Method and device for supporting a plurality of automated driving systems

Abstract

The invention relates to a method and a device for supporting a plurality of automated driving systems (ADS) by a technical supervisor (TA), comprising the steps: Establishing a communication link for at least one driving situation between at least one of the automated driving systems and the technical supervisor, Acquiring situation information for the at least one driving situation, wherein the situation information comprises environmental information of the at least one of the automated driving systems and ADS information of the at least one of the automated driving systems; Recording of processing data generated in the technical supervision that arise when processing a driving situation, and Automatic analysis of whether a driving situation represents an exceptional condition, and if so, Combining at least part of the recorded situation information and the recorded processing data to form a driving situation data set, wherein the recording of the processing data comprises recording annotations and a classification of the driving situation data set is carried out at least by utilizing the annotations and the driving situation data set is stored together with the classification so that the driving situation data set can be used with other of the many automated driving systems.

Description

[0001] The invention generally relates to automated vehicles that travel at least part of the route autonomously, i.e., without human control. Such a vehicle is generally referred to as an automated driving system (ADS). Such an automated driving system has a plurality of sensors that capture information about the environment of the automated driving system. These include, for example, cameras, radar systems, lidar systems, ultrasonic sensors, humidity sensors, global satellite navigation systems, etc., to name a few examples. Furthermore, information is captured by sensors that characterize the automated driving system itself, e.g., speed, orientation, etc. In addition to the information captured in the automated driving system, further environmental information can also be stored in the automated driving system, e.g.,Digital maps, such as maps, are transmitted to the automated driving system via communication technology and incorporated into the control system. For example, information can originate from vehicle-to-vehicle communication or from infrastructure-to-vehicle communication. Information collected and provided by other vehicles or information collected from traffic infrastructure can be incorporated into the control system. The collected information is processed and evaluated as data in a data processing device, and characteristics or parameters are derived, and the automated driving system is controlled based on these parameters.

[0002] To ensure safety when using automated driving systems, it is necessary to define conditions and / or operating ranges within which such an automated driving system can and may be operated safely. These conditions and / or operating ranges, which determine whether and how the vehicle can and may be operated automatically, are summarized in an Operational Design Domain (ODD), sometimes also referred to as the Operational Domain (OD). For example, it specifies which environmental information from an automated vehicle must be available with a sufficiently determined reliability so that the automated driving system can and may be moved.For example, a speed limit of 100 km / h may be set for a rural road journey, and it may be required that both a radar signal be present, the evaluation of which indicates a clear road ahead of the automated vehicle, and captured camera images be present, the evaluation of which provides the lane markings of the road space and also indicates a clear road space. These two conditions are greatly simplified. In practice, a multitude of additional conditions define the operational domain.

[0003] To implement such control systems that can analyze the large, constantly changing, and always unique volumes of data, data processing systems that implement so-called artificial intelligence are used. Artificial intelligence encompasses systems that are trained using test data for which the corresponding characteristics to be derived are specified, or the associated control behavior is specified. These principles are familiar to experts in the field of automated driving systems, so they will not be explained in detail here.

[0004] In principle, the frequently occurring driving situations and the information generated by the automated driving system and the environment of the automated driving system can be easily generated and trained.

[0005] However, situations repeatedly arise in real operation in which automated movement of the automated driving system is not possible because the derived parameters occur in a combination that characterizes a state outside the operational design domain of the automated vehicle. For example, such a case can arise if an obstacle not included in the digital map of the automated driving system, e.g. a construction site barrier, is detected in a lane whose length the automated driving system cannot determine. Such driving situations are also referred to as corner cases, in which a unique combination of several already rare parameters occurs, leading to a unique scene. Such cases can usually be easily recognized in practice, as intervention by a higher-level authority, a human vehicle passenger, or a technical supervisor (TA) is required.

[0006] This technical supervisor is capable of communicating with the automated driving system and evaluating the information it collects. If necessary, the technical supervisor has access to additional environmental information, such as from surveillance cameras, to assess the driving situation. The technical supervisor is also capable of intervening either directly in the control of the automated driving system or indirectly, for example, by specifying a trajectory for the vehicle to safely avoid the obstacle, along which the automated driving system then avoids the obstacle.

[0007] However, situations also occur in which the automated driving system reaches the edge of its operational design domain with one or more of the relevant parameters, particularly in driving situations where this is not expected. These driving situations are referred to as edge cases. Edge cases also include driving situations in which the automated driving system behaves differently than expected, for example, it does not use the traffic areas designated for vehicle traffic to avoid an obstacle, or instead of braking the automated driving system evenly and continuously with a moderate deceleration before reaching a bottleneck that can only be negotiated at reduced speed, it first increases the speed and then brakes the automated driving system to the reduced speed before reaching the bottleneck with maximum deceleration.These are highly simplified examples that are intended only to illustrate the principle.

[0008] In the following, both corner cases and edge cases are referred to as exceptional states.

[0009] US 2017 / 0192426 A1 describes a system and method for controlling an autonomous vehicle. In one embodiment, a method for controlling an autonomous vehicle includes determining that assistance is required for continued movement of the autonomous vehicle; and enabling movement of the autonomous vehicle by implementing manual instructions, by a processor onboard the autonomous vehicle, provided by a user remote from the autonomous vehicle.

[0010] US 2008 / 033603 A1 describes a system for the automated driving of a motor vehicle to a target position. It comprises a sensor system and an evaluation unit for determining the position of the motor vehicle relative to the target position, as well as devices for planning a collision-free journey from the current position to the target position and devices for implementing the planned journey. A first part of the sensor system is arranged on or in the motor vehicle, and a second part of the sensor system is fixed in a defined position near the target position.

[0011] From DE 10 2020 205 315 A1 a method and a device for identifying critical driving situations during operation of an autonomous vehicle, selecting similar data and retraining an automatic driving system of the autonomous vehicle is known, comprising the steps: identifying a critical driving situation during operation of the autonomous vehicle; requesting and obtaining data from a backend server, which are similar to the critical driving situation, by an intelligent assistant; and using the similar data to train the automatic driving system of the autonomous vehicle.

[0012] DE 10 2017 217 443 A1 discloses a method for providing training data for machine learning for a control model of an automatic vehicle control system. A vehicle is controlled along a trajectory, with sensor data sets being acquired by the vehicle's sensors. The sensor data sets comprise position data, time data, and control data of the vehicle. Transmission data is generated from the sensor data sets and transmitted to a vehicle-external processing unit. The vehicle-external processing unit acquires additional information from the transmission data, and the training data is generated from the transmission data and the additional information, with classification information being assigned to the training data. The training data is stored with the associated classification information.DE 10 2017 217 443 A1 further relates to a system for providing training data for machine learning for a control model of an automatic vehicle control system.

[0013] From DE 10 2019 204 941 A1 a system for teleoperated driving is known, wherein the system comprises a vehicle, a backend and remote control devices, wherein the vehicle, the backend and the remote control devices are configured to communicate via a mobile network.

[0014] From DE 10 2022 107 845 A1, a method for selecting concrete scenarios, in particular for testing and training driver assistance systems and automated driving functions, is known, wherein a scenario represents a traffic situation in a temporal sequence and is defined by a selection of parameters and associated parameter values, and wherein the parameters and associated parameter values ​​are specified for a concrete scenario, comprising determining data by means of sensors of properties and features of a motor vehicle and of objects and events in the surroundings of the motor vehicle when driving along a route, processing the data to classify a real scenario to characterize the driving situation, extracting parameters and parameter values ​​for the determined set of real scenarios,analyzing the determined set of real-world scenarios with regard to the occurrence of clusters of scenario types with similar parameters and parameter values ​​and / or rarely occurring scenario types such as corner cases, classifying a relevant set of real-world scenarios, calculating concrete scenarios using the classified real-world scenarios, and outputting the calculated concrete scenarios for testing and training purposes.

[0015] DE 10 2013 007 502 A1 discloses a learning dialog system and a computer-implemented method for the semantic training of a dialog system. Semantic annotations are automatically generated based on received speech inputs, wherein the semantic annotations are intended for controlling devices or for communicating with a user. For this purpose, at least one speech input is received during an interaction with a user. The meaning of the speech input is captured and evaluated by classifying the speech input based on a trainable semantic model in order to provide a semantic annotation for the speech input. Additional user information related to the speech input is considered if the captured meaning is incorrect, incomplete, and / or assessed as untrustworthy.The meaning of the voice input is automatically learned based on the additional user information.

[0016] US 10,444,754 B2 describes example systems and methods that enable an autonomous vehicle to request assistance from a remote operator when the vehicle's confidence in autonomous operation is low. An example method includes operating an autonomous vehicle in a first autonomous mode. The method may also include identifying a situation in which a confidence level of autonomous operation in the first autonomous mode is below a threshold. The method may further include sending a request for assistance to a remote assistant, wherein the request includes sensor data representing a portion of an environment of the autonomous vehicle. The method may additionally include receiving a response from the remote assistant, wherein the response indicates a second autonomous mode of operation.The method may also include operating the autonomous vehicle in the second autonomous operating mode in accordance with the response of the remote assistant.

[0017] The invention is based on the object of improving the support of a large number of automated driving systems through technical supervision.

[0018] The object is achieved according to the invention by a method having the features of patent claim 1 and a device having the features of patent claim 7. Advantageous embodiments of the invention emerge from the subclaims.

[0019] The invention is based on the idea of ​​segmenting the data streams generated during technical supervision and assigning them to driving situations and classifying them. This allows the knowledge contained therein to be used for improved control not only of one automated system, but also for improved control of other automated driving systems. For this purpose, annotations relating to the driving situation are generated and / or recorded during technical supervision. This makes it possible to classify the driving situations and an associated driving situation data set, which is also referred to as labeling, and to save this data set together with the classification. This makes the data available for further use.

[0020] In particular, a method for supporting a plurality of automated driving systems (ADS) by a technical supervisor (TA) is created, which comprises the steps: Establishing a communication connection for at least one driving situation between at least one of the automated driving systems and the technical supervisor for data exchange, via which at least one of the automated driving systems can be controlled at least directly or indirectly by the technical supervisor, Acquiring situation information for the at least one driving situation of the at least one of the automated driving systems, wherein the situation information comprises environmental information of the at least one of the automated driving systems and ADS information of the at least one of the automated driving systems for the one driving situation; Recording the processing data generated by the technical supervisor when processing a driving situation, and Automatic analysis of whether a driving situation represents an exceptional condition, and if so, Combining at least part of the recorded situation information and at least part of the recorded processing data to form a driving situation data set, wherein the recording of the processing data comprises recording annotations and a classification of the driving situation data set is carried out at least by utilizing the annotations and the driving situation data set is stored together with the classification, so that the driving situation data set can be used in connection with other of the multitude of automated driving systems to improve their automated control.

[0021] Furthermore, a device for improving the automated driving of a plurality of automated driving systems is provided, which comprises a technical supervisor, which comprises at least one communication device for establishing a communication connection for at least one driving situation between at least one of the automated driving systems and the technical supervisor for a data exchange, by means of which the at least one of the automated driving systems can be controlled at least directly or indirectly by the technical supervisor, an information collection device coupled to the at least one communication device for collecting situation information for the at least one driving situation of the at least one of the automated driving systems,wherein the situation information comprises environmental information about the environment of the at least one of the automated driving systems and ADS information of the at least one of the automated driving systems for the one driving situation, at least one processing device which is designed to record processing data which arise when processing the driving situation in the technical supervision, an analysis unit which is designed to automatically analyze whether the driving situation represents an exceptional condition, and if this is the case, to combine at least part of the recorded situation information and at least part of the recorded processing data into a driving situation data set, and a classification device is designed to classify the driving situation data set on the basis of the processing data comprising recorded annotations, at least by utilizing the annotations, and a storage device configured to store the driving situation data set together with the classification, so that the driving situation data set can be used in connection with further of the plurality of automated driving systems to improve their automated control.

[0022] The advantage of the invention is that driving situation data sets are generated during the assistance process, which can be used directly by other automated driving systems or for their development. For example, if the technical supervisor at a new construction site defines a trajectory that several automated driving systems are unable to calculate due to their operational design domain, these automated systems can use this driving situation data set to circumvent the construction site along the trajectory contained therein as processing data.

[0023] For example, the technical supervisor connects to the automated driving system via communication technology after the automated driving system has transmitted a support request because, for example, it is unable to determine a trajectory for driving through a section of road due to its own operational design domain. Such a situation can arise, for example, if the road width is restricted due to snowfall and / or at least parts of the road markings are no longer visible due to snow. In one embodiment, the annotations can be derived from comments that an operator in the technical supervisor makes while thinking aloud. The operator summarizes his thoughts in spoken words and sentences, which are recorded audio-technically. Using speech recognition, parameters and content can be assigned to the comments, which are then used for classification.

[0024] One embodiment therefore provides that the annotations are at least partially captured by means of think-aloud comments.

[0025] An embodiment of the device therefore provides that the processing device comprises a detection device for acoustic signals and a speech recognition device in order to detect comments of a thinking aloud during the processing of the at least one driving situation and to detect expected dynamic behavior of the at least one automated driving system or control-relevant parameters recognized by the technical supervisor.

[0026] The annotation can also be recorded as text or similar, entered via a keyboard or other input device of a computer.

[0027] Alternatively or additionally, the annotations can be derived at least partially automatically based on the recorded situation data and a portion of the processing data collected as part of the assistance provided by the technical supervisor in a particular driving situation. If an operator is processing the support for the driving situation specified above, they can, for example, create waypoints for the trajectory to be traveled and transmit them to the automated driving system. An annotation determination device can use the environmental information to determine that the road is at least covered in snow. Furthermore, the geoposition of the automated driving system can be derived from the situation data.Based on the processing actions, the definition of waypoints, and transmission to the automated driving system, the annotation determination device can derive annotations for the snow-covered roadway at geoposition x and the driving trajectory at the current time, so that the driving situation data can be saved with the classification "Route trajectory for driving through the road section at geoposition X under snow cover." Of course, other parameters can be evaluated and used to improve the classification, for example, the road width between the snow masses narrowing the lanes, etc.

[0028] In one embodiment, it is provided that the operational design domain of the automated driving system is adapted. This can consist, for example, in changing a mode, for example from a normal operating mode to a winter mode. Likewise, the addition of the trajectory for the route section with snow barriers, with the further proviso that the width between the snow masses at the lane edges deviates by a maximum of a specified percentage, can be used as an extension of the operational design domain. The vehicle can now drive through this section under similar or identical conditions in the future without encountering an exceptional driving situation, i.e., a corner case, in which support from the technical supervisor is necessary.

[0029] However, the driving situation data recorded in this way can also be used by other automated driving systems. If another automated driving system arrives at the same geolocation and determines that the parameters for roadway detection are not met and there is snow on the road, the stored driving situation data can be identified based on the classification. The additional automated driving system can evaluate this data and recognize that the trajectory contained therein allows driving through the road section in snowy conditions, provided that the clear width between the snow masses at the lane edges does not fall below the specification set by the technical supervisor during the original processing of the exceptional driving condition, during which the driving situation data set was recorded and classified.

[0030] One embodiment thus provides that a geoposition is assigned to the driving situation data set, which is stored with the driving situation data set, and the driving situation data set is transmitted to at least one further automated driving system whose position is in the vicinity of the one geoposition, so that the at least one further automated driving system uses the information contained in the driving situation data set for automated driving.

[0031] An embodiment of the device therefore provides that the technical supervision comprises a distribution device which is designed to transmit the driving situation data set to at least one further one of the plurality of automated driving systems in order to improve its automated driving thereby.

[0032] Exceptional driving situations can be detected by an automated driving system itself or by other automated driving systems, for example, if they detect unexpected or conspicuous behavior. For example, if automated driving systems repeatedly detect on a section of road that another automated driving system at a specific geoposition significantly reduces its speed and / or changes to the oncoming lane for a short section of road when the section is clear, as if bypassing an obstacle even though the lane is clear, the technical supervisor can be notified that vehicles exhibiting unexpected driving behavior have been detected. The technical supervisor can then communicate with vehicles traveling through this section as in a normal support situation and evaluate the information recorded by the automated driving system.

[0033] For this support, the operator preferably uses the think-aloud method. This involves verbally describing observations of the driving situation, preferably using standardized words and phrases. Furthermore, comments on the expected and / or observed dynamic behavior of the automated driving system are preferably also verbally expressed. These comments are recorded and evaluated using speech recognition. Certain parameters are preferably assigned to the comments. These can be used to derive the classification. However, they can also be compared with parameters that are derived, for example, from the automated driving system. In the last example described, the following parameters could be derived from speech recognition: "Road is clear; steady, non-accelerated driving expected; unexpected braking of the vehicle."Here, it can be deduced from the detected parameters that this is an edge case, because the automated driving system brakes unexpectedly. Using this driving situation data, a vehicle manufacturer, for example, can then determine why automated driving systems of certain types behave unexpectedly at the corresponding geolocation.

[0034] When analyzing the captured image data from the driving situation data, it becomes apparent, for example, that a stationary, moving object (e.g., a sculpture with elements that move in the wind) is located at the corresponding geoposition at a distance from the roadway. Human drivers and the technical supervisor correctly identify this as a stationary object and classify it accordingly. However, the automated driving systems detect, for example, a person moving toward the lane, causing them to reduce the vehicle's speed.

[0035] It should be noted that this is a highly simplified example, and the causes of unexpected vehicle behavior are often very complex. Therefore, it is desirable to capture and classify such driving situations using a variety of environmental information and information about the automated driving system.

[0036] Some edge cases cannot be directly detected by a human operator. Rather, their existence is revealed, for example, by evaluating the parameters derived from thinking aloud and the corresponding parameters derived from the automated driving system.

[0037] One embodiment provides that during the automatic analysis of whether an exceptional situation exists, it is evaluated whether an expected dynamic behavior of at least one of the automated driving systems derived from the recorded comments deviates from the dynamic behavior determined based on the environmental information and ADS information and / or whether control-relevant parameters derived from the recorded comments deviate from equivalent control parameters derived from the ADS information. If this is the case, an exceptional driving situation exists. This is then automatically classified, for example, based on the control parameters for which a deviation has been determined. If the comments contain classification information that, for example, identify the driving situation as an edge case or corner case, these are used additionally and / or alternatively.

[0038] The invention is explained in more detail below with reference to a drawing. Herein: Fig. 1 a schematic diagram to explain the support of an automated vehicle by a technical supervisor; Fig: 2 a schematic representation of a vehicle driving through a road section which is narrowed by snow at the lane edges.

[0039] In Fig. Figure 1 schematically illustrates a process for support by a technical supervisor (TA) for an automated driving system (ADS). First, there is so-called vehicle-to-X communication 10, with which, for example, an automated driving system transmits information to a technical supervisor and / or other, particularly automated, driving systems and vehicles. This can include information about the vehicle's surroundings as well as information concerning the vehicle itself. The former is referred to as environmental information, and the latter as ADS information or vehicle information.

[0040] The technical supervisor also receives information, for example about a road section, via so-called vehicle-to-infrastructure communication 20. The information can be transmitted from the infrastructure to a vehicle and then from there to the technical supervisor. Alternatively and / or additionally, the infrastructure facilities can also transmit information directly to the technical supervisor. In addition, there is communication from traffic control centers 30 to the technical supervisor, which also transmits information. In addition, further information, for example general environmental information such as weather information, etc., can be transmitted 40. Based on these various communications 10, 20, 30, 40, the technical supervisor receives a request for a support service 50. The technical supervisor then contacts the automated driving system 60, which requires a support service.The automated driving system, i.e. the vehicle, accepts the request transmitted by the technical supervisor and establishes a safe state or a risk-minimal state itself 70, unless it is already in a safe state.

[0041] The technical supervisor optionally communicates with passengers of the automated driving system 80 to inform them that the technical supervisor is intervening in the control of the automated driving system. Through this communication, the technical supervisor can also collect additional information about the current driving situation of the automated driving system.

[0042] The technical supervisor then processes the current driving situation of the automated driving system 100. The technical supervisor first records and compiles all information relevant to the current driving situation of the automated vehicle 110. This includes, on the one hand, vehicle information 120, also referred to as ADS information, and environmental information 130. This information can be recorded and / or recorded by the technical supervisor via the various communication channels 10, 20, 30, 40. The necessary data can be transmitted, for example, via mobile radio and / or WLAN connections of the automated driving system 100.

[0043] The technical supervisor then obtains an overview of the current driving situation 140. The technical supervisor then records processing actions 150 that are carried out either by an automated support system of the technical supervisor or, for example, by an operator in the technical supervisor. For example, a support request may consist of the automated driving system requesting a specific feature or parameter 160 that is required for decision-making in the automated control system. For example, an object detected by a camera that cannot be identified by the automated driving system, such as colored markings on a roadway that are not clearly identifiable as road markings, may trigger such a request.The technical supervisor shall independently identify a relevant feature and / or determine a parameter that the vehicle requires for decision-making 170.

[0044] As part of processing the driving situation, the technical supervisor records various inputs 180. These may include, for example, annotations about the driving situation 190. These annotations can be recorded, for example, through speech capture and analysis of the recorded speech of an operator in order to capture parameters through thinking aloud. The annotations can also be recorded using other input methods.

[0045] The think-aloud method involves a human operator speaking their thoughts aloud while performing processing actions. It is preferable to use language that is as standardized as possible to facilitate speech recognition and assignment to parameters and features. Thinking aloud involves verbalizing the human operator's observations and preferred actions, as well as expressing expected behavior from the automated driving system. In addition, observations of the automated vehicle's dynamic behavior are also preferably verbalized. This method allows a wide range of information to be captured without impairing and / or delaying processing of the driving situation.

[0046] In order to provide assistance for an automated driving system, it may, for example, be provided that the technical supervisor defines waypoints for a trajectory 200 along which a vehicle can travel a section of the route for which it is not able to determine a valid trajectory itself due to its internal control systems.

[0047] Furthermore, the technical supervisor may, for example, check the Operation Design Domain of the automated driving system during processing and, if necessary, adapt or change it 210. For example, a different mode, such as a winter or rain mode, may be activated due to weather conditions.

[0048] It is understood that only exemplary, very simplified possibilities are described here. In addition to annotations relating to the general driving situation, annotations relating to vehicle behavior can also be recorded 220. A variety of other information generated by operating actions can also be recorded 230.

[0049] When processing the driving situation, the technical supervisor intervenes in the vehicle's control system directly or indirectly. 240 An indirect intervention, for example, involves transmitting to the vehicle the trajectory along which the automated driving system will travel the corresponding section of the route. With direct control, the technical supervisor steers the automated driving system itself along the trajectory it has defined.

[0050] On the one hand, the technical supervisor becomes active when it receives a request for support services, as in the case described so far. However, if the traffic volume and workload allow, the technical supervisor can also preferably take action on its own initiative in certain situations and contact individual automated driving systems. This can also take place while the automated driving system itself retains control over its own driving behavior. Such proactive support from the technical supervisor can, for example, be based on empirical knowledge if an automated driving system enters a section of the route in which other automated driving systems have previously encountered an exceptional driving condition, such as a corner case. The technical supervisor can then also individually provide its support service for this additional vehicle, i.e.automated driving system, or even simply annotate and comment on the automated journey. This process can include, for example, checking and comparing features and parameters identified by the technical supervisor through annotation with comparable parameters and features identified by the automated driving system. Any deviations found may indicate an exceptional driving condition, a so-called edge case.

[0051] As mentioned above, the annotations can also be recorded using other input devices. For example, an operator can enter inputs via a touch display while processing a particular driving situation. It is also possible for the operator to simulate a manual drive in a so-called vehicle seat box simultaneously with the real drive of the automated driving system. This allows parameters of the automated driving system to be compared with those "generated" by the operator's virtually steered vehicle. This ad hoc simulation is particularly useful for identifying edge cases. In addition to haptic and mechanical input devices, acoustic and other input devices can also be used. For example, the operator's eyes can be tracked while viewing camera images or point clouds captured by the vehicle and / or the infrastructure, which represent the dynamic behavior of one or more automated driving systems.All input modalities can be combined.

[0052] In general, regardless of whether the technical supervisor has acted in response to a request from the automated driving system or proactively, it is determined whether an exceptional driving situation exists 250. If this is the case, the available information, which includes the environmental information, the ADS information and the information recorded from the processing actions, is combined into a driving situation data record 260. This is classified, among other things, based on the recorded annotations 270. Furthermore, the driving situation data record is stored 280 in order to be able to use it for the improved automated driving of other automated driving systems.For example, via communication with traffic infrastructure, which in turn communicates with vehicles, a stored driving situation data set can be output and communicated to vehicles whose geoposition in an environment coincides with the geoposition for which the support service is contained in the corresponding driving situation data set 290. Further automated driving systems can thus evaluate the driving situation data set 300 and, if necessary, use the information contained therein to avoid falling into a corner case themselves, but rather, for example, use the trajectory defined by the traffic authorities themselves 310 to be able to drive through a specific section of the route. The further automated driving system can adapt its own operational design domain 300.However, the driving situation data can also be used to adapt the operational design domain of other automated driving systems, for example by a vehicle manufacturer who evaluates the driving situation data sets 320.

[0053] In Fig. Figure 2 depicts a situation in which an automated driving system 1070 intends to travel through a road section 1000. The guideposts 1010, which provide orientation for the course of the road section, are clearly visible. The automated driving system 1070, which is a passenger car, is moving toward the viewer in a lane 1020. Next to it is another lane 1030 for oncoming traffic. An original and unrestricted total lane width 1050 is limited at the lane edges by snow mounds 1040, so that only a restricted lane width 1060 is available for driving.

[0054] It can be seen that road markings are obscured by snow.

[0055] Also shown is the trajectory 1080 calculated by the automated driving system 1070, which is shown as a dotted line. The automated driving system calculates this, for example, based on its digital map information, which is stored internally in the automated driving system.

[0056] Based on internal sensors (not explicitly shown) installed in the automated driving system 1070, the automated driving system 1070 determines that it must maintain a minimum distance from the snow mounds 1040 at the edge of the lane. The self-calculated trajectory 1080, along which a center position 1100 of the automated driving system 1070 is supposed to move, cannot therefore be followed by the automated driving system.

[0057] Also shown is a trajectory 1090 generated and classified by the technical supervisor, along which the automated driving system 1070 can travel through the partially snow-covered road section 1000. The classification of the trajectory is also referred to as labeling. This trajectory 1090, determined by the technical supervisor, can now also be used by other vehicles that want to travel through the same road section 1000 at a time close to the time determined by the technical supervisor. Thus, a driving situation data record generated during the support service is stored and made accessible, for example, by the technical supervisor to other automated driving systems that want to travel through the same road section 1000 at a time close to the time specified by the technical supervisor, for example, one to two hours after the support service was provided on the same day, without further snowfall being detected based on the recorded weather conditions.

[0058] It is understood that only exemplary embodiments are described here. The individual features of the various embodiments can be implemented together in combination to form other embodiments. Individual method steps mentioned may also be omitted in other embodiments. Reference symbol 10 Vehicle-to-X communication 20 Vehicle-to-Infrastructure Communication 30 Communication with traffic control centers 40 Transmission of general environmental information (e.g. weather information) 50 Request arrives at technical supervision 60 technical supervisor contacts automated driving system 70 automated driving systems accept requests 80 technical supervision Communicates with passengers 100 technical supervision processes driving situation 110 technical supervision records information 120 technical supervision records ADS information 130 technical supervision records environmental information 140 technical supervision provides an overview 150 processing operations are recorded in the technical supervision 160 automated driving system requests specific parameter / feature 170 technical supervision identifies feature / determines parameter 180 Entering inputs 190 annotations on the driving situation 200 speech capture and analysis to capture parameters by thinking aloud 200 Capturing the definition of a trajectory 210 Checking and / or re-determining the OD of the automated driving system 220 Annotation of the behavior of the automated driving system 230 Recording further assistance actions 240 direct or indirect intervention in the vehicle control system 250 Determine whether an exceptional driving situation exists 260 Summarizing the environment, ADS information and processing actions into a driving situation data set 270 Classifying the driving situation dataset based on the annotations 280 Saving the driving situation data record 290 Distribution of at least one driving situation data set to other automated driving systems 300 Evaluation of the driving situation data set 310 Adapting driving behavior based on the driving situation data set 320 Adapting the ODD of other automated driving systems 1000 road sections 1010 delineators 1020 lanes (for oncoming vehicles) 1030 additional lanes 1040 snow hills 1050 unrestricted total lane widths 1060 restricted lane width 1070 automated driving system 1080 calculated trajectory 1090 generated and classified (labeled) 1100 Center position of the automated driving system

Claims

[1] Method for supporting a plurality of automated driving systems (ADS) (1070) by a technical supervisor (TA), comprising the steps: Establishing a communication connection for at least one driving situation between at least one of the automated driving systems (1070) and the technical supervisor for a data exchange, via which the at least one of the automated driving systems (1070) can be controlled at least directly or indirectly by the technical supervisor, Acquiring (110, 120, 130) situation information for the at least one driving situation of the at least one of the automated driving systems (1070), wherein the situation information comprises environmental information of the at least one of the automated driving systems (1070) and ADS information of the at least one of the automated driving systems (1070) for the one driving situation; Recording of processing data generated in the technical supervision that arise when processing a driving situation, and Automatic analysis of whether a driving situation represents an exceptional condition, and if so, Combining at least part of the acquired situation information and at least part of the acquired processing data to form a driving situation data record (260), wherein the acquisition of the processing data comprises the acquisition of annotations and the driving situation data record is classified (270) at least using the annotations, and the driving situation data record is stored together with the classification (280), so that the driving situation data record can be used in conjunction with further automated driving systems (1070) to improve their automated control. [2] Method according to claim 1, characterized bythat the annotations are derived at least partially automatically from the situation information recorded and part of the processing data recorded as part of the assistance provided by the technical supervisor in one driving situation. [3] Method according to claim 1 or 2, characterized by that the processing data comprise an adaptation of the operational domain (OD) of at least one of the automated driving systems (1070). [4] Method according to one of the preceding claims, characterized by that the annotations are at least partially captured by comments from thinking aloud. [5] Method according to claim 4, characterized bythat during the automatic analysis of whether an exceptional state exists, it is evaluated whether an expected dynamic behavior of at least one of the automated driving systems (1070) derived from the recorded comments deviates from the dynamic behavior determined on the basis of the environmental information and ADS information and / or whether control-relevant parameters derived from the recorded comments deviate from equivalent control parameters derived from the ADS information. [6] Method according to one of the preceding claims, characterized bythat a geoposition is assigned to the driving situation data set, which is stored with the driving situation data set and the driving situation data set is transmitted to at least one further one of the automated driving systems (290), the position of which is in the vicinity of the one geoposition, so that the at least one further one of the automated driving systems (1070) uses the information contained in the driving situation data set for automated driving. [7] Device for improving the automated driving of a plurality of automated driving systems (1070) comprising a technical supervisor, which at least one communication device for establishing a communication connection for at least one driving situation between at least one of the automated driving systems (1070) and the technical supervisor for a data exchange, by means of which the at least one of the automated driving systems (1070) can be controlled at least directly or indirectly by the technical supervisor, an information collection device coupled to the at least one communication device for collecting situation information for the at least one driving situation of the at least one of the automated driving systems (1070),wherein the situation information comprises environmental information about an environment of the at least one of the automated driving systems and ADS information of the at least one of the automated driving systems (1070) for the one driving situation, at least one processing device which is designed to record processing data which arise when processing the driving situation in the technical supervision, an analysis unit which is designed to automatically analyze whether the driving situation represents an exceptional state and, if this is the case, to combine at least part of the acquired situation information and at least part of the acquired processing data into a driving situation data set (260), and a classification device is designed to classify the driving situation data set based on the processing data comprising acquired annotations, at least by utilizing the annotations (270), and a storage device configured to store (280) the driving situation data set together with the classification, so that the driving situation data set can be used in conjunction with further of the plurality of automated driving systems (1070) to improve their automated control. [8] Device according to claim 7, characterized bythat the processing device comprises a detection device for acoustic signals and a speech recognition device in order to detect comments of a thinking aloud during the processing of the at least one driving situation and to detect expected dynamic behavior of the at least one of the automated driving systems (1070) or control-relevant parameters detected by the technical supervisor. [9] Device according to claim 7 or 8, characterized by that the technical supervision comprises a distribution device which is designed to transmit (290) the driving situation data set to at least one further one of the plurality of automated driving systems (1070) in order to improve its automated driving thereby.

Images