DE102017006260A1 - Method for determining detection properties of at least one environmental sensor in a vehicle and vehicle, designed to carry out such a method - Google Patents

Abstract

The invention relates to a method for determining detection properties of at least one environmental sensor (3) in a vehicle (1) comprising the following steps: reading out a plurality of environmental sensors (3) of a vehicle (1) each in a plurality of measurement cycles; Identifying a same target object (7) in each measurement cycle of the plurality of measurement cycles, wherein it is checked for at least one environmental sensor (3) of the plurality of environmental sensors (3) to be tested for each measurement cycle of the plurality of measurement cycles, if the target object (7) of the environmental sensor (3) to be tested has been detected in the respective measuring cycle; Forming a detection number for the environmental sensor (3) to be tested as a number of those measurement cycles of the plurality of measurement cycles in which the target object (7) was detected by the environment sensor (3) to be tested, and determining a functional dimension for the environmental sensor (3) to be tested from the detection number.

Description

The invention relates to a method for determining detection properties of at least one environmental sensor in a vehicle as well as to a vehicle that is set up for carrying out such a method.

From the German patent application DE 10 2015 005 961 A1 is a method for monitoring the measurement signal of a sensor from a sensor composite forth, in which a cross comparison of measurement signals of the individual sensors of the sensor network is made, wherein average values of the deviations of the measurement signals are detected to each other, and wherein a permanent deviation of the average values from zero to a relative decalibration of one of the sensors is closed. Using this method, it is not or at least not readily possible to obtain a numerical function measure for a single sensor, since the individual sensor is always examined only with regard to the other sensors in the sense of a relative decalibration. However, it is also important or at least advantageous to determine independent functional dimensions for individual sensors. One such relevant measure of performance that indicates the quality of a sensor is the rate of detection, which indicates how often a sensor is capable of recognizing a real existing object in the real world and displaying it via an appropriate interface. The detection rate defined in this way is also called the "true positive rate" (TP rate for short). Modern driver assistance systems place very high demands on the detection rates of the sensors used, since this is the only way to ensure accident-free driving. In a known manner, detection rates for individual environmental sensors during a development cycle are measured by using the environmental sensors in an environment with known objects and comparing the objects detected by the sensors with the real world (ground truth).

The detection rates can be determined directly and tested under different conditions. A monitoring of the sensors in real, regular operation of a vehicle is not possible in this way, since the properties of the real environment and in particular occurring there objects are not previously known and can not be determined in real driving with independent of the vehicle technology.

The invention has for its object to provide a method and a vehicle which is adapted to carry out such a method, wherein said disadvantages do not occur.

The object is achieved by providing the subject matters of the independent claims. Advantageous embodiments emerge from the subclaims.

The object is achieved in particular by providing a method for determining detection properties of at least one environmental sensor in a vehicle, comprising the following steps: A plurality of environmental sensors of a vehicle are respectively read in a plurality of measurement cycles. The plurality of measurement cycles preferably includes a predetermined total number of measurement cycles. This means, in particular, that in each measuring cycle of the plurality of measuring cycles each environmental sensor of the plurality of environmental sensors is read out. In each measurement cycle of the plurality of measurement cycles, a same target object - for all environmental sensors - is identified, wherein it is checked for at least one environmental sensor to be tested of the plurality of environmental sensors for each measurement cycle of the plurality of measurement cycles, if the target object from the environmental sensor to be tested in the respective Measuring cycle was recorded. The measuring cycles are thus evaluated individually for the at least one sensor to be tested as to whether the target object was detected by the ambient sensor to be tested in the respective measuring cycle considered. A detection number for the sensor to be tested is then formed, wherein the detection number is the number of measurement cycles of the plurality of measurement cycles in which the target object was detected by the environmental sensor to be tested. A function measure for the environment sensor to be tested is determined from the detection number. In this way, an absolute, that is independent of the other environmental sensors Funktionsmaß for the at least one environmental sensor to be tested - in particular independent of its calibration or decalibration compared to the other environmental sensors - are obtained, so directly and directly the functionality of the environmental sensor to be tested can be judged. Essentially, it is tested with which relative frequency, in the measuring cycles performed, the environmental sensor to be tested was able to detect an object existing in the real world. The fact that the target object actually exists in the real world is ensured by identifying the same target object, in particular for all environmental sensors, in each measurement cycle of the plurality of measurement cycles. The further sensors of the plurality of environmental sensors are thus not used in the context of the method to determine a calibration or decalibration of the environment sensor to be tested, but rather to ensure that the target object, whose detection is evaluated by the sensor to be tested, actually in the real world is available. With the aid of the method, it is thus possible to determine an absolute functional dimension for the at least one environmental sensor to be tested even during a real, regular driving operation of a vehicle without prior knowledge of the real vehicle environment.

The identification of the target object takes place in particular by means of a so-called sensor fusion, in which measurements of various redundant environmental sensors are combined, so that a comprehensible overall picture of a current driving situation is obtained. The advantage of this is that the target object is still displayed by the consideration of a plurality of environmental sensors even if one of the environmental sensors involved can not detect this in the real world.

In the context of the method, a plurality of environmental sensors which are at least partially different from each other with respect to their measuring principle are preferably read out. In this way, the safety of the method can be increased because different environmental conditions affect the different measuring principles of the environmental sensors in different ways, and therefore not all environmental sensors are at the same time adversely affected by certain actual environmental conditions in their detection.

At least partially mutually redundant environmental sensors in the plurality of environmental sensors are preferably used, which means in particular that the mutually redundant environmental sensors - preferably by means of different or even by means of at least partially the same measuring principles - are able to an identical target object, such as a preceding vehicle or like to capture. Particularly preferably, only those environmental sensors are included in the plurality of environmental sensors capable of detecting an identical target object. For example, it is possible to use exclusively environmental sensors set up for monitoring a front half-space of a vehicle if, for example, a preceding vehicle is used as the target object. Environmental sensors oriented on a rear half-space of the vehicle can then make no relevant contribution to the process. However, it is also possible to apply the method to a plurality of environmental sensors which together can detect another target object, for example a following vehicle, wherein the method can be carried out, for example, for a plurality of environmental sensors which monitor a rear half-space of the vehicle. It is obvious that the principle described here is arbitrarily expandable and applicable in various ways.

Preferably, the environmental sensors are each evaluated in a plurality of directly successive measuring cycles. This increases the safety and stability of the process.

In particular, the same number of environmental sensors is evaluated in each measurement cycle. In particular, the same environmental sensors are read out in each measurement cycle.

As a functional measure, the detection number itself can be used, so that it does not require any further investigation steps.

According to a preferred embodiment of the method, it is provided that a detection rate is determined as a functional measure, which is determined as a quotient of the detection number and a total number of measurement cycles. For the at least one environmental sensor to be tested, therefore, in particular a counting statistic is determined. In the mathematical sense, the detection rate corresponds to the conditional probability that the environmental sensor to be tested positively detects the target object on the condition that the target object actually exists in the real world. The detection rate is a particularly suitable variable for describing the functionality of an environmental sensor and - as already explained - a particularly critical variable with a view to the usability of the environment sensor to be tested for driver assistance systems.

According to one embodiment of the invention, it is provided that the functional dimension for a plurality of environmental sensors of the plurality of environmental sensors is determined. Particularly preferably, the functional dimension is determined for all environmental sensors of the plurality of environmental sensors. Since each environment sensor of the plurality of environmental sensors is read out anyway in each measurement cycle, it is readily possible to check for more than one of the environmental sensors, in particular for all environmental sensors, whether the target object was detected by these environmental sensors in the individual measurement cycles, in which case also for each of the environmental sensors to be tested, in particular for all environmental sensors, in each case a detection number in the manner described above can be determined, again from the detection number, a function measure for the various environmental sensors to be tested, preferably for all environmental sensors of the plurality of environmental sensors, are determined can. In particular, this requires no additional acquisition or read-out steps, but only a more extensive evaluation of the measurement data obtained in any case. The environmental sensors In this case, quasi-alternating, on the one hand, are used to ensure the actual existence of the target object, and, on the other hand, to determine the respectively assigned functional dimension.

According to one embodiment of the invention, it is provided that the method-preferably in the regular driving mode of the vehicle-is preferably carried out cyclically, continuously, at predetermined time intervals and / or randomly. In this way, the at least one environment sensor to be tested can be monitored during the driving operation of the vehicle. Particularly preferably, a temporal development of the functional dimension for the at least one environmental sensor to be tested is monitored. In this way, not only in particular a deterioration, a defect or a failure of the at least one environmental sensor to be tested can be determined, but also temporarily occurring disturbances, for example due to changing environmental conditions, can be detected for the at least one environmental sensor to be tested. Such temporal developments can then be reacted in a suitable manner.

According to one development of the invention, it is provided that only such measuring cycles of the plurality of measuring cycles are used to test the at least one environmental sensor to be tested, in which the target object has been detected by at least two mutually different environmental sensors of the plurality of environmental sensors. In this way, the probability can be increased that the target object actually exists in the real world, which in turn increases the security of the method. Particularly preferably, only those measurement cycles of the plurality of measurement cycles are included in the evaluation, in which the target object has been detected by at least two environmental sensors of the plurality of environmental sensors that are different from each other and from the environmental sensor to be tested. This further increases the safety of the process. The safety of the method can be additionally increased by further increasing the number of environmental sensors that must have detected the target object in the individual measurement cycles, wherein preferably only those measurement cycles can be used in which the target object of at least three or at least four various environmental sensors of the plurality of environmental sensors has been detected. Measurement cycles of the plurality of measurement cycles that do not meet at least one of the aforementioned criteria, however, are preferably not evaluated, that is, such measurement cycles are included neither in the detection count nor in the total number of measurement cycles, since it is assumed that the target object in these measurement cycles was not present or - for whatever reason - could not be recorded properly, so that such measurement cycles have no usable information content and should therefore be removed from the counting statistics.

According to one embodiment of the invention, it is provided that a warning-in particular to a driver of the vehicle-is output if the functional dimension for the at least one environment sensor to be tested exceeds a predetermined first limit value in the direction of a degraded function, for example below the limit value. Alternatively or additionally, the at least one environmental sensor to be tested is preferably removed from the evaluation for a driving function of the vehicle, in particular for a driver assistance system, if the functional dimension for the at least one environmental sensor to be tested exceeds a predetermined second limit in the direction of a degraded function, for example the Limit value falls below. Alternatively or additionally, a driving function of the vehicle, in particular a driving assistance system, for which the plurality of environmental sensors is used, terminated when the functional dimension for the at least one environment sensor to be tested exceeds a predetermined third limit in the direction of a degraded function, for example, falls below the limit. Thus, in particular, a reaction tailored to the specific functional measure for the at least one environmental sensor to be tested can take place, which can be simply outputting a warning - without further consequences - by removing the sensor, which at least temporarily appears to be at least temporarily unreliable, from the evaluation for a Driving function, to the termination of such a driving function can reach. If the functional measure - such as the detection rate - represents a measure in which higher values indicate greater reliability and / or functionality of the environmental sensor, the first limit value is preferably greater than the second limit value, the second limit value preferably being greater than the third limit value , in each case an underrun of the limit values is checked.

Thus, if the reliability and / or operability of the at least one environmental sensor to be tested is only slightly impaired, issuing a warning can be sufficient, while all other functions can be continued without restriction. On the other hand, if the reliability and / or functionality of the environmental sensor is further impaired, it is advantageously removed from the evaluation for a driving function, in particular a driver assistance system, in which case it may be assumed that other environmental sensors the removed environmental sensor can replace at least temporarily.

If, on the other hand, the reliability and / or functionality of the at least one environmental sensor are even more severely impaired, a driving function of the vehicle based on the at least one environmental sensor, in particular a driver assistance system, is terminated in an advantageous manner. This approach may be based, for example, on the consideration that the environment sensor could fail permanently, and not just temporarily, or that there is such a serious impairment that it could affect other environmental sensors as well so that the driving function of the vehicle does not continue with the necessary certainty can be. This could, for example, be due to a general electronic disturbance of the vehicle or to special environmental conditions, for example a bad weather situation.

However, such a multi-stage approach is not necessarily provided. It is also possible for particularly particularly critical driving functions, for example completely autonomous driving of the vehicle, to be ended immediately as soon as the functional dimension for at least one of the environmental sensors exceeds a predetermined limit value in the direction of a degraded function.

According to one embodiment of the invention, it is provided that a vehicle immediately ahead of the vehicle is used as the target object. Such an immediately preceding vehicle represents a particularly suitable target object since it can typically be detected constantly by a plurality of environmental sensors via a longer period of time, in particular on a time scale of the measuring cycles. Such an immediately preceding vehicle thus represents a stable target object in the sense of the method.

According to one embodiment of the invention, it is provided that the detection of the target object is subjected to a plausibility check. This additionally increases the safety of the process. In particular, the plausibility check may be carried out in each measurement cycle of the plurality of measurement cycles, or on a random basis in a subset of measurement cycles of the plurality of measurement cycles. In particular, when the plausibility check is carried out in each measurement cycle, preferably only those measurement cycles are included in the evaluation in which the target object could be successfully plausibilized. Such a plausibility check can be carried out, for example, by detecting at least one parameter selected from a group comprising a longitudinal speed, a transverse speed, a longitudinal acceleration, and a lateral acceleration for the target object to be detected, for example a preceding vehicle, by means of the plurality of environmental sensors. wherein the at least one parameter thus detected is compared with at least one predetermined plausibility limit value. Such a plausibility limit value is chosen in particular such that it can not be exceeded or fallen short of an object that actually exists in the real world under normal circumstances. If, therefore, such a plausibility limit value is exceeded and / or undershot, an error in the detection of the target object must be assumed.

According to one embodiment of the invention, it is finally provided that at least one environmental sensor of the plurality of environmental sensors, a sensor is selected which is selected from a group consisting of an optical camera, an ultrasonic sensor, a radar sensor, a laser scanner, and a LiDAR sensor , Such environmental sensors are highly suitable for detecting an external environment of a vehicle and in particular for monitoring the external environment even under different environmental conditions due to the various underlying measurement principles.

According to one embodiment of the invention, it is provided that the method is carried out in an offline mode, wherein the performance of the environmental sensors and detection algorithms is checked on the basis of large databases without prior labeling.

According to one embodiment of the invention, it is provided that the execution and evaluation of the plurality of measuring cycles is limited in time with a time-slice method. As a result, it may possibly be possible to react faster to a collapse in the functional dimension of an environmental sensor.

Finally, the object is also achieved by providing a vehicle which has a plurality of environmental sensors and a control unit, wherein the control unit is operatively connected to the plurality of environmental sensors and is set up to carry out a method according to one of the previously described embodiments. In connection with the vehicle, in particular, the advantages that have already been explained in connection with the method arise.

With the aid of the method, it is possible, in particular, to continuously monitor the functional dimensions and, in particular, the detection rates of the environmental sensors installed in the vehicle while the vehicle is traveling. If given default Limit values for the functional dimensions of the environmental sensors are violated, a corresponding system reaction can take place. This can include displaying a warning or switching off a driving function. In particular, the method can be used to detect temporary disturbances, for example system dampings due to fog and rain, contamination and / or blockage, in environmental sensors such as a camera, a LiDAR sensor, a radar sensor, or the like, in the environment of autonomous driving functions. Due to the implicit comparison with other environmental sensors in the vehicle via a sensor fusion, the detection of such a disturbance can sometimes take place more quickly than would be possible for the respective individual environmental sensors in the context of a self-diagnosis.

The invention will be explained in more detail below with reference to the drawing. The single figure shows a schematic representation of an embodiment of a method for determining detection properties of at least one environmental sensor.

The single figure shows a schematic representation of an embodiment of a method for determining detection properties of at least one environmental sensor in a vehicle 1 , where the vehicle 1 a plurality of environmental sensors 3 , here are three environmental sensors 3.1 . 3.2 . 3.3 which, by way of example here, all points to a front half space of the vehicle 1 are aligned. The environmental sensors 3.1 . 3.2 . 3.3 are at least partially different from each other in terms of their measurement principle. In particular, the environmental sensors can 3 be selected from a group consisting of an optical camera, an ultrasonic sensor, a radar sensor, a laser scanner and a LiDAR sensor.

The vehicle has a control unit 5 on that with the environmental sensors 3 is operatively connected and set up to read this, the control unit 5 is further set up to perform the method described in more detail below.

In the context of the embodiment of the method described here, the environmental sensors 3 read out in a plurality of preferably consecutive measuring cycles, wherein a same target object 7 is identified in each measurement cycle of the plurality of measurement cycles. As a target object 7 here is the vehicle 1 immediately preceding another vehicle 9 used.

For at least one ambient sensor to be tested 3 the environmental sensors 3 , preferably for all environmental sensors 3 , is checked for each measurement cycle of the plurality of measurement cycles, whether the target object 7 from the environmental sensor to be tested 3 was recorded in the respective measuring cycle. It then becomes a detection number for the environmental sensor to be tested 3 formed as the number of those measuring cycles of the plurality of measuring cycles in which the target object 7 through the environmental sensor to be tested 3 was recorded. Subsequently, a functional dimension for the environmental sensor to be tested 3 determined from the detection number.

As a function measure, a detection rate is preferably determined as a quotient of the detection number and a total number of the plurality of measuring cycles in one run of the method. In particular, there is therefore a counting statistic for the at least one environment sensor to be tested 3 created.

The method is preferably - in particular in the regular driving operation of the vehicle 1 - cyclically, continuously, at predetermined time intervals, and / or randomly performed, preferably a temporal evolution of the functional dimension for the at least one environmental sensor to be tested 3 is monitored. During each run of the method, a predetermined total number of measuring cycles is preferably run through, preferably an identical total number of measuring cycles during each run of the method. In the figure on the right is schematically such a run of the method for the environmental sensors 3.1 . 3.2 . 3.3 represented with a total of ten measuring cycles.

Preferably, only such measuring cycles of the plurality of measuring cycles are used for the evaluation, in which the target object 7 of at least two environmental sensors 3 the majority of environmental sensors 3 was recorded. Will - as shown here - the process only for three different environmental sensors 3.1 . 3.2 . 3.3 applied, close the at least two environmental sensors necessarily at least temporarily also the environmental sensor to be tested 3 one. Become more than three environmental sensors 3 The safety of the method can be increased by using only those measurement cycles of the plurality of measurement cycles for testing the at least one environmental sensor to be tested 3 used in which the target object 7 at least two environmental sensors different from each other and from the environmental sensor to be tested 3 the majority of environmental sensors 3 was recorded.

The detection of the target object is preferably subjected to a plausibility check. Such a plausibility check can take place in each measuring cycle of the plurality of measuring cycles or on a random basis. Advantageously, only such Measuring cycles used for the evaluation, in which the target object 7 could be made plausible.

As already indicated, the method is shown schematically in more detail in the figure on the right. Here are the results of measurements for each environmental sensor 3.1 . 3.2 . 3.3 shown in each of the ten measuring cycles performed, with a 1 each meaning that the target object 7 through the appropriate environmental sensor 3 was detected, where 0 means that the target object 7 in the appropriate measuring cycle by the respective environmental sensor 3 was not recorded.

For every environment sensor 3.1 . 3.2 . 3.3 Now a detection rate DR (3.1), DR (3.2), DR (3.3) is determined by the detection number is determined as the number of those measuring cycles of the plurality of measuring cycles in which the target object 7 through the environmental sensor to be tested 3 was detected, wherein the detection number for the first environmental sensor 3.1 eight, for the second environment sensor 3.2 ten and for the third environment sensor 3.3 seven.

As a functional measure, the detection rate DR (3.1), DR (3.2), DR (3.3) is now determined as the quotient of this detection number and the total number of the plurality of measurement cycles, in this case ten, which is the case for the first environmental sensor 3.1 a detection rate DR (3.1) of 0.8, for the second environment sensor 3.2 a detection rate DR (3.2) of FIG. 1, and for the third environment sensor 3.3 gives a detection rate DR (3.3) of 0.7.

At least one predetermined limit can be defined for the determined functional dimension, when it is exceeded in the direction of a deteriorated function of the environmental sensor 3 , here if it falls below, a certain measure takes place. In particular, it is possible for a warning to be output if the functional dimension falls below a predetermined first limit value. Alternatively or additionally, it is possible for the environmental sensor assigned to the functional dimension 3 from the evaluation for a driving function of the vehicle 1 is excluded if the functional dimension for the respective environmental sensor 3 falls below a predetermined second threshold. Alternatively or additionally, it is possible that a driving function of the vehicle for which the plurality of environmental sensors is used is terminated when the functional dimension for at least one environmental sensor 3 the majority of environmental sensors 3 falls below a third predetermined limit.

By means of the method proposed here and the vehicle 1 It is possible in regular driving an absolute functional measure for individual environmental sensors 3 to determine the operation of each environmental sensors 3 to monitor in time, and especially temporary disturbances of individual environmental sensors 3 capture.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 102015005961 A1 [0002]

Claims (10)

Method for determining detection properties of at least one environmental sensor ( 3 ) in a vehicle ( 1 ), comprising the following steps: - reading a plurality of environmental sensors ( 3 ) of a vehicle ( 1 ) each in a plurality of measuring cycles; Identifying a same target object ( 7 ) in each measurement cycle of the plurality of measurement cycles, wherein - for at least one environment sensor to be tested ( 3 ) of the plurality of environmental sensors ( 3 ) is checked for each measurement cycle of the plurality of measurement cycles, whether the target object ( 7 ) of the environmental sensor to be tested ( 3 ) was detected in the respective measurement cycle; Forming a detection number for the environmental sensor to be tested ( 3 ) as the number of those measuring cycles of the plurality of measuring cycles in which the target object ( 7 ) through the environmental sensor to be tested ( 3 ), and - determining a functional dimension for the environmental sensor to be tested ( 3 ) from the detection number. A method according to claim 1, characterized in that a detection rate as a ratio of the detection count and a total number of measurement cycles is determined as a function measure. Method according to one of the preceding claims, characterized in that the functional dimension for a plurality of environmental sensors ( 3 ) of the plurality of environmental sensors ( 3 ), preferably for all environmental sensors ( 3 ) of the plurality of environmental sensors ( 3 ), is determined. Method according to one of the preceding claims, characterized in that the method is cyclically, continuously, at predetermined time intervals, and / or randomly performed, preferably a temporal evolution of the functional dimension for the at least one environmental sensor to be tested ( 3 ) is monitored. Method according to one of the preceding claims, characterized in that only such measuring cycles of the plurality of measuring cycles for testing the at least one environmental sensor to be tested ( 3 ) in which the target object ( 7 ) of at least two environmental sensors ( 3 ) of the plurality of environmental sensors ( 3 ) was recorded. Method according to one of the preceding claims, characterized in that a) a warning is issued when the functional dimension for the at least one environment sensor to be tested ( 3 ) exceeds a predetermined first limit in the direction of a degraded function, and / or that b) the at least one environment sensor to be tested ( 3 ) from the evaluation for a driving function of the vehicle ( 1 ) is excluded if the functional dimension of the at least one environment sensor to be tested ( 3 ) exceeds a predetermined second limit in the direction of a degraded function, and / or that c) a driving function of the vehicle ( 1 ) for which the plurality of environmental sensors ( 3 ) is terminated when the functional dimension for the at least one environmental sensor to be tested ( 3 ) exceeds a predetermined third limit in the direction of degraded function. Method according to one of the preceding claims, characterized in that as a target object ( 7 ) to the vehicle ( 1 ) immediately preceding vehicle ( 9 ) is used. Method according to one of the preceding claims, characterized in that the detection of the target object ( 7 ) is subjected to a plausibility check. Method according to one of the preceding claims, characterized in that as at least one environmental sensor ( 3 ) of the plurality of environmental sensors ( 3 A sensor selected from a group consisting of an optical camera, an ultrasonic sensor, a radar sensor, a laser scanner, and a LiDAR sensor is used. Vehicle ( 1 ), with a plurality of environmental sensors ( 3 ), and with a control unit ( 5 ) associated with the plurality of environmental sensors ( 3 ) is operatively connected, wherein the control device ( 5 ) is arranged for carrying out a method according to one of claims 1 to 9.

Images