Method and apparatus for determining triggers for data recording

Abstract

The invention relates to a method for determining triggers (T) for the recording of data (L3, L4), comprising the following steps: - carrying out (S1) an actual movement (B1) of at least one vehicle (C1), - carrying out (S2) a simulated movement (B2) of at least one vehicle (C1), - detecting (S3) at least one deviation (D) between the actual movement (B1) and the simulated movement (B2), - Determining (S4) at least one anomaly (A) in the actual movement (U, F, V1, L1) and / or in the simulated movement (B2), - determining (S5) at least one trigger (T) as a function of the at least one anomaly (A) determined, and a device (100) for determining triggers (T) for the recording of data (L3, L4).

Description

[0001] The invention relates to a method and a device for determining triggers for the recording of data.

[0002] The present invention and the problem underlying it are explained with regard to use in automotive technology, but are of course applicable to any field of application, in particular to the field of aviation.

[0003] When partially or fully autonomous vehicles are being driven, unexpected events can occur, so-called delta scenarios, which are not covered by standard methods in vehicle development - such as in the so-called Operational Design Domain. Investigating these unexpected events is particularly interesting for the further development of the vehicle's autonomous functions or for preparing reports for authorities. However, it is not effective to permanently record all data from a vehicle in operation in order to investigate these unexpected events, as the amount of data generated is large and recording is therefore uneconomical. However, data recording can be initiated using so-called triggers, for example the recording of object data, CAN, Ethernet, LiDAR, radar, or camera data.When a trigger is triggered, the data can be recorded specifically to investigate the unexpected events in more detail.

[0004] US 2021 / 0 291 843 A1 relates to a method and apparatus for acquiring sample deviation data and an electronic device, which can be used in particular in the fields of artificial intelligence technology, automatic driving technology, intelligent transportation technology and deep learning technology.

[0005] The problem, however, is that suitable triggers are unknown or can change over time.

[0006] The technical problem is to create a method and a device that enable better data recording. In particular, the method and device improve the determination of triggers.

[0007] The solution to the technical problem is achieved by the subject matter having the features of the independent claims. Further advantageous embodiments of the invention are set forth in the subclaims.

[0008] A method for determining triggers for data recording is proposed, comprising the following steps: - Carrying out an actual movement of at least one vehicle, - Executing a simulated movement of at least one vehicle, - Detecting at least one deviation between the actual movement and the simulated movement, - Detecting at least one anomaly in the actual movement and / or in the simulated movement, - Determining at least one trigger depending on the at least one anomaly detected.

[0009] The trigger determined using the method can serve as an indicator of the imminent occurrence of an unexpected event and thus trigger targeted data recording. This targeted data collection can improve the further development of a vehicle's autonomous capabilities.

[0010] Further proposed is a device for determining triggers for data recording, wherein the device is configured to carry out a method according to an embodiment described in this disclosure. The device can, in particular, be configured to carry out one, several, or all steps of the method explained in this disclosure. The technical effects and advantages cited in this disclosure for the method naturally also extend to the device, and vice versa.

[0011] The vehicle can be a motor vehicle, e.g., a passenger car. The vehicle can also be an aircraft, a rail vehicle, or a marine vehicle. In particular, the vehicle is a drone. The device can be arranged at least partially within the at least one vehicle or encompass the at least one vehicle. The device can be arranged at least partially outside the vehicle, e.g., in the form of a stationary data server.

[0012] The execution of the actual movement and / or the execution of the simulated movement of the at least one vehicle can be carried out, for example, with the aid of at least one control device of the device. The at least one control device can be designed, for example, as a microcontroller or comprise such a microcontroller.

[0013] The actual movement of the at least one vehicle can be executed in a manual driving mode, in an at least partially autonomous driving mode, or in a fully autonomous driving mode. In the manual driving mode, the vehicle can be controlled by a driver. In the at least partially autonomous driving mode, the vehicle can be controlled at least partially by software and / or at least partially by a driver. In the fully autonomous driving mode, the vehicle can be controlled entirely by software. A fully autonomous driving mode can, for example, be a driving mode according to SAE Level 3 or higher.

[0014] The simulated movement of the at least one vehicle can be executed, for example, in an at least partially autonomous shadow mode and / or in a fully autonomous shadow mode. The simulated movement refers, in particular, to a movement that the at least one vehicle does not actually execute. The simulated movement is executed, in particular, simultaneously with the actual movement, e.g., while a human driver and / or software actually controls the vehicle.

[0015] Executing the actual movement and / or executing the simulated movement can comprise detecting and / or generating and / or implementing at least one parameter set. The at least one parameter set can, for example, have at least one vehicle parameter and / or at least one environmental parameter and / or at least one control parameter. The vehicle parameter can, for example, represent a speed of the vehicle and can be detected, for example, by means of a speed sensor. The environmental parameter can, for example, represent an intensity of rain in the environment of the vehicle and can be detected, for example, by means of a rain sensor. The at least one control parameter can, for example, represent a steering angle.

[0016] The control parameter of the actual movement can be referred to as the actual control parameter. The control parameter of the simulated movement can be referred to as the simulated control parameter. The control parameter can in particular be designed to actually and / or simulate the at least one vehicle and / or at least one actuator of the vehicle. The control parameter can be generated and / or detected by the at least one control device. Of course, the actual control parameter can also be generated by the driver, e.g., by a steering input via a steering wheel and / or by a steering angle sensor. The actual control parameter can, for example, be implemented by the vehicle and / or the at least one actuator. For example, based on at least one detected vehicle parameter and / or environmental parameter, at least one actual control parameter can be generated in order to, for example,to actually control the vehicle in response to the vehicle parameter and / or the environmental parameter. Furthermore, for example, based on at least one detected vehicle parameter and / or environmental parameter, a simulated control parameter can be generated in order to simulate the control of the vehicle, e.g., in response to the vehicle parameter and / or the environmental parameter. In particular, the actual control parameter can be converted by the vehicle into an actual steering movement. Furthermore, the simulated control parameter can be converted by the control device into a simulated steering movement.

[0017] The detected deviation may represent an unexpected event. For example, the deviation may be an unexpected actual and / or simulated steering movement of the vehicle as a result of a detected environmental parameter. The deviation may be detected between a parameter detected during the execution of the actual movement and a parameter detected during the execution of the simulated movement. The deviation may, for example, be detected between the at least one actual control parameter and the at least one simulated control parameter, e.g., between an actual steering angle and a simulated steering angle. The deviation may, in particular, be a critical difference between a parameter set of the actual movement and a parameter set of the simulated movement. The deviation may, for example,be detected if a (magnitude) difference between a value of a parameter of the actual movement and a value of a parameter of the simulated movement is greater than a predefined, critical difference. The detection of the at least one deviation can be carried out, for example, by means of a detection device of the device. The detection device can be designed, for example, as a microcontroller or comprise such a microcontroller.

[0018] The anomaly can occur, for example, during the execution of the actual movement and / or the simulated movement. In particular, parameters of a recorded parameter set can be identified as an anomaly. The anomaly can be, for example, a value of a parameter of the actual movement and / or the simulated movement. In particular, one or more values ​​of a parameter can be identified as an anomaly. By identifying the at least one anomaly, a connection can be established, for example, between the detected deviation and a relevant, in particular causally relevant, parameter. It can, for example, be predefined that one or more critical values ​​of a parameter are causally relevant for the detected deviation. For example, it can be predefined that rain complicates autonomous driving functions and can therefore, for example, lead to an unexpected steering movement. If, for example,If a difference between the actual and simulated steering angle is detected as a deviation, it can be determined whether, for example, in the 30 seconds prior to the detected deviation, the environmental parameter reached a predefined, critical value that is considered causally relevant to the deviation, such as a predefined critical intensity value of rain. The determination of at least one anomaly can be carried out, for example, with the help of at least one control device.

[0019] The at least one trigger can be determined, for example, with the aid of the at least one control device. The trigger can be determined, for example, as a threshold value. The threshold value can be, for example, a value of the at least one parameter that is determined as an anomaly. The trigger can, in particular, be designed such that the recording of data is triggered when the threshold value is exceeded or undershot. For example, the at least one trigger can be determined as a threshold value for the environmental parameter, e.g., a threshold value for the intensity of rain.

[0020] In one embodiment, the at least one deviation between at least one control parameter of the actual movement and at least one control parameter of the simulated movement is detected, wherein the at least one anomaly is determined as at least one critical value of at least one detected parameter of the actual movement and / or the simulated movement, wherein the at least one trigger is determined as the at least one critical value. This is particularly advantageous since unexpected events are generally associated with a deviation between actual and simulated control parameters, and an anomaly in the previously detected parameters can be the cause of the detected deviation. The trigger determined in this way can thus be used to ensure that data is only recorded when the critical value of the at least one detected parameter is detected again. The critical value can, for example,be an intensity value of the rain.

[0021] In one embodiment, the data to be recorded is determined as a function of the at least one trigger. In this way, the scope of the recording can be limited to the data relevant for investigating an unexpected event. The data to be recorded can, for example, comprise one or more specific parameters determined as a function of the specific trigger. Alternatively or cumulatively, the data to be recorded can comprise a recording duration determined as a function of the specific trigger, such as a recording of data over a period of 30 seconds. For example, depending on the specific trigger, an actual and / or a simulated control parameter can be determined as data to be recorded and in this way recorded specifically when the specific trigger is triggered, e.g. because it starts to rain.

[0022] In one embodiment, the specific at least one trigger is validated by checking whether the data recorded using the at least one trigger exhibits at least one deviation. In this way, the determination of a suitable trigger can be further improved. In particular, an unsuitable trigger can be rejected by the validation. Further in particular, the specific trigger can be considered successfully validated if the recorded data exhibit the detected deviation. The validation can be carried out, for example, using the at least one control device. For example, the validation can be carried out by using the specific trigger in the actual movement and / or the simulated movement in another vehicle. The data recorded using the trigger can, for example,exhibit a deviation between a parameter of the actual movement and a parameter of the simulated movement that corresponds to the detected deviation. This indicates the suitability of the specific trigger as an indicator of unexpected events.

[0023] In one embodiment, the specific at least one trigger is validated using data recorded in a plurality of vehicles. This allows for more accurate and faster validation, as the trigger can be validated using a more comprehensive database. The specific trigger can be considered successfully validated, for example, if a predominant number, e.g., a majority, of the recorded data exhibits the detected deviation. The plurality of vehicles can, for example, be part of a vehicle fleet used to validate the specific trigger.

[0024] In one embodiment, the simulated movement is carried out with the aid of software, wherein the software is a version of an at least partially autonomous or fully autonomous actual driving mode of the vehicle. In this way, for example, a reaction of the software to unexpected events can be tested before an actual movement is carried out using the software. The software can, for example, comprise newly developed functions of the actual driving mode. For example, the software to be tested can be executed simultaneously with an at least partially and / or fully autonomous driving mode of the actual movement, wherein the driving mode in the actual movement is carried out, for example, in an older version of the software. In this way, for example, triggers can be determined to verify the new software.

[0025] In one embodiment, the at least one anomaly is determined using a rate of change of the actual movement and / or the simulated movement. In this way, a relationship between the detected deviation and a temporal and / or spatial change in the actual and / or simulated movement can be examined more precisely. The rate of change can in particular be a temporal and / or spatial derivative of a parameter detected during the actual and / or simulated movement. For example, the rate of change can be a temporal derivative of the steering angle, e.g., a steering angular velocity. The rate of change can comprise one or more values ​​of a temporal and / or spatial change in a parameter. The anomaly can in particular be determined as a critical difference between the rate of change of the actual movement and the rate of change of the simulated movement. The anomaly can, for example,be determined if a (magnitude) difference between a value of the rate of change of the actual movement and a value of the rate of change of the simulated movement is greater than a predefined, critical difference.

[0026] In one embodiment, the actual movement and / or the simulated movement of the at least one vehicle is temporarily stored as a ring buffer. The ring buffer has the effect that, for example, the oldest stored data is overwritten with more recent data of the actual and / or simulated movement. In this way, the data temporarily stored for determining the at least one trigger can be reduced to a predefined amount. For this purpose, the device can, for example, have at least one buffer. In particular, a temporal profile of the actual movement and / or the simulated movement can be temporarily stored. The ring buffer can, for example, comprise the temporal profile of the actual and / or simulated movement over a period of 30 seconds. In this case, for example, the data of the first second of the temporal profile can be continuously overwritten by data of the thirtieth second of the temporal profile.The cached data can, in particular, comprise at least one parameter set of the actual and / or simulated movement. Of course, the cached actual and / or simulated movement can also be stored permanently.

[0027] In one embodiment, the actual movement and / or the simulated movement of the at least one vehicle is permanently stored when the at least one deviation is detected. In this way, the amount of data to be permanently stored can be limited. In particular, the actual movement and / or the simulated movement temporarily stored as a ring buffer can be permanently stored when the at least one deviation is detected within the ring buffer. In this way, the amount of data to be permanently stored can be further reduced. For the purpose of permanent storage, the device can, for example, have at least one memory.

[0028] The invention is explained in more detail using exemplary embodiments. The figures show: Fig. 1 a schematic flow diagram of an embodiment of a method, Fig. 2-A is a schematic diagram of a time course of parameters of an actual and a simulated movement, Fig. 2-B is a schematic diagram of a time course of a rate of change of parameters of an actual and a simulated movement and Fig. 3 a schematic representation of a validation of a specific trigger for recording data.

[0029] In the following, the same reference symbols refer to elements with the same technical features.

[0030] Fig. Figure 1 shows a schematic flow diagram of an embodiment of a method for determining triggers T for the recording of data L3, L4. The method comprises several steps S1, ..., S6, which are explained below.

[0031] In a step S1, an actual movement B1 of a vehicle (not shown) is executed. Executing the actual movement B1 involves recording a parameter set U, F, L1. The recorded parameter set U, F, L1 has an environmental parameter U in the form of a rain level, a vehicle parameter F in the form of a speed, and a control parameter L1 in the form of the actual steering angle. The control parameter L1 is generated, for example, in response to the recorded parameters U, F by software V1. The software V1 is, for example, a version of a fully autonomous driving mode of the actual movement B1. The control parameter L1 can be converted, for example, by means of an actuator (not shown), into an actual steering movement K of the vehicle.

[0032] In a step S2, a simulated movement B2 of the vehicle (not shown) is executed. In particular, step S2 is executed simultaneously with step S1. Executing the simulated movement B2 also includes recording a parameter set U, F, L2. The recorded parameter set U, F, L2 also has the environmental parameter U in the form of the rain level, the vehicle parameter F in the form of the speed, and a control parameter L2 in the form of a simulated steering angle. The simulated control parameter L2 is generated by software V2 in response to the recorded parameters U, F. The software V2 is, for example, a version of the fully autonomous driving mode of the actual movement B1 to be tested, which, however, initially only generates a simulated control parameter L2, which is not converted into an actual steering movement K.

[0033] In a step S3, a deviation D between the actual movement B1 and the simulated movement B2 is detected by comparing a time course of the actual control parameter L1 with a time course of the simulated control parameter L2.

[0034] Fig. 2-A shows a schematic diagram of a temporal progression of the actual control parameter L1 and the simulated control parameter L2. The temporal progression of the control parameters L1, L2 is qualitatively represented in a coordinate system, with time t on the abscissa and values ​​of the control parameters L1, L2 recorded as the steering angle L on the ordinate. The temporal progression of the control parameters L1, L2 is buffered as a ring buffer R in a buffer (not shown). The temporal progression of the actual control parameter L1 is linear, while the temporal progression of the simulated control parameter L2 is quadratic. This results in an ever-increasing difference between the values ​​of the control parameters L1, L2, whereby a critical difference can be detected as deviation D. The deviation D is, for example,detected when a difference between the actual control parameter L1 and the simulated control parameter L2 is greater than a predefined critical difference.

[0035] In a step S4, an anomaly A in the actual movement B1 is determined (cf. Fig. 1). For this purpose, for example, the parameters U, F recorded during the actual movement B1 are checked for anomalies A. In particular, it is checked whether a value of the parameter set U, F is relevantly causal for the detected deviation D. For example, it can be predefined that a critical value of the environmental parameter U is considered causally relevant for the detected deviation D. For example, rain can complicate autonomous driving functions and therefore lead to, for example, an unexpected steering movement. Therefore, for example, a critical value of the environmental parameter U can be identified as anomaly A.

[0036] Fig. Figure 2-B shows a schematic diagram of a time course of the change rates Ĺ1, Ĺ2 of the recorded control parameters L1, L2. The anomaly A can also be determined using the change rates Ĺ1, Ĺ2 of the actual movement B1 and the simulated movement B2.

[0037] The temporal progression of the change rates Ĺ1, Ĺ2 of the control parameters L1, L2 is qualitatively represented in a coordinate system, with time t on the abscissa and values ​​of the steering angle velocity Ĺ on the ordinate. The temporal progression of the change rates Ĺ1, Ĺ2 corresponds to the Fig. 2-A shows the temporal progression of the control parameters L1, L2. The rate of change Ĺ1 of the actual control parameter L1 is therefore constant, while the rate of change Ĺ2 of the simulated control parameter L2 is linear. This results in a difference between the values ​​of the rate of change Ĺ1, Ĺ2 right at the beginning of the temporal progression, which can be identified as anomaly A. Anomaly A is identified, for example, when a difference between the rate of change Ĺ1 of the actual control parameter L1 and the rate of change Ĺ2 of the simulated control parameter L2 is greater than a predefined, critical difference.

[0038] In a step S5, a trigger T is determined depending on the detected anomaly A (cf. Fig. 1). The trigger T can be determined in such a way that the trigger T is triggered, for example, when the value of the environmental parameter U determined as anomaly A is detected again. Alternatively, the trigger T can be determined in such a way that the trigger T is triggered, for example, when the difference determined as anomaly A between the rate of change Ĺ1 of the actual control parameter L1 and the rate of change Ĺ2 of the simulated control parameter L2 again shows the critical difference. In other words: the trigger T can be designed as a threshold value, wherein the recording of data L3, L4 is triggered when the threshold value is exceeded or undershot. With the help of the determined trigger T, it is achieved in particular that the recording of data L3, L4 is only triggered when the circumstance determined as anomaly A occurs again.

[0039] The control parameters L3, L4 can be assigned to the specific trigger T as data to be recorded. In this way, the data L3, L4 to be recorded when the trigger T is triggered can be specified as a function of the at least one trigger T. This reduces the amount of data to be recorded when the trigger T is triggered. The control parameter L3 can, for example, correspond to an actual steering angle that is recorded after the trigger T is triggered during the actual movement B1 of a vehicle. The control parameter L4 can, for example, correspond to a simulated steering angle that is recorded after the trigger T is triggered during the simulated movement B2 of a vehicle. The control parameters L3, L4 can, in particular, be specified as data to be recorded since the deviation D has previously been detected in the control parameters L1, L2.

[0040] In step S6, the specific trigger T is used to record data L3, L4 in a vehicle (not shown). Trigger T can be triggered, for example, when it starts to rain. This means that only the environmental parameter U, recorded as the rain level, needs to be monitored before trigger T is triggered. The vehicle parameter F or other parameters do not need to be recorded or monitored. Only when trigger T is triggered are the control parameters L3, L4 additionally recorded as data.

[0041] Fig. 3. shows a schematic representation of a validation of a specific trigger T for the recording of data L3, L4. The specific trigger T can be validated, for example, by checking whether the data L3, L4 recorded using the trigger T exhibit a deviation D or not.

[0042] In the Fig. 3, the trigger T is used in a vehicle C1 designed as a passenger car by means of the Fig. 1. The vehicle C1 comprises a device 100 configured to carry out a method according to steps S1 to S5.

[0043] Following the determination of the trigger T, the trigger T is used to record data L3, L4 in a plurality of vehicles C2, C3, C4. Each of the vehicles C2, C3, C4 also comprises a device 100. The device 100 is further configured for the deployment S6 of the trigger T as well as for the validation S7 of the trigger T. The use of the plurality of vehicles C2, C3, C4 serves to enable faster and more accurate validation S7 of the trigger T, since a larger database can be accessed.

[0044] In the Fig.In the scenario shown in Figure 3, the specified control parameters L3, L4 are recorded using vehicles C2, C3. These exhibit the deviation D as expected. Only with vehicle C4 could no deviation D be detected between the control parameters L3, L4 after trigger T was triggered. Instead, the control parameters L3, L4 in vehicle C4 match during recording. This is illustrated by graph G, which shows two parallel lines representing the control parameters L3, L4 recorded in vehicle C4. Nevertheless, validation S7 of trigger T is considered successful because a majority of the recorded control parameters L3, L4 also exhibit the previously detected deviation D. The successfully validated trigger T can therefore be released for use in further development. List of reference symbols A anomaly B1 actual movement B2 simulated movement C1, ..., C4 vehicle D Deviation F Vehicle parameters K Steering movement L Values ​​of the control parameters L1 Control parameters of the actual movement L2 Control parameters of the simulated movement L3 control parameter of the actual movement to be recorded L4 control parameter of the simulated movement to be recorded Ĺ Values ​​of the rates of change Ĺ1 Rate of change of the actual control parameter Ĺ2 Rate of change of the simulated control parameter R ring buffer S1, ... S7 step T Trigger t time U Environmental parameters V1 software version of the actual movement V2 software version of the simulated movement QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This 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

[0000] US 2021 / 0 291 843 A1

[0004]

Claims

[1] Method for determining triggers (T) for the recording of data (L3, L4), comprising the following steps: - carrying out (S1) an actual movement (B1) of at least one vehicle (C1), - carrying out (S2) a simulated movement (B2) of the at least one vehicle (C1), - detecting (S3) at least one deviation (D) between the actual movement (B1) and the simulated movement (B2), - Determining (S4) at least one anomaly (A) in the actual movement (B1) and / or in the simulated movement (B2), - Determining (S5) at least one trigger (T) depending on the at least one anomaly (A) determined. [2] Method according to claim 1, characterized bythat the at least one deviation (D) between at least one control parameter (L1) of the actual movement (B1) and at least one control parameter (L2) of the simulated movement (B2) is detected, wherein the at least one anomaly (A) is determined as at least one critical value of at least one detected parameter (U, F, L1, L2) of the actual movement (B1) and / or the simulated movement (B2), wherein the at least one trigger (T) is determined as the at least one critical value. [3] Method according to claim 1 or 2, characterized by that the data to be recorded (L3, L4) are determined as a function of at least one trigger (T). [4] Method according to one of the preceding claims, characterized by that the determined at least one trigger (T) is validated by checking whether the data (L3, L4) recorded with the aid of the at least one trigger (T) have at least one deviation (D). [5] Method according to one of the preceding claims, characterized by that the determined at least one trigger (T) is validated using the data (L3, L4) recorded in a plurality of vehicles (C2, C3, C4). [6] Method according to one of the preceding claims, characterized by that the execution (S2) of the simulated movement (B2) is carried out with the aid of software, wherein the software (V2) is a version of an at least partially autonomous or fully autonomous actual driving mode of the vehicle (C1). [7] Method according to one of the preceding claims, characterized by that the at least one anomaly (A) is determined with the aid of a rate of change (Ĺ1) of the actual movement (B1) and / or a rate of change (Ĺ1) of the simulated movement (B2). [8] Method according to one of the preceding claims, characterized bythat the actual movement (B1) and / or the simulated movement (B2) of the at least one vehicle (C1) is temporarily stored as a ring buffer (R). [9] Method according to one of the preceding claims, characterized by that the actual movement (B1) and / or the simulated movement (B2) of the at least one vehicle (C1) is permanently stored when the at least one deviation (D) is detected. [10] Device (100) for determining triggers (T) for the recording of data (L3, L4), characterized by that the device (100) is designed to carry out a method according to one of claims 1 to 9.

Images