AU2014203462A1 - Estimating Wheel Speed for In-wheel Electric Vehicles - Google Patents

Abstract

Abstract A method of estimating the wheel speed of a wheel of an in-wheel electric vehicle, the vehicle having an electric 5 motor which produces a back EMF when driving that wheel, the method comprising: obtaining two or more signals which are related to the wheel speed, at least one signal being obtained from the back EMF of the electric motor that drives the wheel; 10 processing the two or more signals to derive an estimate of the wheel speed; and producing an output indicative of the wheel speed from the wheel speed estimate. 5520179_1 (GHMatters) P97310.AU 25/06/14 Left Right Front Front Signal Sigal ABS Transmission Transmission ABS Sensor Motor Driver: I Braking I 15 Command Local Local Controller Controller 12 Communication Network I Local o Controller t > Controller 13 Central Brake CoContrllere1 Left embedded in Rigt 10 Re\ ECU Ra ABS SgaSi1ABS Sensor in-Wheel Transmission Transmi ion Sor Motor Motor 14 FigI damper brake motor car wheel wheel simulating relative road motion driving motor . encoder brake Fig. 2 5487714_1 (GHMatters) P97310.AU

Description

- 1 Estimating Wheel Speed for In-wheel Electric Vehicles Technical Field 5 The present disclosure relates to methods and apparatuses for estimating the wheel speed of a wheel in an in-wheel electric vehicle. Some methods and apparatuses disclosed herein have particular application where the electric motor used to drive a given wheel in an in-wheel electric 10 vehicle is a multiple phase motor such as a permanent magnet brushless (PMBL) motor. The present disclosure also relates to an antilock braking system for an in-wheel electric vehicle. The vehicle may be a car or a motorbike for example. 15 Background The measured wheel speed is an input to a number of systems in modern vehicles. One such system is the 20 antilock braking system (ABS), which is a crucial safety measure. The role of the ABS is to maintain the steerability of the vehicle whilst trying to minimise the vehicle stopping 25 distance in panic braking or challenging braking scenarios. In conventional vehicles (i.e. petrol or diesel engine driven vehicles), the ABS includes an ABS sensor on at least one of the wheels of the vehicle to measure the rotational speed of that wheel. A typical 30 commercially available ABS sensor comprises a toothed ring, a permanent magnet and a winding (pick up). The toothed ring is configured to rotate with the vehicle wheel at the same rotational speed. The rotational motion of the toothed ring in the magnetic field of the permanent 35 magnet induces a voltage in the winding. This voltage contains information about the wheel speed and is used in conventional antilock braking systems to estimate the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 2 wheel rotational speed. The wheel speed estimated from the sensor is then used by the ABS to control the braking force supplied to the wheel or wheels of the vehicle. 5 Unlike conventional vehicles, in which all the wheels are driven from a central powertrain, in-wheel electric vehicles are powered by independent electric motors at each wheel. The phrase "in-wheel" refers to the motors of such vehicles being mounted inside the wheel hubs of each 10 wheel. Different types of electric motors have been incorporated into in-wheel electric vehicles including brushed and brushless motors. Brushless motors have advantages over brushed motors in their high efficiency and low maintenance, although their dynamics are more 15 complicated. Conventional antilock braking systems can be incorporated into in-wheel electric vehicles. However, because the ABS sensors also have to be fitted into the wheel hub, along 20 with the electric motor, suspension and braking systems, design difficulties arise with the limited space. Furthermore, the electric motor is a significant source of disturbance for the output signal from the ABS sensor to the ABS controller. This can distort the ABS sensor's 25 signal and cause inaccuracies in the performance of the antilock braking system. Accordingly, it is desirable to develop an alternative antilock braking system for in-wheel electric vehicles. 30 Summary of the Disclosure The present disclosure relates to improvements in antilock braking systems for in-wheel electric vehicles and in 35 particular to methods for estimating the wheel speed of a wheel of an in-wheel electric vehicle. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 3 In one embodiment the present disclosure provides a method of estimating the wheel speed of a wheel of an in-wheel electric vehicle, the vehicle having an electric motor which produces a back EMF when driving that wheel, the 5 method comprising: obtaining two or more signals which are related to the wheel speed, at least one signal being obtained from the back EMF of the electric motor that drives the wheel; processing the two or more signals to derive an 10 estimate of the wheel speed; and producing an output indicative of the wheel speed from the wheel speed estimate. In another embodiment, the present disclosure provides a 15 method of estimating the wheel speed of a wheel of an in wheel electric vehicle, the vehicle having a multiple phase motor which produces a back EMF in each phase when driving said wheel, the method comprising: obtaining a signal from the amplitude of the back EMF 20 of one phase of the motor; and processing the back EMF amplitude signal to derive an estimate of the wheel speed from the back EMF amplitude signal; and producing an output indicative of the wheel speed 25 from the wheel speed estimate. In another embodiment, the present disclosure provides an apparatus for estimating the wheel speed of at least one wheel of an in-wheel electric vehicle, the vehicle having 30 an electric motor which produces a back EMF when driving that wheel, the apparatus comprising: two or more sensors each configured to obtain a signal which is related to the wheel speed, wherein at least one sensor is a back EMF sensor that obtains a 35 signal corresponding to the back EMF of the electric motor that drives the wheel; a processor operable to receive and process the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 4 signals from the two or more sensors to derive an estimate of the wheel speed and produce an output indicative of the wheel speed from the wheel speed estimate. 5 In another embodiment, the present disclosure provides an apparatus for estimating the wheel speed of at least one wheel of an in-wheel electric vehicle, the vehicle having a multiple phase motor which produces a back EMF in each phase when driving said wheel, the apparatus comprising: 10 a sensor configured to obtain a signal from the amplitude of the back EMF of one phase of the motor; a processor operable to receive and process the signal from the back EMF amplitude sensor to derive an estimate of the wheel speed and produce an output 15 indicative of the wheel speed from the wheel speed estimate. In another embodiment, the present disclosure provides an antilock braking system for an in-wheel electric vehicle 20 having a plurality of independent electric motors, each motor producing a back EMF when driving a respective wheel, the system comprising: two or more sensors that are each arranged to obtain a signal related to the wheel speed of one of the wheels, 25 at least one of the sensors comprising a back EMF sensor that obtains a signal corresponding to the back EMF of the motor that drives said wheel; and a processor that receives the signals from each sensor, the processor programmed to derive an estimate of 30 the wheel speed and produce an output indicative of the wheel speed from the wheel speed estimate for use in controlling a braking force applied to said wheel. In a further embodiment, the present disclosure provides 35 an antilock braking system for an in-wheel electric vehicle having a plurality of independent multiple phase motors, 5520179_1 (GHMatters) P97310.AU 25/06/14 - 5 each motor producing a back EMF in each phase when driving a respective wheel, the system comprising: a sensor that is arranged to obtain a signal corresponding to the amplitude of the back EMF of one 5 phase of one of the motors; and a processor arranged to receive the amplitude signal from the sensor and programmed to derive an estimate of the wheel speed of the wheel driven by said motor from the amplitude signal and produce an output indicative of the 10 wheel speed for use in controlling a braking force applied to said wheel. Brief Description of the Drawings 15 Embodiments of the present disclosure will now be described, by way of reference only, in which: Fig. 1 is a schematic view of an architecture for an antilock braking system of an electric vehicle with four 20 in-wheel motors; Fig. 2 is an image of an antilock braking system test rig used in Example 1; 25 Fig. 3 is graphs of the reference measurements for the wheel speed from the optical encoder in Example 1 (top) and the back EMF of the brushed DC motor corresponding to the top graph (bottom); 30 Fig. 4 is a graph of the output of the ABS sensor in Example 1 corresponding to Fig. 3; Fig. 5 is graphs of the reference measurements for the wheel speed from the optical encoder in Example 1 (top) 35 and the absolute values of the ABS sensor output shown in Fig. 4; 5520179_1 (GHMatters) P97310.AU 25/06/14 - 6 Fig. 6 is the wheel speed estimated from the frequency of the ABS sensor output in Example 1; Fig. 7 is the wheel speed estimated from the amplitude of 5 the ABS sensor output in Example 1; Fig. 8 is the wheel speed after the antilock braking system is activated fused from the frequency and amplitude of the ABS sensor output and the back EMF of the brushed DC motor 10 in Example 1 in which the fused wheel speed was calculated by the OWA (dispersion) method; Fig. 9 is the wheel speed after the antilock braking system is activated fused from the frequency and amplitude of the 15 ABS sensor output and the back EMF of the brushed DC motor in Example 1 in which the fused wheel speed was calculated by the median method; Fig. 10 is a circuit diagram of an electronically 20 commutated BLDC motor; Fig. 11 is a circuit diagram of an equivalent circuit of phase A of the BLDC motor of Fig. 10; 25 Fig. 12 is a schematic diagram of a wheel speed estimation method for a BLDC motor driven wheel in an in-wheel electric vehicle; Fig. 13 is an image of an antilock braking system test rig 30 used in Example 2; Fig. 14 is a graph of the reference (optical encoder) measurements for wheel speed and the wheel speed estimated from the frequency and the amplitude of the ABS sensor 35 output in Example 2; Fig. 15 is the wheel speed from the amplitude of the back 5520179_1 (GHMatters) P97310.AU 25/06/14 - 7 EMF of each of the three phases of the BLDC motor in Example 2; Figs. 16 - 18 are the wheel speeds from a fusion of the 5 amplitude of the ABS sensor output and the amplitude of the back EMF of the three phases of the BLDC motor in Example 2, in which the fused wheel speeds have been produced by the OOWA method (Fig. 16), the POWA method (Fig. 17) and the median method (Fig. 18); 10 Fig. 19 is a comparison of the error in different averaging methods, when the amplitude of the ABS sensor output and the amplitude of the back EMF the three phases of the BLDC motor in Example 2 are fused; 15 Fig. 20 is the wheel speeds from a fusion of the amplitude and frequency of the ABS sensor output and the amplitude and frequency of the back EMF of the three phases of the BLDC motor in Example 2, in which the fused wheel speeds 20 have been produced by the median method; Fig. 21 is the implementation structure of a Discrete Wavelet Transform (DWT) in Example 3; Fig. 22 is an image of an antilock braking system test rig 25 used in Example 3; Fig. 23 is the reference wheel speed measurements by the optical encoder in Example 3; 30 Fig. 24 is the unprocessed output of the ABS sensor in Example 3; Fig. 25 is a comparison of the wheel speed estimations from the frequency of the ABS sensor output and the wheel 35 speed estimations from the reference (optical encoder) measurements during antilock braking system activation in Example 3; 5520179_1 (GHMatters) P97310.AU 25/06/14 - 8 Fig. 26 is a graph of the output of the back EMF of one phase of the BLDC motor with 8 pole pairs in Example 3; 5 Fig. 27 is graphs of the reference measurements for the wheel speed from the optical encoder in Example 3 (top) and the back EMF of one phase of the BLDC motor with 8 pole pairs; 10 Fig. 28 is the wheel speed estimations using the application of the peak detection method on the amplitude of the back EMF shown in Fig. 27; Fig. 29 is the wheel speed estimation from the back EMF of 15 a single phase of the BLDC motor with only two pole pairs using the DWT method in Example 3; Fig. 30 is the wheel speed from the fusion by a median operator of the wheel speed estimates derived from the 20 amplitude from each of two phases of the BLDC motor in Example 3; Fig. 31 is the wheel speed estimation from the back EMF a brushed DC motor in Example 3; 25 Fig. 32 is the energy index for different road conditions and initial speeds in Example 3; Fig. 33 is a schematic flow chart of a method of 30 estimating the wheel speed of a wheel of an in-wheel electric vehicle according to an embodiment of the present disclosure; Fig. 34 is a schematic flow chart of a method of 35 estimating the wheel speed of a wheel of an in-wheel electric vehicle according to another embodiment of the present disclosure; 5520179_1 (GHMatters) P97310.AU 25/06/14 - 9 Fig. 35 is a schematic diagram of an antilock braking system for an in-wheel electric vehicle according to an embodiment of the present disclosure; and 5 Fig. 36 is a schematic diagram of an antilock braking system for an in-wheel electric vehicle according to another embodiment of the present disclosure. 10 Detailed Description of Embodiments In the following detailed description, reference is made to the accompanying drawings, which form a part thereof. In the drawings, similar symbols typically identify 15 similar components, unless context dictates otherwise. The illustrative embodiments described in the detailed description, drawings, and claims are not meant to be limiting. Other embodiments may be utilised, and other changes may be made, without departing from the spirit or 20 scope of the subject matter presented herein. It will be readily understood that the aspects of the present disclosure, as generally described herein, and illustrated in the Figures, can be arranged, substituted, combined, separated and designed in a wide variety of different 25 configurations, all of which are explicitly contemplated herein. Disclosed in some embodiments is a method of estimating the wheel speed of a wheel of an in-wheel electric 30 vehicle, the vehicle having an electric motor which produces a back EMF when driving that wheel, the method comprising: obtaining two or more signals which are related to the wheel speed, at least one signal being obtained from 35 the back EMF of the electric motor that drives the wheel; processing the two or more signals to derive an estimate of the wheel speed; and 5520179_1 (GHMatters) P97310.AU 25/06/14 - 10 producing an output indicative of the wheel speed from the wheel speed estimate. It is to be understood that the "back EMF" refers to the 5 "back electromagnetic force" produced by the electric motor. The back EMF is the voltage generated by the rotation of the motor's armature or rotor that opposes the applied voltage. 10 The motor may be a brushed DC motor or may be a multiple phase motor including a permanent magnet brushless (PMBL) motor such as a brushless DC (BLDC) motor. In an embodiment, the method comprises deriving an 15 estimate of the wheel speed from each obtained signal and producing the wheel speed output by averaging the estimated wheel speeds. In an embodiment, the method comprises fusing each of the 20 obtained signals into a fused signal and deriving the estimate of the wheel speed from the fused signal. In an embodiment, the electric motor is a three phase motor and all of the two or more signals are obtained from 25 the back EMF of the electric motor that drives the wheel. In an embodiment, the electric motor is a three phase motor and each signal obtained from the back EMF is obtained from the back EMF of a single phase of the motor. 30 In an embodiment, the electric motor is a three phase motor and at least two of the two or more signals are obtained from the back EMF, wherein each of said back EMF signals are obtained from the back EMF of different phases 35 of the motor. In an embodiment, each signal obtained from the back EMF 5520179_1 (GHMatters) P97310.AU 25/06/14 - 11 is a frequency or an amplitude of the back EMF. In an embodiment, the electric motor is a three phase motor and the obtained signals include a back EMF 5 amplitude and a back EMF frequency from the back EMF of a single phase of the motor. In an embodiment, the electric motor is a three phase motor and the obtained signals include a back EMF 10 amplitude and a back EMF frequency from each of the back EMF of two or more phases of the motor. In an embodiment, at least one of the signals is obtained from an output of an antilock braking system (ABS) sensor. 15 It is to be understood that references herein to an "antilock braking system sensor" or "ABS sensor" are references to a conventional sensor employed in the antilock braking systems of conventional vehicles (ie. 20 vehicles having diesel or petrol engines) to provide information about the wheel speed for the antilock braking system to control the braking of the wheels of those vehicles. 25 In accordance with the above understanding, in an embodiment the ABS sensor comprises a member attached to the wheel so as to rotate with the wheel, a magnetic field generator which generates a magnetic field in which the member rotates and a wire coil in which a current output 30 is generated by the rotation of the member in the magnetic field, wherein each ABS sensor signal is obtained from the current output. In an embodiment, the ABS sensor signal is the frequency 35 or the amplitude of the ABS sensor output. In an embodiment, the obtained signals comprise the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 12 frequency and the amplitude of the ABS sensor output. In an embodiment, averaging the wheel speed estimates comprises applying an averaging algorithm to the derived 5 wheel speed estimates. In an embodiment, fusing each of the obtained signals into a fused signal comprises applying an averaging algorithm to the obtained signals. 10 In an embodiment, the averaging algorithm is selected from the group consisting of ordered weighted averaging (OWA), median averaging and mean averaging. 15 In an embodiment, deriving the wheel speed estimate comprises pre-processing at least one of the two or more signals before processing the two or more signals. In an embodiment, the pre-processing comprises smoothing 20 the signal using local maxima. In an embodiment, the pre-processing comprises applying a discrete wavelet transform to the signal. 25 Disclosed in some embodiments is an apparatus for estimating the wheel speed of at least one wheel of an in wheel electric vehicle, the vehicle having an electric motor which produces a back EMF when driving that wheel, the apparatus comprising: 30 two or more sensors each configured to obtain a signal which is related to the wheel speed, wherein at least one sensor is a back EMF sensor that obtains a signal corresponding to the back EMF of the electric motor that drives the wheel; 35 a processor operable to receive and process the signals from the two or more sensors to derive an estimate of the wheel speed and produce an output indicative of the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 13 wheel speed from the wheel speed estimate. In an embodiment, the electric motor is a three phase motor and the apparatus comprises two or more back EMF 5 sensors each configured to obtain a signal corresponding to the back EMF of the electric motor that drives the wheel. In an embodiment, the electric motor is a three phase 10 motor and each sensor is a back EMF sensor that is configured to obtain a signal corresponding to the back EMF of a single phase of the motor. In an embodiment, the electric motor is a three phase 15 motor and at least two of the two or more sensors are back EMF sensors, each back EMF sensor configured to obtain a signal corresponding to the back EMF of different phases of the motor. 20 In an embodiment, at least one of the sensors is an antilock braking system (ABS) sensor comprising a member attached to the wheel so as to rotate with the wheel, a magnetic field generator which generates a magnetic field in which the member rotates and a wire coil in which a 25 current output is generated by the rotation of the member in the magnetic field, wherein the ABS sensor is configured to obtain a signal corresponding to the current output. 30 In an embodiment, the apparatus comprises a pre-processor operable to receive and pre-process the signal from at least one of the two or more sensors and sending each pre processed signal to the processor. 35 Any one or more of the embodiments of the above method or the above apparatus can be employed in an antilock braking system for an in-wheel electric vehicle. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 14 Disclosed in some embodiments is an antilock braking system for an in-wheel electric vehicle having a plurality of independent electric motors, each motor producing a 5 back EMF when driving a respective wheel, the system comprising: two or more sensors that are each arranged to obtain a signal related to the wheel speed of one of the wheels, at least one of the sensors comprising a back EMF sensor 10 that obtains a signal corresponding to the back EMF of the motor that drives said wheel; and a processor that receives the signals from each sensor, the processor programmed to derive an estimate of the wheel speed and produce an output indicative of the 15 wheel speed from the wheel speed estimate for use in controlling a braking force applied to said wheel. In an embodiment, the processor may receive signals from two or more sensors provided in respective of each wheel 20 of the vehicle and is programmed to derive an estimate of the wheel speed for each wheel. The processor is further programmed to produce an output indicative of the wheel speed of each respective wheel from the estimated wheel speed for the respective wheel for use in controlling 25 independent braking forces applied to respective wheels. In an embodiment, the processor derives an estimate of the wheel speed from each obtained signal and produces the wheel speed output by averaging the estimate wheel speeds. 30 In an embodiment, the processor is programmed to fuse each of the obtained signals into a fused signal and derive the estimate of the wheel speed from the fused signal. 35 In an embodiment, the electric motor is a three phase motor and each of the sensors is a back EMF sensor that obtains a signal corresponding to the back EMF of the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 15 motor that drives the wheel. In an embodiment, the electric motor is a three phase motor and at least two of the sensors are back EMF sensors 5 that obtain a signal corresponding to the back EMF of a single phase of the motor. In an embodiment, the electric motor is a three phase motor and at least two of the two or more sensors are 10 back EMF sensors, wherein each of said back EMF sensors are configured to obtain a signal corresponding to the back EMF of different phases of the motor. In an embodiment, each back EMF sensor obtains a signal 15 derived from a frequency or an amplitude of the back EMF. In an embodiment, the electric motor is a three phase motor and at least two of the sensors are back EMF sensors, one back EMF sensor obtaining a signal derived 20 from a back EMF amplitude and another back EMF sensor obtaining a signal derived from a back EMF frequency of a single phase of the motor. In an embodiment, the electric motor is a three phase 25 motor and at least two of the sensors are back EMF sensors, two back EMF sensors obtaining signals derived from a back EMF amplitude of two phases of the motor respectively and another two back EMF sensor obtaining a signals derived from a back EMF frequency of two or more 30 phases of the motor respectively. In an embodiment, at least one of the two or more sensors comprises an antilock braking system (ABS) sensor comprising a member attached to the wheel so as to rotate 35 with the wheel, a magnetic field generator which generates a magnetic field in which the member rotates and a wire coil in which a current output is generated by the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 16 rotation of the member in the magnetic field, wherein at least one ABS sensor signal is obtained corresponding to the current output. 5 In an embodiment, the at least one ABS sensor signal is the frequency or the amplitude of the ABS sensor output. In an embodiment, the signals obtained from the ABS sensor include the frequency and the amplitude of the ABS sensor 10 output. In an embodiment, the processor is programmed to average the wheel speed estimates by applying an averaging algorithm to the wheel speed estimates derived from each 15 of the signals received by the processor from the transmitter. In an embodiment, the processor is programmed to fuse each of the obtained signals into the fused signal by applying 20 an averaging algorithm to the obtained signals. In an embodiment, the averaging algorithm is selected from the group consisting of ordered weighted averaging (OWA), median averaging and mean averaging. 25 In an embodiment, the system comprises a pre-processor that receives the signal from at least one of the two or more sensors and is programmed to pre-process said received signals and send each pre-processed signal to the 30 processor. In an embodiment, the pre-processor is programmed to pre process the signal received from at least one of the two or more sensors by smoothing those signal(s) using local 35 maxima. In an embodiment, the pre-processor is programmed to pre 5520179_1 (GHMatters) P97310.AU 25/06/14 - 17 process the signal received from at least one of the two or more sensors by applying a discrete wavelet transform to those signal(s). 5 Disclosed in some embodiments is a method of estimating the wheel speed of a wheel of an in-wheel electric vehicle, the vehicle having a multiple phase motor which produces a back EMF in each phase when driving said wheel, the method comprising: 10 obtaining a signal from the amplitude of the back EMF of one phase of the motor; and processing the back EMF amplitude signal to derive an estimate of the wheel speed from the back EMF amplitude signal; and 15 producing an output indicative of the wheel speed from the wheel speed estimate. In an embodiment, the multiple phase motor is a permanent magnet brushless (PMBL) motor having only two pole pairs. 20 In other embodiments, the PMBL motor has more than two pole pairs. The PMBL motor may be a brushless DC (BLDC) motor. 25 In an embodiment, the method comprises obtaining two or more signals, each signal being the amplitude of the back EMF of a respective phase of the motor, processing the two or more signals to derive an estimate of the wheel speed and producing the output indicative of the wheel speed 30 from the wheel speed estimate. In an embodiment, the method comprises obtaining one or more additional signals, each additional signal comprising the frequency of the back EMF of one phase of the motor, 35 processing the obtained signals to derive an estimate of the wheel speed and producing the output indicative of the wheel speed from the wheel speed estimate. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 18 In an embodiment, the method comprising obtaining at least one signal from an output of an antilock braking system (ABS) sensor, processing the obtained signals to derive an 5 estimate of the wheel speed and producing the output indicative of the wheel speed from the wheel speed estimate. In an embodiment, the ABS sensor comprises a member 10 attached to the wheel so as to rotate with the wheel, a magnetic field generator which generates a magnetic field in which the member rotates and a wire coil in which a current output is generated by the rotation of the member in the magnetic field, wherein each ABS sensor signal is 15 obtained from the current output. In an embodiment, the ABS sensor signal is the frequency or the amplitude of the ABS sensor output. 20 In an embodiment, the obtained ABS sensor signals comprise the frequency and the amplitude of the ABS sensor output. In an embodiment, the method comprises deriving an estimate of the wheel speed from each obtained signal and 25 producing the wheel speed output by averaging the estimated wheel speeds. In an embodiment, the method comprises fusing each of the obtained signals into a fused signal and deriving the 30 estimate of the wheel speed from the fused signal. In an embodiment, averaging the wheel estimates comprises applying an averaging algorithm to the wheel speed estimates derived from each signal. 35 In an embodiment, fusing each of the obtained signals into a fused signal comprises applying an averaging algorithm 5520179_1 (GHMatters) P97310.AU 25/06/14 - 19 to the obtained signals. In an embodiment, the averaging algorithm is selected from the group consisting of ordered weighted averaging (OWA), 5 median averaging and mean averaging. In an embodiment, deriving the wheel speed estimate comprises pre-processing before processing the back EMF amplitude signal. 10 In an embodiment, pre-processing comprises smoothing the back EMF amplitude signal using local maxima. In an embodiment, pre-processing comprises applying a 15 discrete wavelet transform to the back EMF amplitude signal. Disclosed in some embodiments is an apparatus for estimating the wheel speed of at least one wheel of an in 20 wheel electric vehicle, the vehicle having a multiple phase motor which produces a back EMF in each phase when driving said wheel, the apparatus comprising: a sensor configured to obtain a signal from the amplitude of the back EMF of one phase of the motor; 25 a processor operable to receive and process the signal from the back EMF amplitude sensor to derive an estimate of the wheel speed and produce an output indicative of the wheel speed from the wheel speed estimate. 30 In an embodiment, the apparatus comprises two or more sensors, each sensor configured to obtain a signal corresponding to the amplitude of the back EMF of a respective phase of the motor, and wherein the processor 35 is operable to process the signals from the two or more sensors to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 20 wheel speed estimate. In an embodiment, the apparatus comprises one or more additional sensors, each additional sensor configured to 5 obtain a signal corresponding to the frequency of the back EMF of one phase of the motor, wherein the processor is operable to process the obtained signals to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the wheel speed 10 estimate. In an embodiment, the apparatus comprises an antilock braking system (ABS) sensor, the ABS sensor comprises a member attached to the wheel so as to rotate with the 15 wheel, a magnetic field generator which generates a magnetic field in which the member rotates and a wire coil in which a current output is generated by the rotation of the member in the magnetic field, wherein the ABS sensor is configured to obtain a signal corresponding to the 20 current output and wherein the processor is operable to process the obtained signals to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the wheel speed estimate. 25 Any one or more of the embodiments of the above method or the above apparatus can be employed in an antilock braking system for an in-wheel electric vehicle. Disclosed in some embodiments is an antilock braking 30 system for an in-wheel electric vehicle having a plurality of independent multiple phase motors, each motor producing a back EMF in each phase when driving a respective wheel, the system comprising: a sensor that is arranged to obtain a signal 35 corresponding to the amplitude of the back EMF of one phase of one of the motors; and a processor arranged to receive the amplitude signal 5520179_1 (GHMatters) P97310.AU 25/06/14 - 21 from the sensor and programmed to derive an estimate of the wheel speed of the wheel driven by said motor from the amplitude signal and produce an output indicative of the wheel speed for use in controlling a braking force applied 5 to said wheel. In an embodiment, the system comprises a sensor in respect of each wheel of the vehicle that is arranged to obtain a signal from the amplitude of the back EMF of one phase of 10 the respective motor that drives the respective wheel. In an embodiment, the system comprises two or more sensors, each sensor arranged to obtain a signal corresponding to the amplitude of the back EMF of a 15 respective phase of said motor and the processor is arranged to receive each signal and is programmed to process the signals from the two or more sensors to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the wheel speed 20 estimate. In an embodiment, the system comprises one or more additional sensors, each additional sensor configured to obtain a signal corresponding to the frequency of the back 25 EMF of one phase of the motor, wherein the processor is operable to process the obtained signals to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the wheel speed estimate. 30 In an embodiment, the system comprises an antilock braking system (ABS) sensor, the ABS sensor comprising a member attached to the wheel so as to rotate with the wheel, a magnetic field generator which generates a magnetic field 35 in which the member rotates and a wire coil in which a current output is generated by the rotation of the member in the magnetic field, wherein the ABS sensor is 5520179_1 (GHMatters) P97310.AU 25/06/14 - 22 configured to obtain a signal corresponding to the current output and wherein the processor is operable to process the obtained signals to derive an estimate of the wheel speed and produce the output indicative of the wheel speed 5 from the wheel speed estimate. In an embodiment, the at least one ABS sensor signal comprises the frequency or the amplitude of the ABS sensor output. 10 In an embodiment, the at least one ABS sensor signal comprises the frequency and the amplitude of the ABS sensor output. 15 In an embodiment, the processor derives an estimate of the wheel speed from each obtained signal and produces the wheel speed output by averaging the estimated wheel speeds. 20 In an embodiment, the processor is programmed to fuse each of the obtained signals into a fused signal and derive the estimate of the wheel speed from the fused signal. In an embodiment, the processor is programmed to average 25 the wheel speed estimates by applying an averaging algorithm to the wheel speed estimates derived from each of the signals received by the processor. In an embodiment, the processor is programmed to fuse each 30 of the obtained signals into the fused signal and derive the estimate of the wheel speed from the fused signal. In an embodiment, the averaging algorithm is selected from the group consisting of ordered weighted averaging (OWA), 35 median averaging and mean averaging. In an embodiment, the system comprises a pre-processor 5520179_1 (GHMatters) P97310.AU 25/06/14 - 23 that receives the amplitude signal received from the sensor and is programmed to pre-process said received signal and send the pre-processed signal to the processor. 5 In an embodiment, the processor is programmed to pre process the amplitude signal received from the sensor by smoothing the signal using local maxima. In an embodiment, the processor is programmed to pre 10 process the amplitude signal received from the sensor by applying a discrete wavelet transform to the signal. Referring to Fig. 33, a flow chart of a method 100 of estimating the wheel speed of a wheel of an in-wheel 15 electric vehicle according to an embodiment of the present disclosure is shown. The method 100 comprises obtaining a back EMF signal 101 from the back EMF of the electric motor that drives the 20 wheel. In an embodiment where the motor is a multiple phase motor such as a three phase brushless DC motor, two or more back EMF signals may be obtained from different phases of the motor. The method also comprises obtaining a signal from an ABS sensor 102. The back EMF signal and 25 the ABS sensor signal may each be subjected to preprocessing 103, 104. The preprocessing is carried out to enable the signal to be more readily processed in subsequent stages of the steps of the method. Techniques such as smoothing the signal using peak detection 30 algorithms and/or applying transforms to the signal, as well as other de-noising techniques may be applied to the signals during the preprocessing steps. The preprocessing may also involve determining the slip conditions of the surface over which the wheel is travelling. 35 Each of the signals (after any optional preprocessing) are then processed to derive an estimate for the wheel speed 5520179_1 (GHMatters) P97310.AU 25/06/14 - 24 105, 106. The estimates for the wheel speed are subjected to an averaging step 107 in which an averaging algorithm such as ordered weighted averaging, median averaging or mean averaging is used to combine the wheel speed 5 estimates from each of the signals and derive a single wheel speed estimate. From this averaging step the method 100 produces an output 108 which is indicative of the wheel speed. This output may be a value for the wheel speed or in other embodiments may be a command to a 10 control system or any other type of output required for a system within the vehicle that requires the wheel speed as an input. Referring now to Fig. 34, a flow chart of a method 200 of 15 estimating the wheel speed of a wheel of an in-wheel electric vehicle according to another embodiment of the present disclosure is shown. In the method 200 of Fig. 34, the wheel is driven by a multiphase motor, such as a three phase brushless DC motor. 20 The method 200 comprises obtaining a back EMF signal 201 from the back EMF of one phase of the multiphase motor and obtaining another back EMF signal 202 from the multiphase motor. The other back EMF signal may be from the same 25 phase of the multiphase motor that the signal obtained in step 201 is from or may be from a different phase of the motor. It is to be appreciated that the method 200 can comprise obtaining further back EMF signals from each of the phases of the multiphase motor, each back EMF signal 30 subjected to the further processes of the method described below. The back EMF signals obtained in steps 201 and 202 are optionally subjected to preprocessing steps 203, 204. 35 The preprocessing is carried out to enable the signal to be more readily processed in subsequent stages of the steps of the method. Techniques such as smoothing the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 25 signal using peak detection algorithms and/or applying transforms to the signal, as well as other de-noising techniques may be applied to the signals during the preprocessing steps. The preprocessing may also involve 5 determining the slip conditions of the surface over which the wheel is travelling. Each of the signals (after any optional preprocessing) are then processed to derive an estimate for the wheel speed 10 205, 206. The estimates for the wheel speed are subjected to an averaging step 207 in which an averaging algorithm such as ordered weighted averaging, median averaging or mean averaging is used to combine the wheel speed estimates from each of the signals and derive a single 15 wheel speed estimate. From this averaging step the method 200 produces an output 208 which is indicative of the wheel speed. This output may be a value for the wheel speed or in other embodiments may be a command to a control system or any other type of output required for a 20 system within the vehicle that requires the wheel speed as an input. It is also noted that in the above method, the wheel speed estimate is obtained only from the back EMF of the phases 25 of the multiphase motor. No ABS sensor or other type of sensor is required. In a variation of the method shown in Fig. 34, only one back EMF signal is obtained and the wheel speed estimated 30 from this one signal is processed to produce an output indicative of the wheel speed. In this embodiment, steps of obtaining another back EMF signal 202, pre-processing that signal 204, estimating the wheel speed from that signal 206 and averaging the wheel speed estimates 207 are 35 not carried out. Referring now to Fig. 35, a schematic diagram of an 5520179_1 (GHMatters) P97310.AU 25/06/14 - 26 antilock braking system 300 for an in-wheel electric vehicle according to an embodiment of the present disclosure is shown. 5 The system 300 operates the method 100 shown and described in respect of Fig. 33. As illustrated in Fig. 35, the system 300 is provided in respect of a wheel 301 in an in wheel electric vehicle which is driven by an electric motor 302 that produces a back EMF. A back EMF sensor 303 10 is provided to detect the back EMF produced by the motor 302 and obtain a signal corresponding to that back EMF. In an embodiment where the electric motor is a multiphase motor such as a three phase brushless motor, the system may comprise multiple back EMF sensors detecting the back 15 EMF of different phases of the motor to obtain multiple back EMF signals corresponding to the back EMF of different phases of the motor. The system also comprises and ABS sensor 304 which detects a current produced by a toothed ring rotating with the wheel in a magnetic field 20 to obtain a signal corresponding to this current. Each of the signals generated by the one or more back EMF sensors 303 and the ABS sensor 304 is optionally received by a pre-processor 305 which subjects each signal received 25 by the pre-processor to one or more pre-processing functions to improve the subsequent processing of the signals to estimate the wheel speed. One process that may be carried out the by the pre-processor 305 is to smooth one or more of the obtained signals 306 using a peak 30 detection module 307 and/or a transform module 308. The peak detection module 307 applies an algorithm in which the local maxima of the amplitude of the signal are identified in order to smooth the signal and reduce the frequency content. The local maximum is defined as a data 35 sample of the amplitude of the signal that is larger than its neighbouring values. The transform module 308 is operable to apply a discrete wavelet transform (DWT) in 5520179_1 (GHMatters) P97310.AU 25/06/14 - 27 which a discrete wavelet transform is used to decompose the wheel speed signal. In some embodiments, the signal is first preprocessed using the peak detection module 307 and subsequently preprocessed using the transform module 5 308. The pre-processor 305 may also process one or more of the obtained signals to determine the surface condition 309 over which the wheel is travelling. The pre-processor 10 does this by using a road condition module 310. The surface condition module 310 is operable to apply a continuous wavelet transform (CWT) to one of the obtained signals and then using an energy index in which higher calculated values on the energy index indicate that the 15 surface is more slippery. The system 300 also comprises a processor 311 which receives the signals after any optional pre-processing by the pre-processor 305. The processer is programmed to 20 derive an estimate of the wheel speed from the signals and produce an output indicative of the estimated wheel speed. The processor does this by processing the signals to derive a wheel speed estimation 312 and combining information about the wheel speed from the signals 313. 25 In one embodiment, this involves firstly carrying out the wheel speed estimation 312 to derive a wheel speed estimate from each signal and then combining wheel speed estimations 313 by averaging the wheel speed estimates. However, in other embodiments (indicated by the dashed 30 arrow in Fig. 35), the processor first combines the signals 313 using data fusion techniques to obtain a single fused signal and then derives a wheel speed estimate 312 from the single fused signal. 35 In both embodiments, a wheel speed estimator 314 is used to apply algorithms to the respective signals to obtain a wheel speed estimate. Also, an averaging module 315 is 5520179_1 (GHMatters) P97310.AU 25/06/14 - 28 used to apply averaging algorithms such as ordered weighted averaging, median averaging or mean averaging to the wheel speed estimates derived from each of the signals or to other components or derivations of the signals. 5 The processor 311 produces an output indicative of the wheel speed, which may be a value for the wheel speed or any other suitable input for a controller 316 in the antilock braking system 300 to enable the controller 316 10 to send appropriate commands to a brake 317 to control a braking force applied by the brake to the wheel 301. Referring now to Fig. 36, a schematic diagram of an antilock braking system 400 for an in-wheel electric 15 vehicle according to another embodiment of the present disclosure is shown. The system 400 operates the method 200 shown and described in respect of Fig. 34. As illustrated in Fig. 36, the 20 system 400 is provided in respect of a wheel 401 in an in wheel electric vehicle which is driven by a multiphase electric motor 402 that produces a back EMF in each of the phases. For example, the motor may be a three phase brushless DC motor that produces a back EMF in three 25 phases. A first back EMF sensor 403 is provided to detect the back EMF produced by the motor 402 in one phase and obtain a signal corresponding to that back EMF. The system also comprises at least one other back EMF sensor 404 which detects the back EMF in the same phase or a 30 different phase to the first back EMF sensor 403. The system may comprise multiple back EMF sensors detecting the back EMF of different phases of the motor to obtain multiple back EMF signals corresponding to the back EMF of different phases of the motor. 35 Each of the signals generated by the two or more back EMF sensors 403, 404 are optionally received by a pre 5520179_1 (GHMatters) P97310.AU 25/06/14 - 29 processor 405 which subjects each signal received by the pre-processor to one or more pre-processing functions to improve the subsequent processing of the signals to estimate the wheel speed. One process that may be carried 5 out the by the pre-processor 405 is to smooth one or more of the obtained signals 406 using a peak detection module 407 and/or a transform module 408. The peak detection module 407 applies an algorithm in which the local maxima of the amplitude of the signal are identified in order to 10 smooth the signal and reduce the frequency content. The local maximum is defined as a data sample of the amplitude of the signal that is larger than its neighbouring values. The transform module 408 is operable to apply a discrete wavelet transform (DWT) in which a discrete wavelet 15 transform is used to decompose the wheel speed signal. In some embodiments, the signal is first preprocessed using the peak detection module 407 and subsequently preprocessed using the transform module 408. 20 The pre-processor 405 may also process one or more of the obtained signals to determine the surface condition 409 over which the wheel is travelling. The pre-processor does this by using a road condition module 410. The surface condition module 410 is operable to apply a 25 continuous wavelet transform (CWT) to one of the obtained signals and then using an energy index in which higher calculated values on the energy index indicate that the surface is more slippery. 30 The system 400 also comprises a processor 411 which receives the signals after any optional pre-processing by the pre-processor 405. The processer is programmed to derive an estimate of the wheel speed from the signals and produce an output indicative of the estimated wheel speed. 35 The processor does this by processing the signals to derive a wheel speed estimation 412 and combining information about the wheel speed from the signals 413. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 30 In one embodiment, this involves firstly carrying out the wheel speed estimation 412 to derive a wheel speed estimate from each signal and then combining wheel speed estimations 413 by averaging the wheel speed estimates. 5 However, in other embodiments (indicated by the dashed arrow in Fig. 36), the processor first combines the signals 413 using data fusion techniques to obtain a single fused signal and then derives a wheel speed estimate 412 from the single fused signal. 10 In both embodiments, a wheel speed estimator 414 is used to apply algorithms to the respective signals to obtain a wheel speed estimate. Also, an averaging module 415 is used to apply averaging algorithms such as ordered 15 weighted averaging, median averaging or mean averaging to the wheel speed estimates derived from each of the signals or to other components or derivations of the signals. The processor 411 produces an output indicative of the 20 wheel speed, which may be a value for the wheel speed or any other suitable input for a controller 416 in the antilock braking system 400 to enable the controller 416 to send appropriate commands to a brake 417 to control a braking force applied by the brake to the wheel 401. 25 It is noted that in the above system, the wheel speed estimate is obtained only from the back EMF of the phases of the multiphase motor. No ABS sensor or other type of sensor is required. 30 In a variation of the system shown in Fig. 36, only one back EMF signal is obtained and the wheel speed estimated from this one signal is processed to produce an output indicative of the wheel speed. 35 It is to be appreciated that the above methods are carried out and the above systems operated in real time, in 5520179_1 (GHMatters) P97310.AU 25/06/14 - 31 particular when the antilock braking system is required for use by the electric vehicle. Accordingly, the steps of the method and processes of the system are carried out continuously during this period to provide a continuous 5 estimation of the wheel speed. Examples of methods according to embodiments of the present disclosure and their implementation in the antilock braking systems of in-wheel electric vehicles 10 will now be described with reference to the accompanying Figures. Example 1: Brushed DC Electric Motor with ABS sensor 15 Referring to Figure 1, the general architecture of an in wheel electric vehicle antilock braking system 10 with four independently driven brushed DC electric motors is shown. The system 10 includes a processor in the form of a central brake controller 11 that is embedded within the 20 engine control unit (ECU) of the vehicle, four local controllers 12 (one for each wheel) and a communication network 13. In the in-wheel electric vehicle, an electric motor 14 is integrated into each wheel hub of the electric vehicle. Each wheel hub is also equipped with a 25 conventional ABS sensor 15 to realize a four channel antilock braking system (which provides maximum possible safety in emergency and panic braking situations). Data from each electric motor deriving from the motor's back EMF and each ABS sensor is transmitted to the central 30 brake controller via the communication network. The central brake controller receives the information from each wheel and processes the information to generate an output relating to the wheel speed for each wheel. Based on this output, the central brake controller generates the 35 appropriate commands to each local controller to optimize the braking performance of that wheel during activation of the antilock braking system. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 32 Slip ratio control in antilock braking systems requires the measurement of the slip ratio values at each wheel of the vehicle, because the road condition for each wheel can be 5 different. The antilock braking system 10 having the architecture described above and shown in Fig. 1 makes use of the synergistic combination of all three sources of wheel speed information for each wheel. Two sources of information reside in the ABS sensor output and the third 10 source is the back EMF output of the in-wheel motor. The first source of information for estimating the wheel speed is the frequency of the ABS sensor output. The frequency of the ABS sensor output is directly related to 15 the wheel rotational speed. To obtain and utilise a frequency signal from the ABS sensor output, the output signal is converted to a pulse signal (with the same frequency as the analogue output signal of the ABS sensor). The pulse signal is processed by the Electronic Control 20 Unit (ECU) of the vehicle into a value for the wheel speed from the period of the pulse signal. In this respect, the wheel speed (w) can be calculated as: 2wr 1 Z T 25 where r is the radius of the wheel and Z is the number of gear teeth. The wheel speed signal period (T) can be calculated by: T=N- T where N represents the number of adjacent pulses and T is 30 the cycle of a reference clock. The second source of information is the amplitudes of the ABS sensor output. Although wheel speed estimation from the amplitude of the ABS sensor signal is known to be noise 5520179_1 (GHMatters) P97310.AU 25/06/14 - 33 sensitive and has therefore not been used previously, as will described below, the inventors have surprisingly found that the use of this information in the antilock braking system 10 increases the overall accuracy of the wheel speed 5 determined by the system. The third source of information is the back EMF of the electric motor. The following equations describe the dynamics of a brushed DC motor during activation of the 10 antilock braking system: {Va=E+Raza E=KePW where w is the speed of armature (same as the wheel angular velocity), - is the flux per pole, K, is a 15 constant, Ra is the resistance of the armature circuit, Va is the armature voltage, Ia is the armature current and E is referred to as back EMF. It is common to disengage the regenerative braking torque 20 during activation of an antilock braking system. This means that, for a brushed DC motor, the armature circuit is an open circuit and Ia is equal to zero while the antilock braking system is activated. In these conditions, the armature voltage is the same as the back EMF and is 25 linearly related to the wheel speed. By measuring the armature voltage (ie. by obtaining a signal derived from the back EMF of the brushed DC motor), the third source of speed information for wheel speed estimation is captured. 30 Each of the above described signals (frequency and amplitude of the ABS senor output and the back EMF of the motor) are used to derive an estimate of the wheel speed and then fused or aggregated to produce an output relating to the wheel speed for each wheel. Fusing refers to the 35 process of combining data from a plurality of different sources, or combining data derived from a plurality of 5520179_1 (GHMatters) P97310.AU 25/06/14 - 34 different sources. It was found that by fusing the wheel speeds estimated from each of the signals, the reliability, fault-tolerance and accuracy of the wheel speed estimation was substantially improved. A number of different fusion 5 methods were used including using the minimum or maximum value and median and mean averaging. Another such method was Ordered Weighted Averaging (OWA). OWA operators provide a solution for the problem of 10 "aggregating multicriteria" which combine multiple criteria into one fused format. In the present disclosure, OWA is used to fuse the result of any number of different wheel speed signals into one unified result. OWA operators can be obtained by a number of different methods. It is within 15 the scope of a skilled person to be able to determine appropriate OWA operators including by the dispersion method and by using the exponential class of OWA operators including optimistic exponential OWA operators (OOWA) and pessimistic exponential OWA operators (POWA). 20 The performance of above described wheel speed estimation method for a brushed DC motor was demonstrated by modifying a standard ABS experimental test rig (shown in Fig. 2) to include an ABS sensor and a brushed DC motor. 25 The test rig is designed to simulate braking scenarios in which the antilock braking system is activated. The test rig is equipped with two wheels that simulate the vehicle and wheel motion. The "vehicle wheel" is the upper wheel 30 shown in Fig. 2. The upper wheel is driven by friction contact with the lower wheel. The lower wheel simulates relative road motion and is directly driven by a PWM controlled DC motor. Each of the two wheels are connected to an optical encoder with a resolution of 1024 CPR. The 35 encoder connected to the lower wheel is used to measure the vehicle speed (V) and the encoder connected to the upper wheel is used to measure the wheel speed (w). 5520179_1 (GHMatters) P97310.AU 25/06/14 - 35 The braking phase commences when the upper wheel reaches a predefined speed. At that point, the PWM-controlled DC drive (connected to the lower wheel) is switched off and 5 the upper wheel can be decelerated by a disk brake system. A separately excited brushed DC motor was coupled to the upper wheel to simulate the electric motor that drives a wheel in an in-wheel electric vehicle. The DC motor can 10 apply regenerative braking torque. However, the regenerative braking torque is not applied during the antilock braking system activation and the car wheel is solely decelerated by the frictional torque. The back EMF of the DC motor during frictional braking is used for wheel 15 speed estimation in the method as described. A commercially available ABS sensor (made by Repco for a family sedan) was also mounted on the test rig using a specifically designed coupling to enable the toothed ring 20 of the ABS sensor to rotate with the upper wheel shaft. The test rig was equipped with a multipurpose digital data acquisition board to measure the wheel speed provided by the optical encoders. A Daqbook/112 data acquisition system 25 available from Iotech was used to capture the back EMF signal of the DC motor (connected to the upper wheel) as well as the signal from the Repco ABS sensor. The Daqbook/112 can sample input signals with a high sampling rate of up to 100 kHz. Although a typical ABS system needs 30 a new measurement every 7ms, the ABS sensors need to be sampled at much higher rate to achieve good measurement accuracy. The inventors found that the ABS sensors need to be sampled at around 5KHz to achieve good accuracy. Similar sampling rates were used for the back EMF signal to achieve 35 good accuracy. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 36 Typical wheel speed measurements from the optical encoder (the reference measurement) before and after activation of the antilock braking system (during a complete cycle that 5 includes the acceleration and deceleration phases) are shown in Fig. 3. The back EMF signal shown in the bottom graph in Fig. 3 closely resembles the reference measurements shown in the top graph. The high correlation between the two signals shows that wheel speed can be 10 estimated from the back EMF signal. The ABS sensor output corresponding to the signal depicted in Fig. 3 is shown in Fig. 4. As mentioned earlier, the frequency and amplitude of this output can be derived and 15 used in the method of estimating the wheel speed. The frequency signal from the ABS sensor is obtained in the manner described above. The absolute values of the ABS output signal are plotted in 20 Fig. 5. This figure shows that there is a correlation between the absolute values of the ABS output and the speeds measured by the optical encoder (reference measurements). Those figures show that the peaks of the absolute values can be used to estimate the wheel speed. 25 The amplitude signal from the ABS sensor was denoised using a peak detection algorithm in which local maxima of the amplitude signal are identified to generate a smoother signal with a lower frequency content (similar to the reference wheel speed signal). A local maximum is defined 30 as a data sample of the amplitude signal that is larger than its neighboring values (immediate neighbors). The frequency and denoised amplitude signals from the ABS sensor output are shown in Figs. 6 and 7. 35 In order to calculate and compare errors in the estimation of wheel speed associated with the back EMF signal, and the frequency and amplitude signals from the ABS sensor as well 5520179_1 (GHMatters) P97310.AU 25/06/14 - 37 as different fusions of the wheel speeds derived from these signals, the following Mean Absolute Error (MAE) index was used: 5 MAE = -=i(i)-Sig(i) In the above definition, n is the length of signals during activation of the antilock braking system and R(i) and Sig(i) represent the measured values for the ith sample 10 point in the reference (encoder) and evaluated signals, respectively. In order to compare the results of fusion, the MAE index of different fusions are compared with the MAE index of the conventional method of calculating the speed using the frequency of the commercial ABS sensor' s 15 output by itself. With three sources of information (ABS sensor frequency and amplitude and motor back EMF), four different fused signals can be obtained, three derived from two sources and the 20 fourth using all three sources. Results of all possible fusion combinations of two sources are given in Table I below, calculated using mean averaging. The best results were achieved by the fusion of the back EMF and ABS sensor amplitude signals. The table shows that the fusion of any 25 two-signal combinations of the three available signals always decreases the error index. TABLE I IMPROVEMENT OF THE MAE INDEX FOR TWO-SIGNAL FUSION Fusion Freq- Amp Freq- EMF EMF-Amp MAE 27% 32% 38% Improvement 30 TABLE II below summarizes improvements of the MAE error index with different fusion methods of all three signals. In this table, the MAE index has been calculated for all sample points in the fused signal during activation of ABS. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 38 While minimum and maximum operators (ie. using the minimum or maximum value of the signals at a given point in time) both degraded the results in several cases, the dispersion based OWA method improved the error by up to 41 percent. 5 The median method also improved the MAE index by up to 46 percent. The mean, optimistic and pessimistic exponential OWA also improved the MAE index by up to 40 percent. Comparison of Table I and Table II shows that the fusion of the three signals always results in better MAE index 10 compared to two-signal method for the brushed DC motor. The results show that the inclusion of the back EMF signal in the fusion process increased the accuracy of wheel speed estimation significantly. 15 TABLE II IMPROVEMENT OF THE MAE INDEX FOR THREE-SIGNAL FUSION Fusion Improvement of Method the MAE Index Min -14% Max -5% Mean 40% Optimistic 39% OWA Pessimistic 39% OWA Dispersion 41% OWA Median 46% Results of fusing the three signals by the OWA (dispersion) and median methods are shown in Figs 8 and 9. Comparison of these two figures shows that the median method is more 20 accurate at estimating the speed at lower wheel speeds. However, the antilock braking system is normally de activated at low speeds and in this respect lower speed inaccuracies are not as significant to the operation of the antilock braking system. Accordingly, the MAE index for the 25 fused signals were recalculated by excluding lower wheel 5520179_1 (GHMatters) P97310.AU 25/06/14 - 39 speeds. The recalculated MAE index (ie. excluding low wheel speeds) found that the OWA dispersion method outperforms the median method by up to 5%. 5 Example 2: Permanent Magnet Brushless Motor with ABS sensor Permanent Magnet Brushless (PMBL) motors are highly efficient, small and require little maintenance and are 10 thus becoming the predominantly employed drives in in wheel electric vehicles. Permanent Magnet Synchronous Motors (PMSM) and Brushless DC (BLDC) motors are two types of PMBL motors that have 15 the same construction. The required magnetic fields of both motors are generated by the rotors having permanent magnets and their armature windings, which are connected to three-phase voltage sources, are located on their stators. The PMSM is fed by a three-phase sinusoidal 20 voltage whereas the BLDC motor is connected to a DC chopped three-phase voltage source. Although the back EMF shape in PMSM and BLDC are different when those are driven, the back EMF of both motors as generators are sinusoidal shape voltages. 25 A BLDC motor consists of a PMSM, a position sensor (Hall effect sensors) and an inverter that electronically implements the commutation for the motor. An electronically commutated BLDC motor is shown in Fig. 10. 30 The equivalent circuit of phase A of the BLDC motor is shown in Fig. 11. It is noted that phases B and C of the BLDC motor will have substantially identical circuits. The voltage of phase A can be expressed as: dI VA=RAIA+L-. +eA 35 where RA and LA are the equivalent resistance and 5520179_1 (GHMatters) P97310.AU 25/06/14 - 40 inductance of the stator phase winding, IA is the current flowing in phase A and eA is the back EMF of this phase. The back EMF of the phase A of a BLDC machine can be 5 expressed as: eA = Koacosee where we and Be are the electrical velocity and position of the rotor and K is a constant. This equation shows that the amplitude of the back EMF increases as the speed 10 increases. The inventors have realized that this is similar to the relationship between the amplitude of an ABS sensor output and the wheel rotational speed. The rotor electrical position (ee) is related to the 15 mechanical position (em) by: ee = Npem where N, is the number of pole pairs of the rotor. This shows that the back EMF signal completes Np cycles per each complete rotation of the BLDC rotor. The inventors have 20 further realized that this is also similar to the relationship between the voltage and frequency of the ABS sensor output which is expressed as: Sensor = K' c4 cos (NteetG) 25 where Vsensor is the voltage of the ABS sensor, w is the rotational speed of the sensor toothed ring, Nteeth is the number of teeth in the toothed ring and K' is a constant. The general architecture of the antilock braking system 30 for an in-wheel electric vehicle incorporating four BLDC motors is similar to the architecture described in respect of an in-wheel electric vehicle having brushed DC motors and illustrated in Fig. 1. The wheel speed information of each BLDC motor and each ABS sensor are transmitted to the 35 central brake controller located in the Electronic Control 5520179_1 (GHMatters) P97310.AU 25/06/14 - 41 Unit (ECU) of the vehicle via the communication network. The central brake controller issues required commands to each local controller to control the braking of that wheel according to the wheel speed calculated by the central 5 brake controller. A schematic diagram of the wheel speed estimation method for a single BLDC In-Wheel motor is provided in Fig. 12. This figure shows that the method can provide up to eight different wheel speed signals (referred to as amplitude 10 and frequency signals, respectively) that can be fused to estimate the wheel speed. An antilock braking system test rig was used to demonstrate the wheel estimation method according to 15 embodiments of the present disclosure for a wheel driven by a BLDC motor. The test rig is shown in Fig. 13 and is similar to the test rig shown in Fig. 2 except that a BLDC motor having 8 pole pairs (in addition to a Brushed DC motor) was incorporated into the rig. The BLDC motor can 20 be arranged to drive the upper wheel (instead of the Brushed DC motor). A commercially available ABS sensor is incorporated into the test rig. Also, a highly accurate optical encoder with a resolution of up to 2048 counts per cycles (CPR) was used to measure the upper wheel speed. 25 This highly accurate encoder provides reference wheel speed measurements in this study for comparison purposes. Simulation of the vehicle wheel motion during antilock braking system activation commences by bringing both the upper and lower wheels to a predefined speed that is the 30 initial speed for the start of antilock braking system activation. At this point, antilock braking system is activated and the upper wheel is decelerated by the frictional disk brake system while the lower wheel spins freely. 35 5520179_1 (GHMatters) P97310.AU 25/06/14 - 42 A coupling system was designed to enable the ABS sensor ring, the rotor of the BLDC motor and the DC motor armature would all rotate with the upper wheel (vehicle wheel) in the modified test rig. The wheel speed can be 5 estimated from the back EMF of either of the motors and then fused with measurements from the commercial ABS sensor. Results of fusing those measurements can then be compared with the reference wheel speed (measured by the accurate optical encoder of the test rig). 10 Trials were run in which the antilock braking system was activated when the upper wheel speed was around 900 RPM. The reference wheel speed measured by the test rig's optical encoder and the corresponding speed estimations 15 from frequency and amplitude of the ABS sensor output using the methods previously described are shown in Fig. 14. The results of estimating the wheel speed from the amplitude signal of three phases of the BLDC motor using corresponding to Fig. 14, are shown in Fig. 15. 20 To compare the accuracy of different wheel speed estimation techniques, the mean absolute error (MAE) index was calculated as previously described. Table III summarizes the results of calculating the MAE 25 index for different wheel speed measurements. Because the antilock braking system is normally deactivated when the vehicle speed is reduced to around 10 km/h, the MAE index calculations only incorporate measurements up to the point where the peak wheel velocity (representing vehicle speed) 30 is above 200 RPM. Calculated values (for non-fused signals) for MAE index listed in Table III show that the accuracy of the amplitude signals are always higher than that of the frequency signals from the same source 35 5520179_1 (GHMatters) P97310.AU 25/06/14 - 43 Table III. Results of calculating the MAE index for different wheel speed measurements. Method MAE index [RPM] 64.26 (conventional ABS ABS Sensor (Frequency Signal) sensor error) ABS Sensor (Amplitude Signal) 56.60 Phase 1 BLDC (Frequency Signal) 38.53 Phase 1 BLDC (Amplitude Signal) 31.24 Phase 2 BLDC (Frequency Signal) 52.10 Phase 2 BLDC (Amplitude Signal) 34.24 Phase 3 BLDC (Frequency Signal) 40.47 Phase 3 BLDC (Amplitude Signal) 33.76 Results of fusing the four amplitude signals (s1 to s4 5 signals in Fig. 13) of figures Figs. 14 and 15 by OOWA, POWA and median methods are shown in Figs. 16-18 respectively. The calculated values for MAE index corresponding to Figs. 16-18 can be compared with those of other fusion methods in Fig. 19. This figure shows that 10 the POWA fusion method has the minimum MAE index value. These tests were repeated for different road conditions (Dry and slippery roads) with different initial speeds for the start of the braking phase. These tests found that the 15 POWA fusion method (of the four amplitude signals: derived from the ABS sensor and the back EMF of three phases) can decrease the MAE value by up to 70 percent in comparison with ABS sensor measurements alone. 20 However, the median method accuracy was higher than all other fusion methods when all eight signals (s1 to s8 signals in Fig. 13) were fused, i.e. all amplitude and frequency signals from the ABS sensor output and the back EMF of the three phases of the BLDC motor. Results of 25 fusing all eight signals using the median method are shown in Fig. 20. Different road conditions showed that the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 44 median fusing method can decrease the MAE value by up to 76 percent in comparison with the commercial ABS sensor measurements alone. It is noted that although fusion of the all eight signals can produce a more accurate result, 5 it does require a higher computational load then fusing only the four amplitude signals. To test the robustness of the proposed fusion-based ABS, a trial was conducted in which the ABS sensor was short 10 circuited. Table IV summarizes the MAE values calculated for OOWA, POWA and median methods for this trial. The MAE values in the table show that all fusion methods (OOWA, POWA, Median), for the case when the ABS sensor was short circuited, could still provide acceptable if not improved 15 wheel speed estimations compared to a commercial ABS sensor alone (which had a MAE value of around 65 RPM). Table IV. MAE values calculated for OOWA, POWA and median methods where ABS sensor is short circuited. MAE Value [RPM] MAE Value [RPM] fusion of s1 to s4 fusion of s1 to s8 OOWA 39.96 112.54 POWA 30.27 71.58 Median 37.85 19.59 20 The accuracy of fusion based wheel speed estimation for a BLDC motor was also compared with that of a Brushed DC motor. These trials found that the MAE value for fusion of signals in respect of BLDC motors was less than fusion in 25 respect of Brushed DC motors. This was due, at least in part, to the BLDC motor having three phases (as opposed to the single phase of a Brushed DC motor) and thus is able to provide more signals to fuse. 30 5520179_1 (GHMatters) P97310.AU 25/06/14 - 45 Example 3: Brushless Permanent Magnet Motor without ABS sensor In a further embodiment, tests were carried out on an 5 antilock braking system which did not include a conventional ABS sensor. Rather, all information about the wheel speed is derived from the back EMF of the PMBL motor, specifically a BLDC motor that is used to drive the wheel. 10 ABS sensors used in conventional antilock braking systems have several disadvantages. The toothed ring that attaches to the wheel and rotates in the magnetic field of the ABS sensor, is relatively bulky and makes the dimensions of 15 the sensor large. Several components (such as the propulsion, suspension and braking systems) are integrated inside the limited space of the wheel hub in an in-wheel vehicle. As such, omitting the ABS sensor saves critical space and provides for a better in-wheel design. The 20 toothed ring is also vulnerable to dust and corrosion. Further, the ABS sensor output is sensitive to the gap space between the toothed ring and the permanent magnet of the sensor (which generates the magnetic field in which the toothed ring rotates). This space is hard to adjust 25 and a slight misalignment of this space influences the sensor's output. Additional wiring is also required for transmitting the output signal of the ABS sensor to the controller of the antilock braking system. This signal can be influenced by external disturbances. In contrast to 30 conventional vehicles, for which the conventional ABS sensors are designed, the existence of a powerful electric motor in the hubs of in-wheel electric vehicles, close to the ABS sensor, means that a significant source of disturbance, that can distort the ABS signal and 35 deteriorate its performance, is unavoidable. ABS sensor signals are also susceptible to mechanical faults in the wheel bearing assembly and electrical faults such as the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 46 occurrence of an open/short circuit and a change of wire resistance. The embodiment that is described below in which a 5 conventional ABS sensor is not included in the antilock braking system mitigates at least some of the above issues. It should also be noted that the assembly, inspection and maintenance of in-wheel vehicle hubs are more complex and time-consuming compared to conventional 10 vehicle wheel hubs. This embodiment of the antilock braking system simplifies the inspection and maintenance procedures of in-wheel electric vehicles and reduces the cost and on-going maintenance costs of the vehicle. 15 The measurement resolution of a conventional ABS sensor is related to the number of teeth in the toothed ring (typically 48 teeth). Without any signal processing, the BLDC motor in an in-wheel electric vehicle must have enough pole pairs (at least 24 pole pairs) to provide the 20 same measurement accuracy and resolution compared to an ABS sensor. To overcome the above restriction, the antilock braking system without an ABS sensor processes the signals from 25 the BLDC motor's back EMF to improve the accuracy of the wheel speed estimation from these sources. These processing techniques alleviate the constraint on the number of pole pairs, in some cases down to only two pole pairs, where the wheel speed measurement accuracy is 30 comparable to conventional ABS sensor measurements. For the amplitude of the back EMF, one such technique is using the local maxima of the amplitude signal to generate a smoother signal with a lower frequency content. A local 35 maximum is defined as a data sample of the amplitude signal that is larger than its neighboring values (immediate neighbors). The combination of peak detection 5520179_1 (GHMatters) P97310.AU 25/06/14 - 47 (that identifies those local maximums) and extrapolation algorithms are then used to reconstruct the back EMF amplitude signal, which provides a measurement of the wheel speed. 5 Another technique involves using a Discrete Wavelet Transform. A Discrete Wavelet Transform (DWT) of a discrete signal x[n] is given by: DWT U, k] = Yxn]hj (n - 2'k] where n represent sample points of the signal and j and k 10 (j]EN,k EZ) are scale and shift parameters, respectively. Discrete wavelet transform uses a dyadic time-frequency decomposition structure that provides low and high frequency components of a given signal at each 15 decomposition level. The low and high frequency components are referred to as Approximations (Ai) and Details (Di) at level i, respectively (see Fig. 21). While Approximations of a given signal represent the smooth variations, the Details are related to abrupt changes in the signal. 20 The above mentioned amplitude peak detection technique is able to provide wheel speed estimation using the BLDC back EMF during ABS activation. However, the wheel speed accuracy obtained by using this method deteriorates for 25 BLDC motors with a very low number of pole pairs (e.g. two pole pairs). The wheel speed provided by the amplitude signal for a BLDC motor with two pole pairs has a higher frequency content compared to its counterpart signal for a BLDC motor with eight pole pairs. This high frequency 30 content contains abrupt changes that degrade the accuracy of wheel speed estimations for BLDC motors with a low number of pole pairs. The accuracy of wheel speed estimations for BLDC motors 35 with a very low number of pole pairs, decomposing the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 48 amplitude signal using DWT decomposition with a structure that is shown in Fig. 21 is carried out. The technique based on DWT identifies the low frequency component (Approximations) that contains accurate wheel speed 5 estimations. The wheel speed estimation error significantly decreases by using the Approximations of the inaccurate wheel speed estimation (from the amplitude signal of the BLDC back EMF) at level seven (A 7 ), for a BLDC motor with only two pole pairs. 10 The present inventors have also observed that the braking process on roads with different surface conditions (such as dry or icy) results in different features in the back EMF signal of the BLDC motor. These features can be 15 extracted by the application of a continuous wavelet transform on the back EMF signal during ABS activation to provide an identification of the road condition and thus optimize performance of the antilock braking system. 20 A Continuous Wavelet Transform (CWT) of signal x(t) is given by: 1 r+* tt- a ) CWT (a, b) = x(t)* ( dt where 0*(t) is a complex conjugate of a wavelet function and a and b are scale and shift values, respectively. 25 Unlike DWT, the scale and shift parameters in the continuous wavelet transform are real values. This enables an effective time-frequency representation of a signal that can be used for local feature extraction. The implementation detail of the CWT-based feature 30 extraction algorithm for road identification is as follows. The first step is to compute the CWT of the amplitude of the BLDC back EMF signal using the Morlet wavelet. The following energy index is then used to distinguish between different road conditions: 5520179_1 (GHMatters) P97310.AU 25/06/14 - 49 EnergyIndex(n) = Coef 1 (s) 2 where n represent sample points of the BLDC back EMF signal and s and Coef 1 are CWT scales and coefficients, respectively. 5 The road condition features can be extracted by calculating the above energy index for the amplitude of the back EMF of the BLDC motor. The above energy index for road identification is an appropriate measure for road identification purposes, because its values are 10 significantly higher for slippery roads (e.g. icy or oily roads) compared to roads with higher friction coefficients (e.g. dry asphalt). To demonstrate the feasibility of the antilock braking 15 system for a BLDC driven in-wheel electric vehicles without a conventional ABS sensor, an antilock braking system test rig, shown in Fig. 22 was used. The test rig included different BLDC motors (with different number of pole pairs) representing the propulsion system of a 20 vehicle. The test rig includes two wheels which simulate the car wheel motion and the relative motion of the vehicle by the upper and lower wheels, respectively. Tests were carried out using the test rig which each 25 commenced by bringing both the upper and the lower wheels to a predefined initial speed, which is the vehicle wheel speed at the start of the braking phase. For this purpose, a driving motor is coupled to the lower wheel to achieve the desired velocity. When the speed of the upper wheel 30 (vehicle wheel) reached a predefined value (initial speed of the braking phase), the driving motor was switched off and the braking phase started. The braking mechanism in the test rig is based on using a disk brake, which is mounted on the upper wheel. The antilock braking system of 5520179_1 (GHMatters) P97310.AU 25/06/14 - 50 this test rig used the on-off (bang-bang) controller to avoid locking of the upper wheel. Rotational velocities of both wheels were measured using two highly accurate optical encoders. 5 In addition to including two different BLDC motors (with two and eight pole pairs), a commercial ABS sensor (used in many sedan vehicles, made by Repco) was also coupled to the upper wheel to compare the result of wheel speed 10 estimation with and without a commonly used commercial ABS sensor. To compare the results of the BLDC motor with a DC motor back EMF speed estimation during activation of the antilock braking system (and without an ABS sensor), a DC motor was also connected to the upper wheel shaft. 15 The reference wheel speed provided by the optical encoder before and after antilock braking system activation (at around 900 RPM), and its corresponding ABS sensor output are shown in Figs. 23 and 24. The wheel speed measurements 20 (during ABS activation) from the frequency of the ABS sensor output signal in accordance with the method described in Example 1 is shown in Fig. 25. To compare the accuracy of different wheel speed 25 measurement methods, we calculated the Mean Absolute Error (MAE), as described above, and the Root Mean Square Error (RMSE) indices for measurement results of those techniques. The Root Mean Square Error (RMSE), which emphasizes the importance of larger errors, is defined as: RMSE=

>

1 Sig(i) - Ref(i)) n 30 The MAE and RMSE values for wheel speed estimations using the frequency of the ABS sensor output alone during antilock braking system activation are calculated to be 70.6 and 127.2, respectively. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 51 The amplitude signal of the back EMF of one phase of the BLDC motor with 8 pole pairs is shown in Fig. 26. In order to show that the wheel speed can be estimated from the amplitude of the BLDC motor back EMF, the absolute values 5 of the amplitudes of the back EMF signal shown in Fig. 26 are likewise plotted in Fig. 27. This figure shows that there is a high correlation between the absolute values of the amplitude signal of the BLDC motor back EMF and the reference measurements produced by the optical encoders. 10 The signal shown in the bottom part of Fig. 27 has a higher frequency content compared with the reference signal (top part of Fig. 27). The combination of peak detection and extrapolation algorithms as described above was therefore applied to the amplitude signal from the 15 BLDC back EMF. The result of applying these algorithms on the amplitude signal from the BLDC back EMG is shown in Fig. 28. The MAE and RMSE values for the wheel speed estimation 20 using the peak detection method (shown in Fig. 28) are 29.5 and 57.5, respectively. These results show that the MAE and RMSE of the amplitude method are 58.2 and 54.8 percent lower than their corresponding values for the ABS sensor measurements. Results of calculating the MAE and 25 RMSE values for numerous experiments (with different road conditions and initial speeds for the start of the braking phase) showed that the accuracy of wheel speed estimation from applying the peak detection method to the amplitude signal of the back EMF of one phase of the BLDC motor is 30 always significantly higher than the accuracy of the ABS sensor measurements. The peak detection method and the DWT method (described above) was applied to the amplitude signal from the back 35 EMF of one phase of the BLDC motor with only two pole pairs. Fig. 29 shows that the DWT method successfully 5520179_1 (GHMatters) P97310.AU 25/06/14 - 52 extracted an underlying smooth wheel speed signal whereas the peak detection does not. In applying the DWT method, a suitable wavelet shape has 5 to be selected. Different wavelet shapes; Haar, db2-dblO, sym2-sym8, coif1-5, biorl.1-bior6.8, rbiol.1-rbio6.8 and dmey wavelets were tested. The dmey wavelet extracted the most similar underlying wheel speed signal as shown in Fig. 29. Results of calculating the MAE and RMSE values 10 for the wheel speed estimation (shown in Fig. 29) by using the DWT method showed that the method's accuracy is significantly higher compared to that of ABS sensor (MAE and RMSE values decrease 34 and 48 percent, respectively). 15 Further tests were conducted in which it was simulated that a phase of the BLDC motor is short-circuited. In these tests, the amplitude signals of the back EMF from the other two phases of the BLDC motor were obtained and fused using a median fusion operator. The results of these 20 tests showed that accuracy of the fused results using the median operator is higher than the accuracy of frequency of the ABS sensor output signal alone, even though there is a fault in the system. Figure 30 shows the result of fusing the amplitude signals of the back EMF from the two 25 phases of the BLDC motor by a median operator (after applying the peak detection method to each of the signals). The figure shows that the MAE for the fused result (16 RPM) is 77 percent less than that of ABS sensor measurements (71 RPM), despite the fact that one phase of 30 the BLDC is short circuited. Wheel speed estimation from the DC motor back EMF is shown in Fig. 31. The RMSE values were approximately the same for both brushed and brushless motors. However, the MAE 35 value for the brushed DC motor was calculated to be seven percent higher compared to that of BLDC motor back EMF speed measurement, which shows that wheel speed estimation 5520179_1 (GHMatters) P97310.AU 25/06/14 - 53 for the BLDC with 8 pole pairs (using the peak detection method) is more accurate compared to DC motor back EMF wheel speed estimation. 5 To demonstrate the road identification method, outlined above, several tests were carried out with two different road conditions (slippery road and dry road) to produce BLDC back EMF signals during antilock braking system activation. To simulate slippery road conditions, oil was 10 applied to the wheels in the test rig. The back EMF signals (related to dry and slippery road condition) were captured and analyzed with different continuous wavelet transforms. Wavelet shape impacts road identification. The inventors found that the Morlet wavelet is the most 15 suitable wavelet for this purpose. Fig. 32 shows the calculated energy index in different tests (with different road conditions and different initial speed for the start of the braking phase), using the Morlet wavelet. The results show that the proposed energy index is sensitive 20 to road conditions as its value significantly increases for slippery road compared to dry road and can be used for road identification purposes. It is to be understood that, if any prior art publication 25 is referred to herein, such reference does not constitute an admission that the publication forms a part of the common general knowledge in the art, in Australia or any other country. 30 In the claims which follow and in the preceding description of the invention, except where the context requires otherwise due to express language or necessary implication, the word "comprise" or variations such as "comprises" or "comprising" is used in an inclusive sense, 35 i.e. to specify the presence of the stated features but not to preclude the presence or addition of further features in various embodiments of the invention. 5520179_1 (GHMatters) P97310.AU 25/06/14

Claims (52)

1. A method of estimating the wheel speed of a wheel of an in-wheel electric vehicle, the vehicle having an 5 electric motor which produces a back EMF when driving that wheel, the method comprising: obtaining two or more signals which are related to the wheel speed, at least one signal being obtained from the back EMF of the electric motor that drives the wheel; 10 processing the two or more signals to derive an estimate of the wheel speed; and producing an output indicative of the wheel speed from the wheel speed estimate. 15 2. A method as claimed in claim 1, wherein the method comprises deriving an estimate of the wheel speed from each obtained signal and producing the wheel speed output by averaging the estimated wheel speeds. 20 3. A method as claimed in claim 1, wherein the method comprises fusing each of the obtained signals into a fused signal and deriving the estimate of the wheel speed from the fused signal. 25 4. A method as claimed in any one of the preceding claims, wherein the electric motor is a three phase motor and all of the two or more signals are obtained from the back EMF of the electric motor that drives the wheel. 30 5. A method as claimed in any one of claims 1 - 3, wherein the electric motor is a three phase motor and each signal obtained from the back EMF is obtained from the back EMF of a single phase of the motor. 35 6. A method as claimed in any one of claims 1 - 3, wherein the electric motor is a three phase motor and at least two of the two or more signals are obtained from the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 55 back EMF, wherein each of said back EMF signals are obtained from the back EMF of different phases of the motor. 5 7. A method as claimed in any one of the preceding claims, wherein each signal obtained from the back EMF is a frequency or an amplitude of the back EMF.

8. A method as claimed in any one of the preceding 10 claims, wherein the electric motor is a three phase motor and the obtained signals include a back EMF amplitude and a back EMF frequency from the back EMF of a single phase of the motor. 15 9. A method as claimed in any one of claims 1 - 7, wherein the electric motor is a three phase motor and the obtained signals include a back EMF amplitude and a back EMF frequency from each of the back EMF of two or more phases of the motor. 20

10. A method as claimed in any one of claims 1-3 and 5-9 except when dependent on claim 4, wherein at least one of the signals is obtained from an output of an antilock braking system (ABS) sensor. 25

11. A method as claimed in claim 10, wherein the ABS sensor comprises a member attached to the wheel so as to rotate with the wheel, a magnetic field generator which generates a magnetic field in which the member rotates and 30 a wire coil in which a current output is generated by the rotation of the member in the magnetic field, wherein each ABS sensor signal is obtained from the current output.

12. A method as claimed in claims 10 or 11, wherein the 35 ABS sensor signal is the frequency or the amplitude of the ABS sensor output. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 56 13. A method as claimed in any one of claims 10 - 12, wherein the obtained signals comprise the frequency and the amplitude of the ABS sensor output. 5 14. A method as claimed in any one of the preceding claims when dependent on claim 2, wherein averaging the wheel speed estimates comprises applying an averaging algorithm to the derived wheel speed estimates. 10 15. A method as claimed in any one of claims 1 - 13, when dependent on claim 3, wherein fusing each of the obtained signals into a fused signal comprises applying an averaging algorithm to the obtained signals. 15 16. A method as claimed in claims 14 or 15, wherein the averaging algorithm is selected from the group consisting of ordered weighted averaging (OWA), median averaging and mean averaging. 20 17. A method as claimed in any one of the preceding claims, wherein deriving the wheel speed estimate comprises pre-processing at least one of the two or more signals before processing the two or more signals. 25 18. A method as claimed in claim 17, wherein the pre processing comprises smoothing the signal using local maxima.

19. A method of as claimed in claim 17 or 18, wherein the 30 pre-processing comprises applying a discrete wavelet transform to the signal.

20. An apparatus for estimating the wheel speed of at least one wheel of an in-wheel electric vehicle, the 35 vehicle having an electric motor which produces a back EMF when driving that wheel, the apparatus comprising: two or more sensors each configured to obtain a 5520179_1 (GHMatters) P97310.AU 25/06/14 - 57 signal which is related to the wheel speed, wherein at least one sensor is a back EMF sensor that obtains a signal corresponding to the back EMF of the electric motor that drives the wheel; 5 a processor operable to receive and process the signals from the two or more sensors to derive an estimate of the wheel speed and produce an output indicative of the wheel speed from the wheel speed estimate. 10 21. An apparatus as claimed in claim 20, wherein the electric motor is a three phase motor and the apparatus comprises two or more back EMF sensors each configured to obtain a signal corresponding to the back EMF of the electric motor that drives the wheel. 15

22. An apparatus as claimed in claims 20 or 21, wherein the electric motor is a three phase motor and each sensor is a back EMF sensor that is configured to obtain a signal corresponding to the back EMF of a single phase of the 20 motor.

23. An apparatus as claimed in claims 20 or 21, wherein the electric motor is a three phase motor and at least two of the two or more sensors are back EMF sensors, each back 25 EMF sensor configured to obtain a signal corresponding to the back EMF of different phases of the motor.

24. An apparatus as claimed in any one of claims 20, 21 or 23, wherein at least one of the sensors is an antilock 30 braking system (ABS) sensor comprising a member attached to the wheel so as to rotate with the wheel, a magnetic field generator which generates a magnetic field in which the member rotates and a wire coil in which a current output is generated by the rotation of the member in the 35 magnetic field, wherein the ABS sensor is configured to obtain a signal corresponding to the current output. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 58 25. An apparatus as claimed in any one of claims 20 - 24, wherein the apparatus comprises a pre-processor operable to receive and pre-process the signal from at least one of the two or more sensors and sending each pre-processed 5 signal to the processor.

26. An antilock braking system for an in-wheel electric vehicle having a plurality of independent electric motors, 10 each motor producing a back EMF when driving a respective wheel, the system comprising: two or more sensors that are each arranged to obtain a signal related to the wheel speed of one of the wheels, at least one of the sensors comprising a back EMF sensor 15 that obtains a signal corresponding to the back EMF of the motor that drives said wheel; and a processor that receives the signals from each sensor, the processor programmed to derive an estimate of the wheel speed and produce an output indicative of the 20 wheel speed from the wheel speed estimate for use in controlling a braking force applied to said wheel.

27. An antilock braking system as claimed in claim 26, wherein the processor derives an estimate of the wheel 25 speed from each obtained signal and produces the wheel speed output by averaging the estimated wheel speeds.

28. An antilock braking system as claimed in claim 26, wherein the processor is programmed to fuse each of the 30 obtained signals into a fused signal and derive the estimate of the wheel speed from the fused signal.

29. An antilock braking system as claimed in any one of claims 26 - 28, wherein the electric motor is a three 35 phase motor and each of the sensors is a back EMF sensor that obtains a signal corresponding to the back EMF of the motor that drives the wheel. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 59 30. An antilock braking system as claimed in any one of claims 26 - 29, wherein the electric motor is a three phase motor and at least two of the sensors are back EMF 5 sensors that obtain a signal corresponding to the back EMF of a single phase of the motor.

31. An antilock braking system as claimed in any one of claims 26 - 29, wherein the electric motor is a three 10 phase motor and at least two of the two or more sensors are back EMF sensors, wherein each of said back EMF sensors are configured to obtain a signal corresponding to the back EMF of different phases of the motor. 15 32. An antilock braking system as claimed in any one of claims 26 - 31, wherein each back EMF sensor obtains a signal derived from a frequency or an amplitude of the back EMF. 20 33. An antilock braking system as claimed in any one of claims 26 - 32, wherein the electric motor is a three phase motor and at least two of the sensors are back EMF sensors, one back EMF sensor obtaining a signal derived from a back EMF amplitude and another back EMF sensor 25 obtaining a signal derived from a back EMF frequency of a single phase of the motor.

34. An antilock braking system as claimed in any one of claims 26 - 28, wherein the electric motor is a three 30 phase motor and at least two of the sensors are back EMF sensors, two back EMF sensors obtaining signals derived from a back EMF amplitude of two phases of the motor respectively and another two back EMF sensor obtaining a signals derived from a back EMF frequency of two or more 35 phases of the motor respectively.

35. An antilock braking system as claimed in any one of 5520179_1 (GHMatters) P97310.AU 25/06/14 - 60 claims 26 - 28 and 30 - 34 except when dependent on claim 29, wherein at least one of the two or more sensors comprises an antilock braking system (ABS) sensor comprising a member attached to the wheel so as to rotate 5 with the wheel, a magnetic field generator which generates a magnetic field in which the member rotates and a wire coil in which a current output is generated by the rotation of the member in the magnetic field, wherein at least one ABS sensor signal is obtained corresponding to 10 the current output.

36. An antilock braking system as claimed in claim 35, wherein the at least one ABS sensor signal is the frequency or the amplitude of the ABS sensor output. 15

37. An antilock braking system as claimed in any one of claims 35 or 36, wherein the signals obtained from the ABS sensor include the frequency and the amplitude of the ABS sensor output. 20

38. An antilock braking system as claimed in any one of claims 26 - 37, when dependent on claim 27, wherein the processor is programmed to average the wheel speed estimates by applying an averaging algorithm to the wheel 25 speed estimates derived from each of the signals received by the processor from the transmitter.

39. An antilock braking system as claimed in any one of claims 26 - 37, when dependent on claim 28, wherein the 30 processor is programmed to fuse each of the obtained signals into the fused signal by applying an averaging algorithm to the obtained signals.

40. An antilock braking system as claimed in claim 38 or 35 39, wherein the averaging algorithm is selected from the group consisting of ordered weighted averaging (OWA), median averaging and mean averaging. 5520179_1 (GHMatters) P97310.AU 25/06/14 - 61 41. An antilock braking system as claimed in any one of claims 26 - 40, wherein the system comprises a pre processor that receives the signal from at least one of 5 the two or more sensors and is programmed to pre-process said received signals and send each pre-processed signal to the processor.

42. An antilock braking system as claimed in claim 41, 10 wherein the pre-processor is programmed to pre-process the signal received from at least one of the two or more sensors by smoothing those signal(s) using local maxima.

43. An antilock braking system as claimed in claim 41 or 15 42, wherein the pre-processor is programmed to pre-process the signal received from at least one of the two or more sensors by applying a discrete wavelet transform to those signal(s). 20 44. A method of estimating the wheel speed of a wheel of an in-wheel electric vehicle, the vehicle having a multiple phase motor which produces a back EMF in each phase when driving said wheel, the method comprising: obtaining a signal from the amplitude of the back EMF 25 of one phase of the motor; and processing the back EMF amplitude signal to derive an estimate of the wheel speed from the back EMF amplitude signal; and producing an output indicative of the wheel speed 30 from the wheel speed estimate.

45. A method as claimed in claim 44, wherein the method comprises obtaining two or more signals, each signal being the amplitude of the back EMF of a respective phase of the 35 motor, processing the two or more signals to derive an estimate of the wheel speed and producing the output indicative of the wheel speed from the wheel speed 5520179_1 (GHMatters) P97310.AU 25/06/14 - 62 estimate.

46. A method as claimed in claims 44 or 45, wherein the method comprises obtaining one or more additional signals, 5 each additional signal comprising the frequency of the back EMF of one phase of the motor, processing the obtained signals to derive an estimate of the wheel speed and producing the output indicative of the wheel speed from the wheel speed estimate. 10

47. A method as claimed in any one of claims 44 - 46, wherein the method comprising obtaining at least one signal from an output of an antilock braking system (ABS) sensor, processing the obtained signals to derive an 15 estimate of the wheel speed and producing the output indicative of the wheel speed from the wheel speed estimate.

48. A method as claimed in claim 47, wherein the ABS 20 sensor comprises a member attached to the wheel so as to rotate with the wheel, a magnetic field generator which generates a magnetic field in which the member rotates and a wire coil in which a current output is generated by the rotation of the member in the magnetic field, wherein each 25 ABS sensor signal is obtained from the current output.

49. A method as claimed in claims 48 or 49, wherein the ABS sensor signal is the frequency or the amplitude of the ABS sensor output. 30

50. A method as claimed in any one of claims 48 - 49, wherein the obtained ABS sensor signals comprise the frequency and the amplitude of the ABS sensor output. 35 51. A method as claimed in any one of claims 45 - 50, wherein the method comprises deriving an estimate of the wheel speed from each obtained signal and producing the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 63 wheel speed output by averaging the estimated wheel speeds.

52. A method as claimed in any one of claims 45 - 50, 5 wherein the method comprises fusing each of the obtained signals into a fused signal and deriving the estimate of the wheel speed from the fused signal.

53. A method as claimed in claim 51, wherein averaging 10 the wheel estimates comprises applying an averaging algorithm to the wheel speed estimates derived from each signal.

54. A method as claimed in claim 52, wherein fusing each 15 of the obtained signals into a fused signal comprises applying an averaging algorithm to the obtained signals.

55. A method as claimed in claim 53 or 54, wherein the averaging algorithm is selected from the group consisting 20 of ordered weighted averaging (OWA), median averaging and mean averaging.

56. A method as claimed in any one of claims 44 - 55, wherein deriving the wheel speed estimate comprises pre 25 processing before processing the back EMF amplitude signal.

57. A method as claimed in claim 56, wherein pre processing comprises smoothing the back EMF amplitude 30 signal using local maxima.

58. A method of as claimed in claim 56 or 57, wherein pre-processing comprises applying a discrete wavelet transform to the back EMF amplitude signal. 35

59. An apparatus for estimating the wheel speed of at least one wheel of an in-wheel electric vehicle, the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 64 vehicle having a multiple phase motor which produces a back EMF in each phase when driving said wheel, the apparatus comprising: a sensor configured to obtain a signal from the 5 amplitude of the back EMF of one phase of the motor; a processor operable to receive and process the signal from the back EMF amplitude sensor to derive an estimate of the wheel speed and produce an output indicative of the wheel speed from the wheel speed 10 estimate.

60. An apparatus as claimed in claim 59, wherein the apparatus comprises two or more sensors, each sensor configured to obtain a signal corresponding to the 15 amplitude of the back EMF of a respective phase of the motor, and wherein the processor is operable to process the signals from the two or more sensors to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the wheel speed 20 estimate.

61. An apparatus as claimed in claims 59 or 60, wherein the apparatus comprises one or more additional sensors, each additional sensor configured to obtain a signal 25 corresponding to the frequency of the back EMF of one phase of the motor, wherein the processor is operable to process the obtained signals to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the wheel speed estimate. 30

62. An apparatus as claimed in any one of claims 59 - 61, wherein the apparatus comprises an antilock braking system (ABS) sensor, the ABS sensor comprises a member attached to the wheel so as to rotate with the wheel, a magnetic 35 field generator which generates a magnetic field in which the member rotates and a wire coil in which a current output is generated by the rotation of the member in the 5520179_1 (GHMatters) P97310.AU 25/06/14 - 65 magnetic field, wherein the ABS sensor is configured to obtain a signal corresponding to the current output and wherein the processor is operable to process the obtained signals to derive an estimate of the wheel speed and 5 produce the output indicative of the wheel speed from the wheel speed estimate.

63. An antilock braking system for an in-wheel electric vehicle having a plurality of independent multiple phase 10 motors, each motor producing a back EMF in each phase when driving a respective wheel, the system comprising: a sensor that is arranged to obtain a signal corresponding to the amplitude of the back EMF of one phase of one of the motors; and 15 a processor arranged to receive the amplitude signal from the sensor and programmed to derive an estimate of the wheel speed of the wheel driven by said motor from the amplitude signal and produce an output indicative of the wheel speed for use in controlling a braking force applied 20 to said wheel.

64. An antilock braking system as claimed in claim 63, wherein the system comprises two or more sensors, each sensor arranged to obtain a signal corresponding to the 25 amplitude of the back EMF of a respective phase of said motor and the processor is arranged to receive each signal and is programmed to process the signals from the two or more sensors to derive an estimate of the wheel speed and produce the 30 output indicative of the wheel speed from the wheel speed estimate.

65. An antilock braking system as claimed in claims 63 or 64, wherein the system comprises one or more additional 35 sensors, each additional sensor configured to obtain a signal corresponding to the frequency of the back EMF of one phase of the motor, wherein the processor is operable 5520179_1 (GHMatters) P97310.AU 25/06/14 - 66 to process the obtained signals to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the wheel speed estimate. 5 66. An antilock braking system as claimed in any one of claims 63 - 65, wherein the system comprises an antilock braking system (ABS) sensor, the ABS sensor comprising a member attached to the wheel so as to rotate with the wheel, a magnetic field generator which generates a 10 magnetic field in which the member rotates and a wire coil in which a current output is generated by the rotation of the member in the magnetic field, wherein the ABS sensor is configured to obtain a signal corresponding to the current output and wherein the processor is operable to 15 process the obtained signals to derive an estimate of the wheel speed and produce the output indicative of the wheel speed from the wheel speed estimate.

67. An antilock braking system as claimed in claim 65 or 20 66, wherein the at least one ABS sensor signal comprises the frequency or the amplitude of the ABS sensor output.

68. An antilock braking system as claimed in any one of claims 65 - 67, wherein the at least one ABS sensor signal 25 comprises the frequency and the amplitude of the ABS sensor output.

69. An antilock braking system as claimed in any one of claims 64 - 68, wherein the processor derives an estimate 30 of the wheel speed from each obtained signal and produces the wheel speed output by averaging the estimated wheel speeds.

70. An antilock braking system as claimed in any one of 35 claims 64 - 68, wherein the processor is programmed to fuse each of the obtained signals into a fused signal and derive the estimate of the wheel speed from the fused 5520179_1 (GHMatters) P97310.AU 25/06/14 - 67 signal.

71. An antilock braking system as claimed claim 69, wherein the processor is programmed to average the wheel 5 speed estimates by applying an averaging algorithm to the wheel speed estimates derived from each of the signals received by the processor.

72. An antilock braking system as claimed in claim 70, 10 wherein the processor is programmed to fuse each of the obtained signals into the fused signal and derive the estimate of the wheel speed from the fused signal.

73. An antilock braking system as claimed in claim 71 or 15 72, wherein the averaging algorithm is selected from the group consisting of ordered weighted averaging (OWA), median averaging and mean averaging.

74. An antilock braking system as claimed in any one of 20 claims 63 - 73, wherein the system comprises a pre processor that receives the amplitude signal received from the sensor and is programmed to pre-process said received signal and send the pre-processed signal to the processor. 25 75. An antilock braking system as claimed in claim 74, wherein processor is programmed to pre-process the amplitude signal received from the sensor by smoothing the signal using local maxima. 30 76. An antilock braking system of as claimed in claim 74 or 75, wherein the processor is programmed to pre-process the amplitude signal received from the sensor by applying a discrete wavelet transform to the signal. 35 5520179_1 (GHMatters) P97310.AU 25/06/14