System and procedure for verifying the plausibility of a vehicle environment hypothesis for a motor vehicle

Abstract

Method for verifying the plausibility of a vehicle environment hypothesis for a motor vehicle (10), comprising the steps: - Providing (S10) the vehicle environment hypothesis, which is determined from given data and provides an assumption about an environment property of the motor vehicle (10); - Detection (S12) of environmental features in the vicinity of the motor vehicle (10) by means of a sensor device (16), wherein the sensor device (16) comprises a plurality of different sensors (18, 20, 22); - Evaluating (S14) the recorded environmental features with regard to the plausibility of the vehicle environment hypothesis by checking the respective recorded environmental feature for the presence of one or more predetermined plausibility criteria, wherein the plausibility criteria depend on a type of environmental feature and a sensor type with which the environmental feature was recorded; - Plausibility check (S16) of the vehicle environment hypothesis based on the evaluated environmental characteristics; whereby the plausibility criteria are weighted differently depending on the type of environmental characteristic and / or the sensor type during the evaluation.

Description

[0001] The invention relates to a method for verifying the plausibility of a vehicle environment hypothesis for a motor vehicle, a system for verifying the plausibility of the vehicle environment hypothesis and a motor vehicle with such a system.

[0002] In automated vehicles from SAE Level 3 onwards, the driver temporarily transfers responsibility to the vehicle and thus liability to a vehicle manufacturer. This results in high safety requirements for the entire system due to the Automotive Safety Integrity Level (ASIL) classification. Current driver assistance systems according to SAE Level 2 generally do not meet these high requirements. New driver assistance systems usually use predefined data to increase anticipation of information about the vehicle's surroundings or environmental properties, since the installed environmental sensors have a limited field of view. For example, data from a digital map can be transmitted from a backend to the vehicle, which can then be used to control the vehicle.

[0003] The disadvantage is that, due to the unguaranteed timeliness of the data and potential transmission errors, the safety requirements resulting from an ASIL classification of the automated driving system are almost impossible to meet. This means that this data, or its use in safety-critical driving functions, could lead to a violation of a safety objective. For example, if lane hypotheses stored in an outdated digital map were directly used to guide the vehicle laterally, this could lead to a dangerous situation and thus a violation of a safety objective.

[0004] DE 10 2009 060 600 A1 discloses a method for assigning lanes of a roadway to a motor vehicle, wherein the motor vehicle has a navigation system including a digital map for determining the current position and an image processing device for determining environmental data. The method comprises determining the position of the motor vehicle, determining the number of lanes of the roadway and extracting lane marking features from the position of the motor vehicle and the digital map, extracting lane markings and their marking features from the environmental data of the image processing device, and comparing the marking features extracted from the digital map and the marking features determined by image processing with a classifier to determine the lane of the motor vehicle with respect to the current position.

[0005] DE 10 2015 214 338 A1 discloses a method for determining arrangement information for a vehicle. The arrangement information includes a position of the vehicle and an orientation of the vehicle with respect to a stationary coordinate system. In the method, a local lane arrangement in the environment of the vehicle is detected using vehicle sensors. Depending on the local lane arrangement and previously determined arrangement information for the vehicle, a first lane arrangement is determined with respect to the stationary coordinate system. Depending on predefined map material, a second lane arrangement is determined with respect to the stationary coordinate system.Depending on a geometric similarity between the first lane arrangement and the second lane arrangement, similarity information is determined and, depending on similarity information, previously determined arrangement information for the vehicle and a previously determined lane arrangement hypothesis are updated.

[0006] DE 10 2016 215 314 A1 discloses a driver assistance system, a means of transport, and a method for predicting a traffic situation. The method comprises the steps of determining a current image of a vehicle's surroundings, determining a future state of an environmental object contained in the image using a parametric model, determining a correction value using a machine-based learning method, determining an uncertainty value using the machine-based learning method, and determining a tolerance band for future states using the correction value and the uncertainty value to predict the vehicle's surroundings.

[0007] DE 10 2008 053 ​​531 A1 discloses an evaluation module for evaluating data from a digital map for a vehicle. The evaluation is based on the map's own quality information and measurement data from the vehicle's environmental sensors. The connected driver assistance systems and safety systems of the vehicle use the digital map data based on the evaluation.

[0008] From DE 10 2019 002 487 A1 a method for checking at least one environment detection sensor of a vehicle is known, wherein the vehicle is located in a digital map, features of stored stationary objects in an environment of the vehicle are identified in the digital map, which are expected to be detected by the environment detection sensor, the environment of the vehicle is detected with the environment detection sensor and a degradation of the environment detection sensor is concluded if the features expected to be detected are not detected by the environment detection sensor or if features actually detected by the environment detection sensor deviate significantly from the features expected to be detected.

[0009] The invention is based on the object of evaluating a vehicle environment hypothesis of a motor vehicle.

[0010] This object is achieved by the independent patent claims. Advantageous developments of the invention are disclosed in the dependent patent claims, the following description, and the figures.

[0011] The invention is based on the idea that a vehicle environment hypothesis, for example, a lane hypothesis, is evaluated using environmental features detected by multiple sensors to determine whether predetermined data from which the vehicle environment hypothesis is obtained meet safety requirements. Thus, if the vehicle environment hypothesis is confirmed, the data can be used for vehicle control, in particular for autonomous motor vehicle control.

[0012] The invention provides a method for verifying the plausibility of a vehicle environment hypothesis for a motor vehicle. The method, which can be carried out, for example, by a system with a sensor device and a computing device, comprises the following steps: providing the vehicle environment hypothesis, which is determined from predetermined data and provides an assumption about an environmental property of the motor vehicle; detecting environmental features in an environment of the motor vehicle by means of a sensor device, wherein the sensor device has a plurality of different sensors; evaluating the detected environmental features with regard to a plausibility of the vehicle environment hypothesis by checking the respectively detected environmental feature for the presence of one or more predetermined plausibility criteria, wherein the plausibility criteria depend on a type of environmental feature and a sensor type,with which the environmental feature was recorded, and a plausibility check of the vehicle environment hypothesis based on the evaluated environmental features, whereby the plausibility criteria are weighted differently depending on the type of environmental feature and / or the sensor type in the evaluation.

[0013] In other words, a vehicle environment hypothesis can first be determined from given data, wherein the vehicle environment hypothesis represents an assumption about an environmental property of the motor vehicle. For example, the given data can originate from a high-resolution digital map, and the vehicle environment hypothesis determined therefrom can include, for example, an assumption about the number of existing lanes and / or the precise geometric layout of the lanes and / or a lane width at a current or future position. This means that in this example, the environmental property of the motor vehicle is the number of lanes, which can be provided from the given data.

[0014] In order to check whether this vehicle environment hypothesis corresponds to reality or whether, for example, the specified data is out of date, a sensor device can next record the surroundings of the motor vehicle, wherein in the recording of the surroundings, environmental features are determined which can be used to confirm or refute the put forward vehicle environment hypothesis. For this purpose, the sensor device has at least two sensors, wherein one sensor type differs from the at least two sensors. Preferably, at least three different sensor types can be provided, in particular a camera, a radar, a lidar. Alternatively or additionally, the sensor device can also have an ultrasonic sensor. The sensors of the sensor device can detect the same or different environmental features, which depends in particular on the type of environmental feature and on the field of view of the respective sensor.Environmental features can be, for example, traffic signs, lane markings, edge buildings, in particular crash barriers, beacons, walls and / or curbs, object tracks / column tracks and / or objects, in particular other vehicles.

[0015] After determining the environmental features, each environmental feature can be checked to determine whether it confirms the vehicle environment hypothesis or not. To do this, the respective environmental features can be checked to determine whether one or more predefined plausibility criteria are met for the respective environmental feature. The plausibility criteria can be specified differently for each environmental feature; for example, the plausibility criteria can be predetermined from a simulation. Alternatively or additionally, the plausibility criteria for the respective environmental feature can also depend on the sensor type used to measure the environmental feature. For example, different plausibility criteria can be checked for an environmental feature measured by a camera than for a measurement using a lidar.It can also preferably be determined whether the respective environmental feature could be measured by the respective sensor at all or whether it was outside of the field of view.

[0016] Furthermore, the plausibility criteria are weighted differently during the evaluation depending on the type of environmental feature and / or the sensor type, so that if predetermined plausibility criteria are met, the vehicle environment hypothesis is more likely to be correct. Preferably, a confidence value can be determined for each met plausibility criterion, which can in particular be weighted differently in order to weight the respective environmental feature with respect to the vehicle environment hypothesis. The confidence values ​​can indicate the probability of the vehicle environment hypothesis being correct. Preferably, the plausibility criteria can differ depending on the vehicle environment hypothesis established.For example, in a lane hypothesis, lane markings, object lanes, and / or convoy lanes can be checked. In a speed limit hypothesis, for example, traffic signs and / or objects, in particular a measured speed of the objects, and / or lane markings can be checked. In particular, the color of the lane markings can indicate a reduced speed limit. For example, a yellow lane marking can indicate a construction site and thus a reduced speed.

[0017] The evaluation of the environmental features can be carried out by a computing device which, by means of artificial intelligence which has been trained in particular by machines, checks the presence of the respective plausibility criteria in the recordings of the environmental features.

[0018] Finally, the vehicle environment hypothesis can be validated based on the evaluated environmental characteristics. This means that the vehicle environment hypothesis can be confirmed or rejected. If confirmed, the specified data from which the vehicle environment hypothesis was provided can be marked as valid or current. The specified data can then be used, for example, for vehicle control, preferably for autonomous vehicle control. Alternatively, the vehicle environment hypothesis can also be rejected, whereby the specified data can be marked as not current and thus unsuitable for vehicle control.

[0019] For plausibility checks, the evaluations of the environmental features can be stored, for example, in an evaluation table, which can specify which plausibility criterion is confirmed or rejected by which sensor. The evaluation table can preferably also specify whether an environmental feature could not be measured by a sensor at all because the environmental feature that could confirm the driving environment hypothesis is outside the field of view. In other words, the vehicle environment hypothesis to be tested, or a part of the hypothesis, for example, a lane, can be outside the field of view for an environmental feature. Thus, the environmental feature cannot be used for plausibility checks in this area because it cannot be measured.Preferably, a different field of view can be defined for each environmental feature, since these can be detected differently by the sensors. In particular, the respective confidence values ​​determined can be stored in the evaluation table, which can be used to perform a statistical evaluation regarding the validity of the vehicle environment hypothesis. Statistical methods can be used for this purpose that calculate a statement about the credibility of the vehicle environment hypothesis. In particular, the vehicle environment hypothesis can be verified using a Dempster-Shafer theory.

[0020] The specified data from which the vehicle environment hypothesis is derived can originate from various data sources, some of which do not meet the safety requirements of an ASIL classification, such as data from a digital map, swarm data, and / or data from artificial intelligence. Using the method and a confirmed vehicle environment hypothesis, the data can be qualified in such a way that they meet the ASIL classification. This offers the advantage that the specified data can be used for vehicle control without jeopardizing the vehicle's safety objectives. In particular, the sensor device can be used to check the plausibility of the data / vehicle environment hypothesis during runtime in the motor vehicle. The evaluation is not carried out using a single sensor principle and a single environmental feature, but rather based on a combination of several sensor types and environmental features.This provides redundancy in the assessment, increasing its reliability. Preferably, the data can then be used to control the vehicle, enabling safety-relevant driving functions to be performed, increasing user convenience and thus acceptance.

[0021] The invention also includes embodiments which provide additional advantages.

[0022] One embodiment provides that the predefined data from which the vehicle environment hypothesis is determined is a digital map and / or swarm data and / or data from an artificial intelligence, in particular a neural network. In other words, the predefined data can originate from a non-ASIL-compliant source and be upgraded to ASIL-compliant data using the method. A digital map can, for example, include roadway information, such as a number of lanes, a road layout, and / or speed information for route sections. Swarm data can be aggregated from multiple vehicles in a vehicle fleet, from which environmental properties of the motor vehicle can be determined. An artificial intelligence, in particular a neural network, can be trained using machine learning, for example to perform vehicle functions.This embodiment has the advantage that vehicle environment hypotheses can be verified from multiple sources.

[0023] A further embodiment provides that when checking the respective detected environmental feature for the presence of the at least one plausibility criterion, a current field of view of the respective sensors is taken into account. In other words, it can be taken into account whether a respective sensor can perceive an environmental feature or not. For example, an environmental feature can only be detected by a single sensor due to a field of view of the sensors, whereby other sensors cannot perceive the environmental feature at all due to the field of view. The fields of view of the respective sensors can preferably be predetermined and / or estimated. The field of view of the sensor device can be divided into areas in which the contributing sensors differ.This can be taken into account when checking the plausibility criteria and evaluating the vehicle environment hypothesis, for example, if a sensor or sensor type fails to confirm an environmental feature, the environmental feature was actually not detected by the sensor, or the environmental feature cannot be confirmed because it lies outside the sensor's current field of view. This, in turn, can be taken into account in the resulting plausibility check of the vehicle environment hypothesis. This embodiment offers the advantage of improving the evaluation of the vehicle environment hypothesis.

[0024] Preferably, an installation position of the respective sensor and / or a specification of the respective sensor, in particular an aperture angle and / or a measuring range, are taken into account for the current field of view. Thus, an installation position on the motor vehicle can specify a direction that the sensor can detect. Alternatively or additionally, specifications of the respective sensor, for example, a resolution, an aperture angle or measuring range, and / or a measuring range for the respective sensor, can also be taken into account.

[0025] In a further advantageous embodiment, it is provided that current restrictions of the fields of view of respective sensors are determined by means of sensor data fusion, wherein the current restrictions of the fields of view are taken into account when checking the respectively detected environmental feature for the presence of the at least one plausibility criterion. In other words, the field of view of the respective sensor can also depend on whether the sensor can detect an environmental feature unhindered or whether an object is located in the field of view of the sensor that obscures the environmental feature. In this way, the fields of view of all sensors can be determined from sensor data fusion, in particular overlapping areas, wherein, for example, an environmental feature can be measured by one sensor but not by another sensor because an object, in particular a vehicle, is located in the field of view of the sensor.Thus, in particular, only the plausibility criteria for the sensors that were able to detect the environmental feature can be checked, and for the sensor with the obscured field of view, for example, it can be stated that the environmental feature could not be detected by this sensor and thus the measured values ​​of this sensor are not taken into account when evaluating the environmental feature with regard to the plausibility of the vehicle environment hypothesis.

[0026] A further embodiment provides that the vehicle environment hypothesis is assumed to be plausible if at least two of the detected environmental features confirm the vehicle environment hypothesis with a predetermined confidence. In other words, a single environmental feature is not sufficient to confirm the vehicle environment hypothesis, even if it confirms the vehicle environment hypothesis. In particular, it can be specified that the vehicle environment hypothesis is confirmed by at least two environmental features for which the plausibility criteria apply, wherein the plausibility criteria are preferably confirmed by two different sensors and / or sensor types. This embodiment can improve the plausibility of the vehicle environment hypothesis because the probability of a false-positive assessment can be minimized.This also increases safety when using the specified data for vehicle control.

[0027] A further embodiment provides that the at least one plausibility criterion to be checked for the respective environmental feature is predetermined by means of a simulation. In other words, the simulation predetermines which plausibility criteria are checked for a respective environmental feature, preferably also taking into account the sensor type with which the environmental feature was recorded. Ground truth data of the vehicle environment hypothesis can preferably be used for the simulation; these serve as the input hypothesis and can be modified by means of error injection or fault injection to determine the effects of the sensors on the assessment. For example, ground truth data of the vehicle environment hypothesis can be formed from highly accurate digital maps or labeled data in order to have a reference.Additionally, the environmental characteristics to be evaluated can be falsified using error injection and then evaluated for safety criticality using ground-truth data. By reviewing the evaluation states for safety-critical initial hypotheses, the criteria that lead to safe initial hypotheses can be determined. These parameters or criteria can be determined statistically through various error cases and simulations. This has the advantage of allowing the determination of evaluation criteria or plausibility criteria that must be met for the respective environmental characteristics to be considered sufficiently safe. This implementation can increase reliability and confidence for various states.

[0028] Preferably, the sensor device comprises a vehicle camera, a radar, and a lidar. This means that the sensor device comprises at least the three sensors mentioned above to detect the environmental features, with each detected environmental feature preferably being evaluated depending on the respective sensor type.

[0029] A further aspect of the invention relates to a system for checking the plausibility of a vehicle environment hypothesis, comprising a sensor device with a plurality of different sensors that are designed to detect environmental features, and comprising a computing device that is designed to evaluate the detected environmental features with regard to a plausibility of the vehicle environment hypothesis by checking the respective detected environmental feature for the presence of one or more predetermined plausibility criteria, wherein the plausibility criteria depend on a type of environmental feature and a sensor type with which the environmental feature was detected, and to check the plausibility of the vehicle environment hypothesis on the basis of the evaluated environmental features, wherein the computing device is further designed to weight the plausibility criteria differently during the evaluation depending on the type of environmental feature and / or depending on the sensor type.In other words, the system is designed to carry out a method according to one of the preceding embodiments. This results in the same advantages and possible variations as in the method.

[0030] According to the invention, a motor vehicle with such a system is also provided. The motor vehicle according to the invention is preferably designed as a motor vehicle, in particular as a passenger car or truck, or as a passenger bus or motorcycle. In particular, the motor vehicle can be designed as an autonomous or semi-autonomous vehicle.

[0031] The invention also includes the control device for the motor vehicle. The control device can have a data processing device or a processor device configured to carry out an embodiment of the method according to the invention. For this purpose, the processor device can have at least one microprocessor and / or at least one microcontroller and / or at least one FPGA (Field Programmable Gate Array) and / or at least one DSP (Digital Signal Processor). Furthermore, the processor device can have program code configured to carry out the embodiment of the method according to the invention when executed by the processor device. The program code can be stored in a data memory of the processor device.

[0032] The invention also includes further developments of the system according to the invention that have features already described in connection with the further developments of the method according to the invention. For this reason, the corresponding further developments of the system according to the invention are not described again here.

[0033] The invention also encompasses combinations of the features of the described embodiments. The invention therefore also encompasses implementations that each comprise a combination of the features of several of the described embodiments, unless the embodiments are described as mutually exclusive.

[0034] Exemplary embodiments of the invention are described below. Shown are: Fig. 1 shows a schematically illustrated motor vehicle with a system for checking the plausibility of a vehicle environment hypothesis according to an exemplary embodiment; Fig. 2 is a schematic process diagram according to an exemplary embodiment; Fig. 3 a schematic representation of sensor viewing areas.

[0035] The exemplary embodiments explained below are preferred embodiments of the invention. In the exemplary embodiments, the described components of the embodiments each represent individual features of the invention that can be considered independently of one another, each of which also develops the invention independently of one another. Therefore, the disclosure is intended to encompass combinations of the features of the embodiments other than those shown. Furthermore, the described embodiments can also be supplemented by further features of the invention already described.

[0036] In the figures, the same reference symbols designate elements with the same function.

[0037] In Fig. 1 shows a schematically illustrated motor vehicle 10 with a system 12 for checking the plausibility of a vehicle environment hypothesis according to an exemplary embodiment. The motor vehicle 10 can preferably be an autonomous vehicle configured to drive autonomously using predetermined data. The predetermined data can, for example, be a high-resolution digital map received via a navigation system 14 of the motor vehicle 10. Since it cannot be ensured that the digital map is up-to-date, the system 12 can be configured to check the plausibility of a vehicle environment hypothesis derived from the digital map. For example, a lane hypothesis can be derived as the vehicle environment hypothesis.

[0038] To verify the plausibility of the lane hypothesis, the system 12 may have a sensor device 16 comprising a vehicle camera 18, a radar 20, and a lidar 22. Furthermore, the system 12 may have a computing device 23 that can analyze the sensor measurements, in particular by means of artificial intelligence. The system 12 is preferably designed to Fig. 2 to carry out the procedural steps in order to verify the plausibility of the vehicle environment hypothesis.

[0039] In a step S10, the data from the digital map are used, for example, to form a lane hypothesis. The lane hypothesis can be used, for example, for lateral guidance of a highly automated driving system. Alternatively or in addition to the digital map, swarm data and / or data from artificial intelligence, in particular machine-trained data from a neural network, can also be used as input data to form a vehicle environment hypothesis. As an alternative to the lane hypothesis, further hypotheses can also be formed, for example, regarding traffic signs, traffic regulations (in particular speed limits, overtaking bans and / or right-of-way rules), or stopping points.

[0040] In a step S12, environmental features of the motor vehicle 10 can be detected by the camera 18, radar 20, and lidar 22 of the sensor device 16, wherein the respective sensor data are preferably merged in a sensor data fusion. The various environmental features can be measured by the different sensors. The environmental features can include, for example, traffic signs, lane markings, roadside structures, and object tracks.

[0041] In a step S14, the measured environmental features can then be evaluated with regard to the plausibility of the lane hypothesis. The environmental features can have different significance for the assessment of the safety or plausibility of the lane hypothesis. For example, the various environmental features can have different plausibility criteria with regard to their relevance to the assessment of the lane hypothesis, such as a direct assessment of the safety of the hypothesis, in particular intersections with peripheral buildings; an indirect assessment of the safety of the hypothesis, for example, a match with perceived lane markings; and / or an assessment of the plausibility of the hypothesis, for example, a distance to traffic signs at the edge of the road.Depending on the relevance or significance of the assessed plausibility criterion, this can be included in the assessment of the plausibility of the lane hypothesis with different weighting.

[0042] For example, multiple confidence values ​​can be determined that provide a statement about the plausibility of the lane hypothesis. A confidence value can be determined for each environmental feature, which can also depend on the sensor type of the sensor device 16 used to measure the environmental feature. This means that for each environmental feature, it can also be specified which sensor confirmed the environmental feature or measured the environmental feature. This allows an evaluation table to be compiled that shows which sensors and which environmental features confirm the lane hypothesis, allowing a statement to be made about the quality and plausibility of the hypothesis.

[0043] Traffic signs, for example, can serve as environmental features for evaluating the lane hypothesis. It can be checked whether the number of lanes corresponds to the number of lanes on the traffic sign. It can be checked whether the number of lanes corresponds to the corresponding number of variable message signs on a sign gantry. It can be checked whether the existence of a lane corresponds to a traffic sign that depicts a lane end. It can be checked whether the existence of a lane narrowing corresponds to a traffic sign that depicts a lane end. It can be checked whether the existence of a curve corresponds to a traffic sign that depicts a curve. It can be checked whether detected traffic signs are positioned outside the lane hypotheses. It can be checked,whether existing exits are confirmed by traffic signs and / or it can be checked whether roundabouts are confirmed by traffic signs. Furthermore, as an environmental feature, a lane marking can be checked to determine whether a number of lanes corresponds to a corresponding number of perceived lane markings, it can be checked whether the line type of the lane hypotheses corresponds to the line types of the perceived lane markings, a course of lane markings can be checked and / or it can be checked whether intersections with lane markings are recognized. Furthermore, peripheral buildings, in particular intersections with peripheral buildings, such as concrete walls, beacons, crash barriers or curbs, can be recorded as environmental features and / or object tracks / convoy tracks can be checked as environmental features for parallelism of the object tracks to the lanes,a distance of the object lanes to a lane center and to the boundaries and / or an intersection of object lanes with lane boundaries must be checked.

[0044] Furthermore, when evaluating the individual environmental features, a field of view of the sensors 18, 20, 22 of the sensor device 16 can be taken into account. The field of view of the sensors is, for example, Fig. 3, where Fig.3 shows a schematic plan view of the motor vehicle 10. In this exemplary embodiment, the motor vehicle 10 can have two sensors, for example the camera 18 and the radar 20. Firstly, the field of view of the camera 18 and the radar 20 is determined by an installation position in the motor vehicle 10 and a specification of the respective sensor, which means that an aperture angle and / or a measuring range are predetermined from the specification. For example, the camera 18 can have a field of view 24 in which environmental features can be detected, and the radar 20 can have a field of view 26. The camera 18 and the radar 20 can also have an overlap region 28 in which both sensors can detect environmental features.Thus, for example, environmental features that only occur in the field of view 24 cannot be measured by the radar 20 or vice versa, and environmental features that occur in the overlap area 28 can be detected by both sensors.

[0045] In addition, it can be taken into account whether the fields of vision are restricted, for example by an object 30, which could be another vehicle in the vicinity. This means that environmental features located behind object 30 may not be detected by any of the sensors. This can be taken into account when evaluating the lane hypothesis, for example, whether environmental features that confirm or refute the lane hypotheses are even within the field of vision of the respective sensor. This makes it possible to distinguish whether, in the case of an unconfirmed plausibility criterion, the environmental feature refutes the vehicle environment hypothesis or whether the plausibility criterion could not be confirmed because the environmental feature lies outside the current field of vision of the sensor. This can in turn be taken into account in the resulting quality assessment of the lane hypothesis.A lack of confirmation of the lane hypothesis due to an environmental feature can therefore be treated differently depending on the perceptibility of the environmental feature in the current situation.

[0046] Finally, in a step S16, the lane hypothesis can be checked for plausibility on the basis of the evaluated environmental features, in particular by means of determined confidence values, preferably in order to confirm the lane hypothesis and to use it for lateral guidance of the autonomous motor vehicle 10.

[0047] To determine evaluation criteria that are considered sufficiently safe, the reliability and confidence of various states can be determined in a previous simulation. For example, plausibility criteria can be simulated for various identified environmental characteristics. Ground truth data for the hypotheses can be generated, for example, from high-precision digital maps or labeled data to provide a reference. Additionally, the hypotheses to be evaluated can be falsified using error injection and then evaluated for safety criticality using the ground truth data. By checking the states of an evaluation table for safety-critical initial hypotheses, the criteria in the evaluation table that lead to safe initial hypotheses can be determined.These parameters or plausibility criteria can be determined statistically through various error cases and simulations.

[0048] In a further exemplary aspect, it is provided that a plausibility check and evaluation of a priori data or predetermined data is carried out using sensor-detected environmental features in order to meet safety requirements for the motor vehicle 10 and at the same time benefit from the advantages of the a priori data, for example from digital map data. Data from a digital map can be used, for example, to form lane hypotheses and then compare the resulting result with data from the environmental sensors of a sensor device 16. Various environmental features, which are detected and fused purely by sensors and correspond to the safety requirements for the motor vehicle 10, are used. For each of these environmental features, various criteria that provide a statement about the plausibility or certainty of the hypothesis can be checked.

[0049] Environmental features include, for example, lane markings, peripheral structures, especially guardrails, curbs or walls, traffic signs, or vehicles, including their historical movements (vehicle tracks). For each environmental feature, criteria or plausibility criteria are defined that can be used to determine whether the respective environmental feature confirms the lane hypothesis generated using the given data and thus evaluates it as plausible.

[0050] The plausibility criteria to be checked can be selected to exclude a violation of a safety objective; however, these criteria can vary for different environmental features. Additionally, for each environmental feature, it can be specified which sensor confirmed or measured the environmental feature. This allows an evaluation table to be created showing which sensors and environmental features confirm the lane hypothesis, allowing a statement to be made about the quality and plausibility of the hypothesis.

[0051] If sufficient sensors and environmental characteristics confirm the hypothesis, it can be used for further application in the vehicle control system, since it can be assumed that no safety objective will be violated. A hypothesis that does not meet safety requirements is essentially upgraded to ASIL by performing a plausibility check using environmental characteristic-specific safety criteria (plausibility criteria). A decision as to whether a hypothesis can be considered safe can be made using both hard criteria, which must be met, and soft criteria, which support the confidence of the hypothesis. These confidence criteria can be determined in advance using statistical evaluations with reference data (ground truth).

[0052] When assessing compliance with the defined criteria, the current field of view of the environmental sensors can also be taken into account. A separate field of view can be estimated for each environmental feature by detecting the environmental feature in the current situation. The field of view can be divided into areas in which the contributing sensors differ. This allows a current field of view to be estimated for each environmental feature and for each sensor. Alternatively, a new field of view can be created instead of for each sensor only if the number of sensors differs. This can take into account the installation position of the sensor, sensor specifications, and any limitation of the sensor's field of view due to obscuration in a current situation, for example by other vehicles or peripheral buildings.

[0053] When assessing the plausibility of a lane hypothesis by comparing it with an environmental feature, it can be considered whether the hypothesis from the a priori data actually lies within the currently estimated field of view of the environmental feature and the sensor. This makes it possible to distinguish whether, in the absence of confirmation, the environmental feature or a sensor actually does not confirm the hypothesis from the a priori data or cannot confirm it at all because it lies outside the current field of view. This, in turn, can be taken into account in the resulting quality assessment of the hypothesis. A lack of confirmation of a hypothesis due to an environmental feature is thus treated differently depending on the perceptual capability of the environmental feature in the current situation.

[0054] In addition to lane hypotheses, other hypotheses can also be evaluated from a priori data. These could be traffic signs or stopping points, for example, which can be validated using other environmental features. Hypotheses generated by machine learning methods, such as neural networks, can also be evaluated in this way. Machine learning methods have the disadvantage that they generally do not meet the high safety requirements of the ASIL classification and therefore cannot be used for safety-relevant functions, such as lateral guidance.

[0055] By evaluating environmental features using multiple sensor principles, redundancy in the evaluation can be achieved, which increases reliability. Furthermore, the evaluation capability is taken into account based on the perception capability of the sensors and environmental features, thus distinguishing whether a hypothesis is not confirmed or cannot be evaluated due to the sensor's field of view.

[0056] Overall, the examples show how an evaluation and plausibility check of vehicle environment hypotheses, especially lane hypotheses, can be provided using sensor-based environmental features.

Claims

[1] Method for verifying the plausibility of a vehicle environment hypothesis for a motor vehicle (10), comprising the steps: - Providing (S10) the vehicle environment hypothesis, which is determined from predetermined data and provides an assumption about an environmental property of the motor vehicle (10); - detecting (S12) environmental features in an environment of the motor vehicle (10) by means of a sensor device (16), wherein the sensor device (16) has a plurality of different sensors (18, 20, 22); - Evaluating (S14) the detected environmental features with regard to a plausibility of the vehicle environment hypothesis by checking the respective detected environmental feature for the presence of one or more predetermined plausibility criteria, wherein the plausibility criteria depend on a type of environmental feature and a sensor type with which the environmental feature was detected; - Plausibility check (S16) of the vehicle environment hypothesis based on the evaluated environmental characteristics; the plausibility criteria are weighted differently in the evaluation depending on the type of environmental characteristic and / or the sensor type. [2] Method according to claim 1, wherein the predetermined data from which the vehicle environment hypothesis is determined are a digital map and / or swarm data and / or data from an artificial intelligence, in particular a neural network. [3] Method according to one of the preceding claims, wherein a current field of view (24, 26, 28) of the respective sensors (18, 20, 22) is taken into account when checking the respective environmental feature for the presence of the at least one plausibility criterion. [4] Method according to claim 3, wherein an installation position of the respective sensor (18, 20, 22) and / or a specification of the respective sensor (18, 20, 22), in particular an aperture angle and / or a measuring range, are taken into account for the current field of view. [5] Method according to one of claims 3 or 4, wherein current restrictions of the fields of view (24, 26, 28) of respective sensors (18, 20, 22) are determined by means of a sensor data fusion, wherein the current restrictions of the fields of view are taken into account when checking the respective detected environmental feature for the presence of the at least one plausibility criterion. [6] Method according to one of the preceding claims, wherein the vehicle environment hypothesis is assumed to be plausible if at least two of the detected environmental features confirm the vehicle environment hypothesis with a predetermined confidence. [7] Method according to one of the preceding claims, wherein the at least one plausibility criterion to be checked for the respective environmental feature is predetermined by means of a simulation. [8] Method according to one of the preceding claims, wherein the sensor device (16) comprises a vehicle camera (18), a radar (20) and a lidar (22). [9] System (12) for checking the plausibility of a vehicle environment hypothesis, comprising a sensor device (16) with a plurality of different sensors (18, 20, 22) which are designed to detect environmental features, and comprising a computing device (23) which is designed to evaluate the detected environmental features with regard to a plausibility of the vehicle environment hypothesis by checking the respectively detected environmental feature for the presence of one or more predetermined plausibility criteria, wherein the plausibility criteria depend on a type of environmental feature and a sensor type with which the environmental feature was detected, and to check the plausibility of the vehicle environment hypothesis on the basis of the evaluated environmental features, wherein the computing device (23) is further designed to weight the plausibility criteria differently during the evaluation depending on the type of environmental feature and / or depending on the sensor type. [10] Motor vehicle (10) with a system (12) according to claim 9.

Images