two or more nodes communicating with the network, the multiple signals being modulated with a code determined by an identification of each of the two or more nodes transmitting the multiple signals, the detector further configured to detect and isolate a reference signal from the multiple signals based on the identification; and a processor configured to track and locate the one or more wireless network devices using the reference signal as a ranging signal from each node.

In an embodiment, wherein the plurality of signals from each of the two or more nodes are combined into a frame comprising a reference signal, and wherein the detector is further configured to estimate a process position of a start of the frame from each node.

In an embodiment, wherein the detector is further configured to estimate the process position by correlating the reference signal with a known replica of such reference signal.

In an embodiment, wherein the detector is further configured to isolate the reference signal from any other signals in the frame, and wherein the detector is further configured to isolate the reference signal for each of the two or more nodes.

In an embodiment, wherein the processor is at least one multipath mitigation processor, and wherein the multipath mitigation processor is configured to receive the process position and isolation reference signals and estimate relative arrival times of ranging signals from each node.

In an embodiment, the processor is at least one multipath mitigation processor.

In an embodiment, wherein a plurality of signals from each of two or more nodes are in a frame, wherein the detector is further configured to estimate a process position of a start of the frame from each node, wherein the detector is configured to isolate a reference signal from any other signals in the frame, wherein the detector is further configured to isolate the reference signal for each of the two or more nodes, wherein the detector is configured to pass the process position and the isolated reference signal of each node to a multipath mitigation processor, and wherein the multipath mitigation processor is configured to receive the process position and the isolated reference signal and estimate a relative time of arrival of ranging signals from each node.

In an embodiment, a system for tracking and locating one or more wireless devices communicating with a network includes a node receiver configured to receive a device signal from one or more wireless network devices, the device signal being modulated with a device code determined by a device identification of each of the one or more wireless network devices transmitting the device signal, the node receiver comprising: a device detector configured to detect and isolate a device reference signal from the device signal based on the device identification; and a processor configured to use the device reference signal as a ranging signal from each wireless network device to track and locate the one or more wireless network devices.

Furthermore, the hybrid approach can be transparent to the LTE UE positioning architecture. For example, the hybrid approach can operate within the 3GPP TS 36.305 framework.

In an embodiment, RSTD may be measured and communicated from the UE to the E-SMLC according to 3GPP TS 36.305.

UL-TDOA (U-TDOA) is currently in the research stage and is expected to be standardized in the upcoming Release 11.

Embodiments of UL-TDOA (uplink) are described above and also illustrated in Figures 16 and 17. Figures 18 and 19 described below provide examples of alternative embodiments of UL-TDOA.

Figure 18 presents an environment that may include one or more DAS and/or femto/small cell antennas. In this example embodiment, each NSAU is equipped with a single antenna. As depicted, at least three NSAUs are required. However, in embodiments, additional NSAUs may be added to improve hearability, as each UE must be "heard" by at least three NSAUs.

Furthermore, the NSAUs can be configured as receivers. For example, each NSAU receives information over the air but does not transmit it. In operation, each NSAU can listen to wireless uplink network signals from a UE. Each of the UEs can be a handset, an accessory device, and/or another UE device.

Furthermore, the NSAU can be configured to communicate with a location server unit (LSU) via an interface (e.g., a wired service or LAN). In turn, the LSU can communicate with a wireless or LTE network. Communication can be via a network API, where the LSU can, for example, communicate with an E-SMLC of an LTE network and utilize wired services such as a LAN and/or WAN.

Optionally, the LSU can also communicate directly with the DAS base station and or femto/small cell. Such communication can use the same API or a modified network API.

In this embodiment, a sounding reference signal (SRS) may be used for positioning purposes. Of course, other signals may also be used.

The NSAU may convert the UE uplink transmission signal into a digital format, such as I/Q samples, and may periodically send multiple converted signals to the LSU using a timestamp.

The DAS base station and/or femto/small cell may pass one or all of the following data to the LSU:

1) SRS, I/Q samples and timestamps;

2) List of served UE IDs; and

3) SRS scheduling for each UE with UE ID, the scheduling includes SRS SchedulingRequestConfig information and SRS-UL-Config information.

The information passed to the LSU may not be limited to the above information. It may include any information required to correlate each UE device uplink signal (eg, UE SRS) with each UE ID.

The LSU functions may include ranging calculations and obtaining the UE's position. These determinations/calculations may be based on information passed to the LSU from the NSAU, DAS base station, and/or femto/small cell.

The LSU may also determine a timing offset from available downlink transmission information passed from the NSAU to the LSU.

In turn, the LSU can provide UE positioning and other calculations and data to the wireless or LTE network. This information can be transmitted via the network API.

For synchronization purposes, each NSAU can receive, process, and time-stamp samples of the downlink signal. Each NSAU can also periodically send multiple such samples to the LSU, including the time stamp.

Additionally, each NSAU may include an input configured for synchronization with an external signal.

FIG19 illustrates another embodiment of UL-TDOA. In addition to the components depicted in FIG18 , the environment of this embodiment may also include one or more cell towers, which may be used in place of DAS base stations and/or femto/small cells. Data from the one or more cell towers may be used to obtain the UE's position.

Thus, advantages of this embodiment include obtaining positioning with a single cell tower (eNB). Additionally, this embodiment can be configured to operate in a manner similar to that described in FIG. 18 , except that one or more eNBs can replace DAS base stations and/or femto/small cells.

One method for uplink positioning of a UE is the cell identification method (CID). In the basic CID method, the UE position can be determined at the cell level. This method is purely network-based. Therefore, the UE (e.g., a handset) is unaware of the fact that it is being tracked. While this is a relatively simple method, it lacks accuracy because the positioning uncertainty is equal to the cell diameter. For example, as shown in FIG20 , any handset 2000 within the cell diameter 2002 of a serving cell tower 2004 effectively has the same position, even if they are not in the same location. The accuracy of the CID method can be improved when combined with knowledge of the serving sector identification (sector ID). For example, as shown in FIG21 , sector ID 2100 identifies a portion 2102 within the cell diameter 2002 that contains multiple handsets 2104, which are known to have different locations than other handsets 2000 in other sectors of the cell diameter 2002.

The CID method can be further enhanced through an enhanced cell ID (E-CID) method, which further improves on the basic CID method described above. One enhancement uses timing measurements to calculate how far the UE is from the eNB (network node). This distance can be calculated as half the round-trip time (RTT), or the timing advance (TA) in LTE (LTE TA) multiplied by the speed of light. If the UE is connected, the RTT or TA can be used for distance estimation. In this case, the serving cell tower or sector and the UE (when commanded by the serving eNB) can measure the timing difference between the Rx subframe and the Tx subframe. The UE can report its measurements to the eNB (also under eNB control). It should be noted that LTE Release 9 added TA type 2 measurements, which rely on timing advance estimated from receiving PRACH preambles during the random access procedure. The PRACH (Physical/Packet Random Access Channel) preamble specifies the maximum number of preambles sent during a PRACH ramp-up cycle when no response is received from the tracked UE. LTE Type 1 TA measurements are equivalent to RTT measurements and are as follows:

RTT=TA(Type 1)=eNB(Rx-Tx)+UE(Rx-Tx)

With knowledge of the coordinates of the eNB and the height of the serving cell signal tower antenna, the UE's location can be calculated through the network.

However, the E-CID positioning method is still limited because positioning accuracy depends on the sector width and distance from the serving cell tower in one dimension, while the error in the other dimension depends on the TA (RTT) measurement accuracy. Sector width varies with network topology and is affected by propagation phenomena (specifically, multipath). Sector accuracy estimates vary from 200 meters to over 500 meters. The resolution of LTE TA measurements is 4Ts, which corresponds to a maximum error of 39 meters. The actual error of LTE TA measurements is even greater, however, due to calibration inaccuracies and propagation phenomena (multipath), the actual error can reach up to 200 meters.

As shown in Figure 22, the E-CID method can be further improved by incorporating a feature known as the Angle of Arrival (AoA). The eNB estimates the direction from which the UE is transmitting using a linear array of equally spaced antenna elements 2200. Typically, AoA determination is performed using a reference signal. When a reference signal is received from a UE at two adjacent antenna elements 2200, the reference signal can be phase-rotated, as shown in Figure 23. The amount of phase rotation depends on the AoA, carrier frequency, and element spacing. AoA may require each eNB to be equipped with an antenna array/adaptive antenna. It is also susceptible to multipath and topology variations. However, a precision antenna array can significantly reduce the width 2202 of the sector 2100, which can result in better positioning accuracy. Furthermore, if two or more serving cell towers 2300 (eNB base stations equipped with directional antenna arrays) can be used for handset AoA determination, as shown in Figure 23, accuracy can be significantly improved. In this case, accuracy is still limited by multipath/propagation phenomena.

Deploying antenna arrays/adaptive antenna networks across multiple LTE bands is prohibitively expensive in terms of funding, time, and maintenance. Consequently, antenna arrays/adaptive antennas have not yet been deployed for UE positioning purposes. Other methods, such as those based on signal strength, have not yielded significant accuracy improvements. One such signal strength method is fingerprinting, which requires the creation and continuous updating of a large, ever-changing (over time) fingerprint database (e.g., incurring significant funding and ongoing costs) without yielding significant accuracy improvements. Furthermore, fingerprinting is a UE-based technology, making it impossible to determine UE location without UE assistance at the UE application level.

A solution to the limitations of other uplink positioning methods includes using AoA functionality without the need for antenna arrays/adaptive antennas. Such embodiments may employ TDOA (Time Difference of Arrival) positioning techniques for AoA determination, which may be based on estimating the difference in arrival time of signals from a source at multiple receivers. The specific time difference estimate defines a hyperbola between the two receivers communicating with the UE. When the distance between the receive antennas is small relative to the distance at which the transmitter (handset) is located, TDOA is equivalent to the angle between the baseline of the sensor (receiver antenna) and the incident RF energy from the transmitter. If the angle between the baseline and true north is known, then the line of bearing (LOB) and/or AoA can be determined.

While general positioning methods using TDOA or LOB (also known as AoA) are known, TDOA positioning methods have not been used to determine LOB because the TDOA reference points are too close to each other to make the accuracy of one technique unacceptable. In practice, LOB is typically determined using directional antennas and/or beamforming antennas. However, the super-resolution methods described herein make it possible to use TDOA for LOB determination while significantly improving accuracy. In addition, without the reference signal processing techniques described herein, it may not be possible to "hear" (e.g., detect) reference signals from UEs outside the serving sector (e.g., via non-serving sectors and/or antennas). Without the resolution and processing capabilities described herein, it may not be possible to employ TDOA for LOB determination because at least two reference points, e.g., two or more sectors and/or antennas, are required. Similarly, a UE may not be able to detect reference signals arriving at the UE from a sector other than the serving sector (e.g., from non-serving sectors and/or antennas).

For example, in FIG24 , two antenna spacing scenarios are illustrated: wide spacing and close (small) spacing. In both scenarios, hyperbola 2400 and incident ray 2402 intersect at the location of handset 2000, but in the case of wide spacing of antennas 2404, this occurs at a steeper angle, which in turn significantly reduces positioning errors. At the same time, in the case of close proximity of antennas 2404, hyperbola 2400 becomes interchangeable with RF energy incident ray 2402 or LOB/AoA.

The formula listed below can be used to determine the incident RF energy from the transmitter, where the time difference in the arrival time of the RF energy between the two antennas (sensors) is given by the following formula:

in:

Δt is the time difference in seconds;

x is the distance between the two sensors in meters;

Θ is the angle in degrees between the baseline of the sensor and the incident RF wave; and

c is the speed of light.

Several positioning strategies can be obtained by using TDOA positioning embodiments, including: (1) when TDOA measurements between two or more serving cells (multilateration) are available, for example, wide spacing; (2) when TDOA measurements are from two or more sectors at one or more serving cells, for example, small antenna spacing, such as LOB/AoA; (3) a combination of strategies (2) and (3); and (4) a combination of TA measurements with strategies (1) to (3), for example, improved E-CID.

As further explained below, in the case of closely located antennas, TDOA positioning embodiments can use lines of bearing when the signals from two or more antennas are from the same cell tower. These signals can be detected in the received composite signal. By knowing the tower location and the azimuth angle of each sector and/or antenna, lines of bearing and/or AoA can be calculated and utilized in the positioning process. LOB/AoA accuracy can be affected by multipath, noise (SNR), etc. However, this effect can be mitigated through advanced signal processing that can be based on super-resolution techniques and the multipath mitigation processing techniques described above. Such advanced signal processing includes (but is not limited to) signal correlation/correlation, filtering, averaging, synchronous averaging, and other methods/techniques.

A cell tower 2500 is typically composed of multiple sectors, as shown in FIG. 25 , which illustrates a three-sector configuration (sector A, sector B, and sector C). The illustrated three-sector deployment may include one or more antennas 2502 per sector. A single sector (e.g., sector A) can be controlled by a UE (handset) because the handset's transmissions may be within the main lobe of sector A (the center of the main lobe aligns with the sector's azimuth). Meanwhile, the handset's transmissions may fall outside the main lobes of sectors B and C, for example, into the antenna's side lobes. Therefore, the handset's signal may still be present in the output signal spectrum of sectors B and C, but may be significantly attenuated relative to signals from other handsets located in the main lobes of sectors B or C. Nevertheless, by using advanced signal processing as described above and below, it is possible to achieve sufficient processing gain for ranging signals, enabling them to be detected from the side lobes of neighboring sectors (e.g., the side lobes of sectors B and C). For network-based positioning purposes, the LTE uplink SRS (Sounding Reference Signal) may be employed as the ranging signal.

In other words, even though the UE uplink reference signal may be in the sidelobes of adjacent sector antennas, the processing gain achieved by the reference signal processing method described herein can be sufficient to allow calculation of the TDOA between two (or more) sector antennas. The accuracy of this embodiment can be significantly enhanced by the multipath mitigation processing algorithm described above. Therefore, the LOB/AOA intersecting the circular domain calculated using the LTE TA timing can provide UE positioning within an error ellipse of approximately 20 meters by 100 meters.

Further positioning error reduction can be achieved when the UE can be heard by two or more LTE towers (which is very likely when using the processing gain and multipath mitigation techniques described above). In this case, the intersection of the TDOA hyperbola with one or more LOB/AoA lines can produce an error ellipse of 30 by 20 meters (for cell towers of two sectors). If each cell tower supports three or more sectors, the error ellipse can be further reduced to 10 to 15 meters. If the UE is heard by three or more eNBs (cell towers), an accuracy of 5 to 10 meters can be achieved. In high-value areas, such as shopping malls, office buildings, etc., additional small cells or passive listening devices can be used to form coverage.

As mentioned above, each sector of the cell tower 2500 may contain one or more antennas 2502. In a typical installation, for a given sector, the signals from each antenna are combined at the receiver input of that sector. Thus, for positioning purposes, two or more sector antennas may be treated as a single antenna with a composite directivity profile, azimuth, and elevation. The sector itself may also be assigned an assumed antenna composite directivity, as well as its (main lobe) azimuth and elevation.

In an embodiment, received signals (in digital format) from all sectors of each serving cell tower and neighboring serving cell towers are sent to a location server unit (LSU) for position determination. In addition, SRS scheduling and TA measurements for each served UE are provided to the LSU by each serving sector from each serving cell tower. Assuming the location coordinates of each serving cell tower and each neighboring cell tower, the number of sectors per tower with their assumed (composite) sector antenna azimuth and elevation, and the location of each sector at the cell tower, the LSU can determine the location of each UE relative to the serving cell tower and/or neighboring cell towers. All of the above information can be sent over a wired network (e.g., LAN, WAN, etc.) using one or more standardized or proprietary interfaces. The LSU can also interface with wireless network infrastructure using standardized interfaces and/or network operator-defined interfaces/APIs. Position determination can also be split between network nodes and the LSU or performed solely in the network nodes.

In an embodiment, location determination may be performed in the UE or split between the UE and an LSU or network node. In such cases, the UE may communicate over the air using standard network protocols/interfaces. Alternatively, location determination may be performed by a combination of the UE, LSU, and/or network node, or the LSU functionality may be implemented (embedded) in a SUPL server, an E-SMLC server, and/or an LCS (Location Service) system, which may then be used in place of the LSU.

An embodiment of the downlink (DL) positioning method is opposite to the uplink (UL) positioning embodiment described above. In the DL embodiment, the sector can become a transmitter with a transmission line shape, azimuth and altitude that matches the receive directivity, azimuth and altitude of the sector. Unlike the uplink embodiment, in the DL embodiment, the UE typically has a single receive antenna. Therefore, there is no sensor baseline for the UE that can be used to determine the incidence of RF waves. However, the UE can determine the TDOA between different sectors and therefore the hyperbola between sectors (multilateration), and because the same cell tower sectors are close to each other, the hyperbola becomes interchangeable with the RF energy incident line or LOB/AoA, as described above with reference to Figure 24. Although LOB/AoA accuracy may be affected by multipath, noise (SNR), etc., such effects can be suppressed by using advanced signal processing and multipath suppression processing based on the super-resolution technology described above.

As mentioned, UE DL positioning can be done in a similar manner to UE uplink positioning, except that the RF wave incidence angle cannot be determined according to the above formula. Instead, multilateration techniques can be used to determine the LOB/AoA of each serving cell tower.

UE DL positioning embodiments also employ reference signals. In the DL case, one approach for such network-based positioning can be to use the LTE cell-specific reference signal (CRS) as a ranging signal. Alternatively, the position reference signal (PRS) introduced in LTE Release 9 can be used. Thus, positioning can be accomplished using CRS, PRS, or both.

As with the UE uplink positioning embodiment, for the UE downlink positioning embodiment, a snapshot of the UE received signal in digital format can be sent to the LSU for processing. The UE can also obtain TA measurements and provide them to the LSU. Optionally, the TA measurements for each served UE can be provided to the LSU by each serving sector from each serving cell tower (network node). As previously described, assuming that the location coordinates of each serving cell tower and each neighboring cell tower, the number of sectors of each tower with their respective sector transmission line azimuth and altitude, and the location of each sector at the cell tower are known, the LSU can determine the location of each UE relative to the serving cell tower and/or neighboring cell tower. In an embodiment, location determination can be performed in the UE or split between the UE and the LSU or network node. In an embodiment, all location determination can be performed in the LSU or network node or split between the two.

The UE can transmit/receive measurements and other information over the air using standard wireless protocols/interfaces. Information exchange between the LSU and network nodes can be performed over a wired network (e.g., a LAN, WAN, etc.) using proprietary and/or one or more standardized interfaces. The LSU can interface with the wireless network infrastructure using standardized interfaces and/or network operator-defined interfaces/APIs. Position determination can also be split between the network node and the LSU or performed solely in the network node.

For the UE DL positioning embodiment described above, antenna port mapping information can also be used to determine the position. The 3GPP TS 36.211 LTE standard defines the antenna ports used for DL. The LTE standard defines different reference signals (pilot signals) for each antenna port. Therefore, the DL signal also carries antenna port information. This information is included in the PDSCH (Physical Downlink Shared Channel). The PDSCH uses the following antenna ports: 0; 0 and 1; 0, 1, 2, and 3; or 5. These logical antenna ports are assigned (mapped) to physical transmit antennas, as shown in Figure 26. Therefore, this antenna port information can be used for antenna identification (antenna ID).

For example, antenna port mapping information can be used to determine RF wave incidence and hyperbolic measurements between antennas (assuming known antenna positions). Depending on where the position determination is performed, antenna mapping information must be available to the LSU, UE, or network node. It should be noted that the antenna ports are indicated by placing CRS signals in different time slots and different resource elements. One CRS signal can be transmitted per DL antenna port.

In the case of a MIMO (Multiple Input Multiple Output) deployment in an eNB or network node, the receiver may be able to determine the time difference of arrival from a given UE. Knowing the antenna to receiver mapping, e.g., the MIMO mapping, including the antenna position, it is also possible to determine the RF wave incidence (LOB/AoA) to the antenna and the hyperbola (multilateration) for a given eNB antenna. Similarly, at the UE, the UE receiver may be able to determine the time difference of arrival from two or more eNBs or network nodes and MIMO antennas. Knowing the eNB antenna position and the antenna mapping, it is possible to determine the RF wave incidence (LOB/AoA) from the antenna and the hyperbola (multilateration) for a given eNB antenna. Depending on where the position determination is performed, the antenna mapping information must be available to the LSU or the UE or the network node.

There are other configurations that are subsets of MIMO, such as single-input multiple-output (SIMO), single-output multiple-input (SOMI), single-input single-output (SISO), etc. All of these configurations can be defined/determined by antenna port mapping and/or MIMO antenna mapping information for positioning purposes.

In one aspect, embodiments of the present invention relate to methods and systems for RF-based identification, tracking, and location (including RTLS) of objects. According to one embodiment, the methods and systems employ geographically distributed clusters of receivers and/or transmitters that are precisely synchronized in time, e.g., within 10 ns or better within each cluster, although inter-cluster time synchronization may be less precise or not required at all. While a precise synchronization time of 10 ns or better is described with respect to one specific embodiment, it is important to note that the predetermined synchronization time required to achieve accurate location depends on the devices being utilized. For example, for some wireless system devices where 3 m accuracy is required for accurate location determination, the predetermined time may need to be 10 ns or better, while with other wireless system devices, 50 m location accuracy may be more than sufficient. Thus, the predetermined time is based on the desired location accuracy for the wireless system. The disclosed methods and systems represent a significant improvement over existing implementations of tracking and locating DL-OTDOA and U-TDOA technologies, which rely on geographically distributed independent (individual) transmitters and/or receivers.

For example, in DL-OTDOA technology, the relative timing difference between signals from neighboring base stations (eNBs) is calculated, and the UE position can be estimated in a network with a UE (handset) with or without UE assistance, or in a UE (handset) with network assistance (with a control plane or user plane based only on SUPL) or without network assistance. In DL-OTDOA, upon receiving signals from three or more base stations, the UE measures the relative timing difference between the signals from a pair of base stations and generates a hyperbolic position line (LOP). At least three reference points (the base stations do not belong to the straight line) are required to define two hyperbolas. The position (positioning) of the UE is at the intersection of these two hyperbolas (see Figure 11). The UE positioning is related to the (antenna) position of the RF transmitter of the base station. For example, when using LPP (LTE Positioning Protocol, Release 9), DL-OTDOA positioning is UE-assisted, and the E-SMLC (Evolved Serving Mobile Positioning Center) is server-based.

U-TDOA technology is similar to DL-OTDOA, but with the roles reversed. Here, a nearby location management unit (LMU) calculates the relative time of arrival of uplink signals from the UE (handset) and can estimate the UE's position in the network without UE assistance. Therefore, U-TDOA is LMU-assisted, and the E-SMLC (Evolved Serving Mobile Location Center) is server-based. Once relative time of arrival values from three or more LMUs are available, the network's E-SMLC server generates a hyperbolic line of position (LOP) and the UE's position (position) (see Figure 27). UE positioning is correlated with the LMU antenna position. In one aspect, unlike DL-OTDOA, time synchronization with the eNB (base station) may not be required in the case of U-TDOA; only the LMU may require precise time synchronization for positioning purposes. For example, the LMU is essentially a receiver with computing capabilities. As another example, the LMU receiver utilizes SDR (Software Defined Radio) technology. In another example, the LMU can be a small cell, macro cell, or a dedicated small cell type device that only performs reception.

Regardless of the implementation, correlating the positioning of the SRS to a specific UE as provided by the network can enable identification and positioning of the UE. Positioning of the SRS can be performed at the network level or within a local sector (e.g., a DAS of a building, a small cell, or a combination of a small cell and a macro cell serving a specific area). If the positioning of the SRS for a UE is not known a priori, a solution may be able to correlate the positioning of the UE throughout the coverage area. Doing so can show the location history of the UE as it has traveled. In some cases, it may be necessary to determine the location of the UE even if the network does not provide an indication of where the SRS is located for a specific UE. By determining the location or proximity of the UE to a known point, the UE's location can be correlated with the SRS, thereby correlating the UE with the SRS it transmits. Such positioning can be accomplished through other location/proximity solutions (e.g., Wi-Fi and Bluetooth). Users can also determine their location via a UE application or by walking to a predetermined location in order to determine their UE to a location solution.

A macro base station is shown in Figures 11 and 27. In addition, Figure 27 depicts an LMU co-located with the base station. These depictions are valid options, but the LTE standard does not specify where the LMU can be placed, as long as the LMU placement complies with multilateration/trilateration requirements.

In one aspect, common deployments for indoor environments are DAS (Distributed Antenna Systems) and/or small cells, which are inexpensive base stations with high RF integration. LMUs can also be placed in indoor and/or campus-type environments, for example, U-TDOA can be used in DAS and/or small cell environments. In another aspect, accurate indoor positioning based on U-TDOA can be obtained by a combination of LMUs positioned indoors and macro cells positioned outside, for example, without the need to deploy DAS and/or small cells; or with a reduced number of small cells. Therefore, LMUs can be deployed with or without the presence of DAS and/or small cells. On the other hand, LMUs can be placed in environments where cellular signal amplifiers/boosters are used; with or without the presence of DAS and/or small cells.

LTE Release 11 also contemplates integrating the LMU and eNB into a single unit. However, this may place an additional burden on inter-small cell time synchronization requirements when individual small cell eNBs are geographically distributed, which wireless/cellular service providers are not prepared to meet, especially indoors and/or in other GPS/GNSS-denying environments.

DAS systems are inherently time-synchronized to a much higher degree (precision) than geographically distributed macro/micro/small cells/LMUs. Using a DL-DTOA solution in a DAS environment can alleviate time synchronization issues. However, in a DAS environment, a single base station serves a large number of distributed antennas, enabling multiple antennas to transmit the same downlink signal with the same cell ID (identification number). Therefore, traditional DL-OTDOA methods are ineffective because there are no identifiable neighboring cells (antennas) generating signals with different IDs. Nonetheless, it is possible to use DL-OTDOA technology by employing a multipath mitigation processor and multipath mitigation techniques/algorithms such as those described in U.S. Patent No. 7,872,583, and by extending the use of location consistency algorithms such as those described in U.S. Non-Provisional Application No. 13/566,993, filed on August 3, 2012, entitled “MULTI-PATH MITIGATION IN RANGEFINDING AND TRACKING OBJECTS USING REDUCED ATTENUATION RF TECHNOLOGY,” which are incorporated herein by reference in their entirety. However, these consistency algorithms impose a limit on the number of antennas transmitting with the same ID. One solution is to reduce the number of antennas transmitting with the same ID, for example, by splitting a large number of DAS antennas into two or more time-synchronized clusters with different IDs. Such an arrangement increases system cost (increasing the number of base stations) and requires the handset/UE to support the aforementioned techniques.

Adopting U-TDOA in a DAS environment also increases the cost associated with adding/installing LMU units. However, no changes to the UE (handset) may be required; only the base station software may have to be upgraded to support U-TDOA functionality. In addition, it is possible to integrate multiple LMUs with (into) the DAS system. Therefore, using the U-TDOA method with LMUs has many advantages when used indoors, in campus environments, and in other geographically constrained environments where GPS/GNSS is challenging.

Precise time synchronization between multiple geographically distributed base stations and/or small cells and/or LMUs in indoor and other GPS/GNSS-intolerant environments is more complex than synchronizing time between macrocells and/or LMUs used outdoors (e.g., in GPS/GNSS-friendly environments). This is because macrocells in outdoor environments have elevated and exposed antennas. Consequently, GPS/GNSS signal quality is very good, and GPS/GNSS can be used to synchronize macrocell antenna transmissions and/or LMU receivers to very high accuracy, with a standard deviation of 10ns, over a sufficiently large area.

In one aspect, for indoor and other GPS/GNSS-incompatible environments, time synchronization between multiple distributed base stations and/or small cells/LMUs is achieved through the use of an external synchronization source that generates a synchronization signal shared by the base stations and/or small cells and/or LMUs. This synchronization signal can be derived from GPS/GNSS, such as the IPPS signal, and/or from the Internet/Ethernet, such as PTP or NTP. The latter is a low-cost solution, but it does not provide the time synchronization precision required for precise positioning. External synchronization signals derived from GPS/GNSS are more precise, with standard deviations down to 20ns, but require additional hardware requirements; for example, wiring these signals together can be more complex/expensive. Furthermore, changes to the base station and/or small cell hardware/low-level firmware may be required to achieve a higher level of precision for the external synchronization signal. Furthermore, a standard deviation of 20ns is insufficient to precisely meet the 3-meter requirement; for example, a standard deviation of approximately 10ns is sufficient.

To overcome the aforementioned limitations, as illustrated by the multi-channel LMU high-level block diagram of FIG. 28 , one embodiment utilizes an LMU device 2800 having multiple receive antennas 2802 and signal channels 2804. For example, one or more signal channels 2804 may include signal processing components such as an RFE (RF front end) 2806, an RF downconverter 2808, and/or an uplink-positioning processor 2810. Other components and configurations may be used. In one aspect, the signal channels 2804 are co-located within the LMU device 2800 and tightly time-synchronized (e.g., with a standard deviation of approximately 3 ns to approximately 10 ns). In another example, the antennas 2802 from each LMU signal channel 2804 are geographically distributed (e.g., similar to a DAS). As another example, an external time synchronization component (e.g., GPS/GNSS, Internet/Ethernet, etc.) may communicate with the LMU device 2800. Precise time synchronization is more easily achieved within a device (e.g., LMU device 2800) than attempting to tightly synchronize multiple geographically distributed devices.

For example, when two or more multi-channel LMUs (e.g., LMU device 2800) are deployed, the time synchronization between these LMUs can be relaxed so that a low-cost and low-complexity method (using an external source signal) can be used to synchronize multiple distributed multi-channel LMUs. For example, Internet/Ethernet synchronization can be used or a common sensor (device) can be deployed to provide timing synchronization between different multi-channel LMUs.

On the other hand, the multi-channel LMU approach reduces the number of hyperbolic lines of position (LOPs) that can be used when determining a position fix, but improvements in time synchronization can overcome this shortcoming (see explanation and examples below).

When using multilateration/trilateration methods, UE positioning accuracy depends on two factors: the geometric dilution of precision (GDOP), which is due to the geometric arrangement of macrocell towers/small cells/LMUs; and the accuracy of a single ranging σ _{R_pseudo} measurement (see Günter Seeber, Satellite Geodesy, 2003):

σ _POS = GDOP × σ _{R_pseudo}

The GDOP depends on the geographic distribution of the transmit antennas (in the case of DL-OTDOA) or receive antennas (in the case of U-TDOA). In the case of regularly placed antennas, the two-dimensional GDOP estimate is equal to 2/√N (H.B. Lee, "Accuracy Limitations of Hyperbolic Multilateration Systems," 1973), where, in the case of cellular networks, N is the number of transmitters (macrocell towers/small cells/DAS antennas) that the UE can "hear" (in the case of DL-OTDOA) or the number of LMUs/LMU receive channels that can "hear" the UE's uplink transmissions (in the case of U-TDOA). Therefore, the standard deviation of the UE position error can be calculated as follows:

Assume that eight geographically distributed (indoor) single-receive channel LMUs (regularly placed) are detecting UE uplink transmissions, and that these LMUs are synchronized via a 1PPS signal (e.g., 20ns standard deviation). In this case, N=8, and there may be seven independent LOPs that can be used for UE positioning. Further assume that the ranging error standard deviation σ _R is 3 meters (about 10ns); then the accuracy of a single ranging measurement is:

where σ _SYNC is the standard deviation of the external time synchronization signal (20ns).

In this case (N=8), the single ranging measurement and standard deviation of the UE position error σ _POS is equal to 4.74 meters.

As an example, if two or four receive channel LMUs (e.g., multi-channel LMU device 2800) with regularly placed distributed antennas are detecting UE uplink transmissions, each LMU can generate a set of three tightly time-synchronized LOPs (e.g., with a standard deviation of approximately 3 ns), with N = 4 for three independent LOPs. In this case, two UE position fixes are generated, each with a standard deviation error σ _POS of 3.12 meters. The UE position error can be further reduced by averaging and/or combining these two fixes in other ways/methods. One estimate is that the error reduction is proportional to the square root of the number of UE position fixes. In the present invention, this number is equal to 2, and the final UE position fix error σ _{POS_FINAL} is 2.21 meters, obtained as .

In one aspect, for indoor and other GPS/GNSS-denying environments, multiple multi-channel LMUs (e.g., LMU assembly 2800) can be used with relaxed synchronization between them. For example, within a multi-channel LMU assembly, the LMUs can be tightly synchronized (e.g., with a standard deviation between approximately 3 ns and approximately 10 ns). Another embodiment utilizes the fact that multiple single-channel small cells/LMUs and/or small cells with integrated LMU assembly electronics (LMU functionality embedded in the eNB) can be clustered (e.g., consolidated, co-located, etc.) in a rack-mount housing (Figures 31, 32, and 33) and/or cabinet (e.g., a 19-inch rack). Each single-channel assembly antenna can be geographically distributed, as in a DAS. The units within the cluster can be tightly time-synchronized (e.g., with a standard deviation of less than or equal to 10 ns). Multiple rack-mount housings can be synchronized based on communication requirements (e.g., VoLTE), thereby enabling a low-cost and low-complexity approach. Precise (tight) time synchronization between multiple devices clustered (consolidated) within a rack-mount enclosure/cabinet is easier and less expensive to achieve than tightly time synchronization between multiple geographically distributed devices.

In another aspect, multiple LMUs can be integrated with (or into) a DAS system, as shown in FIG34 . For example, the LMU receivers can share the received signals generated by each DAS antenna, e.g., shared DAS antennas. The actual distribution of these received signals depends on the DAS implementation: active DAS versus passive DAS. However, embodiments of LMU and DAS integration share the received signals generated by each DAS antenna across the LMU receiver channels and form an almanac that matches (correlates) each DAS antenna coordinate with the corresponding LMU/LMU receiver channel. Similarly, clustering and/or employing multi-channel LMUs are preferred approaches for LMU and DAS integration.

Similarly, it is possible to share the received signal generated by each small cell antenna across the LMU receiver channel. Here, small cell time synchronization can be relaxed; for example, positioning requirements may not be met, but precise time synchronization of the LMUs/LMU channels may be required. For this option, clustering and/or the use of multi-channel LMUs are preferred approaches for LMUs.

Integration of the LMU and eNB into a single unit offers cost advantages over combining separate eNB and LMU devices. However, unlike the integrated LMU and eNB receiver, the separate LMU receive channel does not have to process the data payload from the UE. Furthermore, because the UE uplink ranging signals (SRS, Sounding Reference Signal, in the case of LTE) can be repeated and time-synchronized (with the serving cell), each separate LMU receive channel can support two or more antennas (time-division multiplexed with them), for example, to serve two or more small cells. This, in turn, can reduce the number of LMUs (in small cell/DAS and/or other U-TDOA positioning environments) and lower system costs (again, see Figure 28).

If the wireless/cellular network E-SMLC server does not have the functionality required for DL-OTDOA and/or U-TDOA technology, this functionality can be implemented by a positioning server that can communicate with the UE and/or LMU and the wireless/cellular network infrastructure and/or positioning service server (see Figures 29 and 30). Other configurations can be used.

In another aspect, one or more LMU devices (e.g., LMU 2802) can be deployed with WiFi infrastructure, for example, as shown in FIG35. Alternatively, a listening device can be used to monitor the LMU antenna in the same manner as the WiFi infrastructure. Thus, the LMU device and/or the channel antenna serving the LMU can be co-located with one or more WiFi/listening devices 3500 (e.g., one or more WiFi access points (APs)). For example, the WiFi devices 3500 can be geographically distributed.

In one embodiment, the WiFi device 3500 can be connected to a power source. The RF analog portion 3502 (e.g., circuitry) of one or more LMU devices or channels can be integrated with the LMU antenna, allowing the RF analog portion 3502 to share a power source with the WiFi device 3500 (see FIG. 35 ). For example, the RF analog portion 3502 of the LMU device or channel can be connected via a cable to an uplink-positioning processor circuit (e.g., uplink-positioning processor 2810), which can include baseband signal processing. As another example, such embodiments facilitate an improved signal-to-noise ratio (SNR) because signal amplification can occur between the antenna and the connecting cable between the RF analog portion 3502 and the baseband circuitry. Furthermore, the RF analog portion 3502 can downconvert the received signal (e.g., to baseband), and because the baseband signal frequency is several orders of magnitude lower than the received signal at the antenna, cabling requirements can be relaxed. This reduced cabling requirement can translate into reduced connectivity costs and significantly increased transmission range.

Establishing a database of precise antenna locations for multiple geographically distributed base stations and/or small cells and/or LMUs in indoor and other GPS/GNSS-denied environments is more complex than determining antenna locations for macrocell and/or LMU devices used outdoors (e.g., in GPS/GNSS-friendly environments). This is because macrocells and LMUs in outdoor environments have antennas that are elevated and exposed. Consequently, GPS/GNSS signal quality is excellent, and macrocell and/or LMU antenna locations can be determined using GPS/GNSS with a very high level of accuracy (e.g., with a standard deviation of 10 ns or better).

In one aspect, for indoor and other GPS/GNSS-denied environments, a database of antenna locations among multiple distributed base stations and/or small cells/LMUs can be derived from available building/architecture diagrams that include distributed base station and/or small cell/LMU device and antenna placement. However, this approach can be error-prone and lead to inaccuracies in the cellular/wireless antenna location database.

However, when the LMU antenna position can be accurately correlated with at least one of the distributed base station and/or small cell antennas, such as when the antennas are shared or located in close proximity, the aforementioned limitations can be overcome by determining the precise antenna positions of the distributed base stations and/or small cells.

In this case, the precise antenna positions of the distributed base stations and/or small cells can be determined by deploying three or more antenna position calibration units (DL-APCUs) that can be time-synchronized. In an embodiment, the DL-APCU is basically a receiver with computing capabilities. In another embodiment, the DL-APCU receiver may adopt SDR (software defined radio) technology. In another embodiment, the DL-APCU may be a small cell, a macro cell, or a dedicated small cell type device that only performs reception. The DL-APCU may use a downlink reference signal transmitted by the distributed base station and/or small cell as a ranging signal, such as a CRS (cell-specific reference signal). The data collected by the DL-APCU may then be processed by a multipath mitigation processor and a positioning processor to determine the precise antenna position of the distributed base station and/or small cell relative to the antenna position of the DL-APCU. This determination may be performed in real time or offline using a separate server.

This procedure may require precise knowledge of the DL-APCU antenna coordinates to ensure that the antenna positions of the distributed base stations and/or small cells are accurately established. This can be achieved by making the DL-APCU portable so that it can be placed in a GPS/GNSS friendly area next to a window, just outside a building, etc. Also, because the downlink reference signals are known a priori, it is possible to achieve much higher signal processing gain for these signals than for the data payload signals. Therefore, the reception range of the reference signals can be several times larger than the communication range specified for the distributed base stations and/or small cells. This can make it possible to place the DL-APCU in a GPS/GNSS friendly area. Also, the DL-APCU can use GPS/GNSS for time synchronization.

When the LMU antenna position cannot be precisely correlated with at least one of the distributed base station and/or small cell antennas, such as in the case of UL-TDOA, which may include one or more cell towers and multiple LMUs, as illustrated in FIG19 , a single LMU Antenna Position Calibration Unit (UL-APCU) can be used to determine the precise LMU antenna position. As one example, the UL-APCU is essentially a UE (User Equipment) device with or without computing capabilities, similar to a handset. As another example, the UL-APCU may employ SDR technology. In another example, the UL-APCU may be a handset or a dedicated UE device.

In one embodiment, the UL-APCU may be used with an eNB simulator and/or handset test equipment. This simulator/test equipment may generate all signals that enable the UL-APCU to lock onto this simulator/test equipment. In addition, this simulator/test equipment may configure the UL-APCU to generate predetermined uplink reference signals, such as SRS (sounding reference signals) that may be used by the LMU as ranging signals. The data collected by the LMU may then be processed by a multipath mitigation processor as well as a positioning processor to determine the antenna position of the UL-APCU relative to the LMU antenna position. If the UL-APCU antenna position is known, it is possible to determine the precise antenna position of the LMU by performing multiple position measurements of the UL-APCU when the UL-APCU is positioned at several different positions. This determination may be performed in real time or offline using a separate server.

This procedure may require precise knowledge of the UL-APCU antenna coordinates to ensure that the LMU's antenna position is accurately established. This can be achieved by making the UL-APCU portable and also placing the eNB emulator and/or handset test equipment so that the UL-APCU can be positioned in a GPS/GNSS-friendly area, next to a window, just outside a building, etc. Also, because the uplink reference signals are known a priori, the LMU achieves much higher signal processing gain for these signals than for data payload signals. Therefore, the reception range of the reference signals is several times greater than the communication range. This allows positioning of the UL-APCU in GPS/GNSS-friendly areas.

One possible drawback of the described LMU antenna coordinate determination is the possibility of interference from signals generated by the eNB emulator and/or handset test equipment and the UL-APCU. This phenomenon can be mitigated by 1) connecting the UL-APCU receiver to the eNB emulator and/or handset test equipment via a cable, and 2) having the UL-APCU transmitter/antenna generate the uplink reference signal out-of-band, i.e. in an unused LTE band and/or ISM band, such as the 915 MHz ISM band. This latter solution is possible because the LMU is an SDR-based receiver that can be easily tuned to multiple frequency bands. Furthermore, connecting the UL-APCU to the eNB emulator and/or handset test equipment via a cable can further facilitate UL-APCU positioning in GPS/GNSS friendly areas.

In yet another embodiment, the UL-APCU may lock onto one of the operating cells, which may become the serving cell for the UL-APCU; and this cell may be programmed to configure the UL-APCU to generate one or more predetermined uplink reference signals, which may then be used by the LMU as ranging signals.

Similar to distributed base stations and/or small cells and/or other LTE devices, the DL-ACPU and UL-ACPU clocks may be derived from GPS/GNSS for stability.

DL-ACPU is optional because the UL-ACPU system and method can cover two cases: when the LMU antenna position can be accurately correlated with at least one of the distributed base station and/or small cell antennas, and when the LMU's antenna position cannot be accurately correlated with at least one of the distributed base station and/or small cell antennas.

It should be understood that the ranging signal is not limited to SRS and other reference signals including MIMO, CRS (cell-specific reference signal), etc. may be utilized.

Having thus described various embodiments of a system and method, it should be apparent to those skilled in the art that certain advantages of the described method and apparatus have been achieved. Specifically, those skilled in the art will appreciate that a system for tracking and locating an object can be assembled using an FPGA or ASIC and standard signal processing software/hardware combinations with minimal incremental cost. Such a system is suitable for a variety of applications, such as locating a person in indoor or outdoor environments, in harsh and hostile environments, and the like.

It should also be understood that various modifications, adaptations, and alternative embodiments thereof can be made within the scope and spirit of the invention.

Claims (16)

1. A method for determining the location of an antenna in a wireless system, the method comprising: Three or more antenna position calibration units are deployed within the communication range of the antenna, each of the three or more antenna position calibration units having a known location, and at least one of the three or more antenna position calibration units includes a multipath suppression processor; A reference signal is received via the antenna at the three or more antenna position calibration units; Data indicating the position of the antenna is collected from the reference signal; and The collected data is analyzed to determine the position of the antenna relative to the known position of each of the three or more antenna position calibration units. 2. The method according to claim 1, wherein the three or more antenna position calibration units are one or more of a small cell device, a macro cell device, or a small cell receiving device only. 3. The method of claim 1, wherein the three or more antenna position calibration units employ software-defined radio technology. 4. The method according to claim 1, wherein the reference signal is a ranging signal. 5. The method of claim 1, wherein the three or more antenna position calibration units are synchronized in time based on the distance positioning accuracy of the antenna within a standard deviation less than or equal to a predetermined time. 6. The method according to claim 1, wherein the received reference signal is a ranging signal. 7. The method of claim 1, wherein the multipath suppression processor is configured to use coherent summation, incoherent summation, matched filtering, time diversity, or a combination thereof to improve the signal-to-noise ratio of the received reference signal. 8. The method of claim 1, wherein the multipath suppression processor models each of the received reference signals as a linear combination of a complex exponent of frequency and a complex amplitude. 9. A method for determining the location of an antenna in a wireless system, the method comprising: Three or more antenna position calibration units are deployed within the communication range of the antenna, each of the three or more antenna position calibration units having a known position; The reference signal is transmitted to the antenna via the three or more antenna position calibration units; Data indicating the position of the antenna is collected from the reference signal; and The collected data is analyzed by a multipath suppression processor to determine the position of the antenna relative to the known position of each of the three or more antenna position calibration units. 10. The method of claim 9, wherein the three or more antenna position calibration units comprise an emulator or handheld device configured to generate a signal that enables the three or more antenna position calibration units to lock onto the emulator or handheld device. 11. The method of claim 9, wherein the reference signal is a detection reference signal. 12. The method of claim 9, wherein the reference signal is a ranging signal. 13. The method of claim 9, wherein the three or more antenna position calibration units are synchronized in time based on the distance positioning accuracy of the antenna within a standard deviation less than or equal to a predetermined time. 14. The method of claim 9, wherein the collected data includes measured distances between the antenna and each of the three or more antenna position calibration units. 15. The method of claim 14, wherein the measured distance is based on the reference signal. 16. The method of claim 9, wherein the transmitted reference signal is a wide-bandwidth signal comprising different narrow-bandwidth frequency components.