Simulation scene generation method, device, computer equipment and storage medium

Abstract

The present application provides a simulation scene generation method, device, computer equipment and storage medium, the method comprising: obtaining accident road test data, and extracting vehicle status information, map information and motion information of traffic participants from the accident road test data; obtaining manual driving time according to the vehicle status information; using the vehicle status information and map information corresponding to when the road test vehicle is in an automatic driving state and a target driving period as the first driving information, and using the vehicle status information and map information corresponding to when the road test vehicle is in a manual driving state as the second driving information; the end time of the target driving period is the manual driving time; judging whether an avoidance event occurs during the road test according to the first driving information, the second driving information and the motion information of traffic participants; if so, extracting a data segment corresponding to the avoidance event from the accident road test data, and generating an avoidance simulation scene based on the data segment.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a simulation scene generation method, device, computer equipment and storage medium. Background Art

[0002] Before the large-scale application of autonomous driving algorithms, it is necessary to test the autonomous driving algorithms to ensure that the algorithms can safely deal with various dangerous scenarios that the vehicle may encounter during actual driving. When the vehicle is driving on the actual road, the driving environment is complex and the vehicle will face some scenarios that are prone to collision with other traffic participants. In order to test whether the autonomous driving algorithm's response strategy for this scenario is appropriate, the simulation system can be used to simulate the avoidance simulation scenario and obtain the control of the vehicle by the autonomous driving algorithm in this scenario to determine whether the algorithm can control the vehicle to avoid the traffic participant and thus avoid the collision.

[0003] In order to improve the authenticity of the avoidance simulation scene, the accident road test data can be used to construct the scene. However, the amount of accident road test data is very large, and the proportion of scenes that are valuable to autonomous driving technology is small and the types are scattered, so the complete accident road test data cannot be directly used for simulation. The existing technology generally generates avoidance simulation scenes by having safety officers report accident road test data, manually review and classify, manually configure simulation time, and manually configure result evaluation criteria. All links require manual participation, and there is a problem of low scene conversion efficiency. With the continuous expansion of the scale of road testing, there is an urgent need to provide a solution that can efficiently generate avoidance simulation scenes. Summary of the invention

[0004] The purpose of this application is to solve at least one of the above-mentioned technical defects, especially the technical defect of low scene conversion efficiency in the prior art.

[0005] In a first aspect, an embodiment of the present application provides a simulation scene generation method, the method comprising:

[0006] Acquire accident road test data, and respectively extract vehicle status information, map information, and movement information of traffic participants from the accident road test data;

[0007] Acquire a manual driving time according to the vehicle state information, where the manual driving time is the time when the road test vehicle switches to a manual driving state;

[0008] The vehicle state information and map information corresponding to the road test vehicle in the automatic driving state and the target driving period are used as the first driving information, and the vehicle state information and map information corresponding to the road test vehicle in the manual driving state are used as the second driving information; the end time of the target driving period is the manual driving time;

[0009] According to the first driving information, the second driving information and the movement information of the traffic participant, it is determined whether an avoidance event occurs during the road test; if so, a data segment corresponding to the avoidance event is extracted from the accident road test data, and an avoidance simulation scene is generated based on the data segment.

[0010] In one embodiment, the step of determining whether an avoidance event occurs during the road test according to the first driving information, the second driving information, and the motion information of the traffic participant includes:

[0011] Acquire a predicted driving trajectory of the road test vehicle based on the first driving information, and acquire a predicted motion trajectory of a traffic participant based on the motion information of the traffic participant;

[0012] Determine whether the predicted motion trajectory of the traffic participant intersects with the predicted driving trajectory of the road test vehicle, and obtain an intersection determination result;

[0013] Based on the intersection judgment result and the second driving information, it is determined whether the avoidance event occurs during the road test.

[0014] In one embodiment, the step of determining whether the avoidance event occurs during the road test based on the intersection determination result and the second driving information includes:

[0015] Based on the second driving information, determining whether the road test vehicle has an evasive behavior when in the manual driving state, and obtaining an evasive behavior determination result; wherein the evasive behavior is a behavior that can increase the distance between the driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant, or the evasive behavior is a behavior that can delay the intersection time between the driving trajectory of the road test vehicle and the predicted motion trajectory;

[0016] Based on the avoidance judgment result and the intersection judgment result, it is determined whether the avoidance event occurs during the road test.

[0017] In one embodiment, the step of determining whether an avoidance event occurs during the road test according to the first driving information, the second driving information, and the motion information of the traffic participant further includes:

[0018] Acquire a first driving safety score of the road test vehicle in the automatic driving state according to the first driving information;

[0019] Acquire a second driving safety score of the road test vehicle in the manual driving state according to the second driving information, and compare the second driving safety score with the first driving safety score to obtain a comparison result;

[0020] The step of determining whether the avoidance event occurs during the drive test based on the avoidance determination result and the intersection determination result includes:

[0021] Based on the avoidance judgment result, the intersection judgment result and the comparison result, it is judged whether the avoidance event occurs during the road test.

[0022] In one embodiment, the step of extracting a data segment corresponding to the avoidance event from the accident road test data and generating an avoidance simulation scenario according to the data segment includes:

[0023] The time corresponding to the intersection of the predicted driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant is taken as the first target time;

[0024] Acquire a second target time according to the second driving information, and determine a scenario generation period based on the first target time and the second target time; the second target time is the earliest time corresponding to when the traveling direction of the road test vehicle does not intersect with the position of the traffic participant;

[0025] A data segment corresponding to the scenario generation period is extracted from the accident road test data, and an avoidance simulation scenario is generated according to the data segment.

[0026] In one embodiment, the step of determining a scene generation period based on the first target moment and the second target moment includes:

[0027] Calculating the difference between the first target time and a first preset value to obtain a starting time;

[0028] Calculating the sum of the second target time and the second preset value to obtain an end time;

[0029] The time period corresponding to the start time to the end time is determined as the scene generation time period.

[0030] In one embodiment, the simulation scene generation method further includes:

[0031] The time period corresponding to the manual takeover moment to the second target moment is used as a simulation evaluation time period, and the avoidance simulation scenario is configured according to the simulation evaluation time period.

[0032] In a second aspect, an embodiment of the present application provides a simulation scene generation device, the device comprising:

[0033] An information extraction module, used to obtain accident road test data, and respectively extract vehicle status information, map information and movement information of traffic participants from the accident road test data;

[0034] A manual driving time acquisition module, used to acquire the manual driving time according to the vehicle state information, wherein the manual driving time is the time when the road test vehicle switches to the manual driving state;

[0035] A driving information acquisition module, used to use the vehicle state information and map information corresponding to the road test vehicle in the automatic driving state and the target driving period as the first driving information, and use the vehicle state information and map information corresponding to the road test vehicle in the manual driving state as the second driving information; the end time of the target driving period is the manual driving time;

[0036] A simulation scenario generation module is used to determine whether an avoidance event occurs during the road test according to the first driving information, the second driving information and the movement information of the traffic participant; if so, extract a data segment corresponding to the avoidance event from the accident road test data, and generate an avoidance simulation scenario based on the data segment.

[0037] In a third aspect, an embodiment of the present application provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the simulation scene generation method described in any of the above embodiments.

[0038] In a fourth aspect, an embodiment of the present application provides a computer device, the computer device comprising: one or more processors, and a memory;

[0039] The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the simulation scene generation method described in any of the above embodiments are executed.

[0040] The simulation scene generation method, device, computer equipment and storage medium of the present application can extract vehicle status information, map information and traffic participant movement information from the accident road test data, and obtain the moment when the road test vehicle switches to the manual driving state, that is, the manual driving moment, based on the vehicle status information. Based on the manual driving moment, the present application can use the vehicle status information and map information corresponding to the road test vehicle before it enters the manual driving state as the first driving information, and the vehicle status information and map information corresponding to the road test vehicle after it enters the manual driving state as the second driving information, and automatically determine whether the road test vehicle has an avoidance event during the road test based on the first driving information, the second driving information and the traffic participant movement information. If so, extract the data segment corresponding to the avoidance event from the accident road test data, and generate an avoidance simulation scene accordingly.

[0041] In this way, the automatic identification of avoidance events and the automatic generation of avoidance simulation scenes can be realized based on the movement information of traffic participants and the corresponding driving information before and after the road test vehicle enters the manual driving state, which can greatly reduce the degree of human participation in the scene generation and configuration process, and thus improve the scene conversion efficiency. At the same time, using a stable automated algorithm to construct avoidance simulation scenes can ensure that the recall rate of legal scenes is not affected by human factors and reduce missed judgments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 One of the flowcharts of the simulation scene generation method in one embodiment;

[0044] Figure 2 is a schematic diagram of an avoidance event in one embodiment;

[0045] Figure 3 This is one of the flowchart diagrams of the avoidance event determination step in one embodiment;

[0046] Figure 4 This is a second flow chart of the avoidance event determination step in an embodiment;

[0047] Figure 5 A schematic diagram of a flow chart of steps for generating an avoidance simulation scenario in an embodiment;

[0048] Figure 6 The third flowchart of the simulation scene generation method in one embodiment;

[0049] Figure 7 A schematic diagram of the structure of a simulation scene generating device in an embodiment;

[0050] Figure 8 Schematic diagram of the structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0052] In order to test whether the response strategy of the autonomous driving algorithm is appropriate when facing some scenarios that are prone to collision with other traffic participants (i.e., avoidance simulation scenarios), the simulation system can be used to simulate the avoidance simulation scenario and obtain the control of the vehicle by the autonomous driving algorithm in this scenario to determine whether the algorithm can control the vehicle to avoid the traffic participant and thus avoid the collision. Limited by the amount of data of the complete accident road test data, generally speaking, a simulation scenario will extract the key period from the complete accident road test data for simulation. The simulation scenario includes all the data required to run a complete simulation, including the accident road test data of the corresponding period, the start and end time of the scenario configuration, and the evaluation criteria. Among them, the evaluation criteria are used to verify whether the vehicle's behavior meets or violates certain rules during the simulation process, and output a numerical score.

[0053] At present, when constructing avoidance simulation scenarios, the existing technology generally reports the accident road test data through the safety officer on the car, and the engineer classifies and cuts the accident road test data to obtain the avoidance simulation scenario. In this process, a lot of manpower is required, and it is also inconsistent with the goal of autonomous driving - unmanned driving. Moreover, affected by the understanding of autonomous driving technology by safety officers and engineers, the recall rate of avoidance simulation scenarios is very unstable. With the continuous expansion of the scale of autonomous driving and road testing, the growing demand for manpower, manpower costs and training costs have become bottlenecks in the construction of avoidance simulation scenarios. Therefore, it is urgent to provide a solution that can automatically identify avoidance events based on accident road test data and automatically generate avoidance simulation scenarios based on this, so as to improve the efficiency of scene conversion, reduce the cost of simulation scene construction, and ensure that the recall rate of legal scenes is not affected by human factors, and reduce missed judgments.

[0054] In one embodiment, the present application provides a simulation scene generation method, which can be executed on a computer device, such as a server. Figure 1 As shown, the method comprises the following steps:

[0055] S100, acquiring accident road test data, and extracting vehicle status information, map information, and movement information of traffic participants from the accident road test data.

[0056] The accident road test data refers to all data recorded by the road test vehicle during the road test, including but not limited to surrounding obstacle information, traffic light information, and the road test vehicle's own status information.

[0057] The map information may include the surrounding map information of the road test vehicle and the map information of the traffic participants located around the road test vehicle, which is used to analyze whether the driving behavior of the road test vehicle is correct and reasonable. Among them, traffic participants refer to objects that are not saved on the semantic map, including but not limited to pedestrians, ice cream cones, other vehicles other than the vehicle, and other obstacles. In one example, the map information may include lane position, lane line type, lane line position, and the content of traffic signs used to assist in traffic management, the location of stop lines, the status of traffic lights, the location of construction ice cream cones, the content and location of warning signs, etc.

[0058] In one embodiment, the computer device may scan the periphery of the road test vehicle during the road test period according to the semantic map elements to obtain the map information of the periphery of the road test vehicle and the map information of the traffic participants located around the road test vehicle. For example, when the road test vehicle is currently traveling in lane l, the computer device may attempt to search for and record the left and right adjacent lanes l of lane l. l and l r , and record the lane information of the left and right lanes (such as the type and length of the lanes) so as to determine whether lane change to the left lane and / or the right lane is allowed according to the lane information. The computer device can also record the source lane set of the lane {l f} and the accessible lanes {l t}, so as to facilitate the subsequent judgment of the merging and diverging of lanes. The computer equipment can also record the content of the sign corresponding to each lane, the status of the traffic light, the location of the stop line passed, the location of the ice cream cone, etc. In this way, by recording the lane information of the road test vehicles and the locations of different traffic participants at different time periods, map information can be obtained.

[0059] The vehicle status information may be information used to reflect the driving status of the road test vehicle at each moment, including but not limited to the vehicle coordinate position of the road test vehicle at each moment, the driving safety score, the current driving status of the road test vehicle, and other information. In one embodiment, the vehicle status information may include abnormal status information, and the abnormal status information may be status information used to reflect whether the road test vehicle has a driving abnormality, so as to assist in screening and classifying the accident road test data. Furthermore, the abnormal status information may include information used to reflect that the road test vehicle is in a manual driving state, information used to reflect that the road test vehicle brakes suddenly, information used to reflect that the road test vehicle turns suddenly, information used to reflect that a system error occurs inside the road test vehicle, and information used to reflect that there is a high transmission delay inside the road test vehicle.

[0060] The motion information of traffic participants may include the location information, posture information, speed information, acceleration information and trajectory information of traffic participants, which is used to analyze the characteristics and behavioral goals of traffic participants and set reasonable action goals for the road test vehicle accordingly. In one embodiment, in order to reduce the amount of data processing to further improve the efficiency of scene conversion, the present application can extract the motion information of key traffic participants from the accident road test data. Among them, key traffic participants refer to traffic participants who are too close to the road test vehicle at a certain moment, and / or traffic participants whose trajectories will intersect with the road test vehicle in a short period of time.

[0061] In the embodiment of the present application, the computer device can extract map information, vehicle status information and movement information of traffic participants from the complete accident road test data, and perform steps S200, S300 and S400 based on the extracted information. In this way, redundant information in the accident road test data can be removed, and the amount of processed data can be compressed to 1 / 1000 of the accident road test data, thereby improving the operating efficiency of the computer device, reducing resource usage, and further accelerating the efficiency of scene conversion.

[0062] S200, obtaining a manual driving time according to the vehicle state information, where the manual driving time is the time when the road test vehicle switches to a manual driving state.

[0063] During the road test, the road test vehicle can be in an automatic driving state or a manual driving state. When in the automatic driving state, the road test vehicle can travel under the control of the automatic driving algorithm; when in the manual driving state, the road test vehicle can be manually controlled by the safety officer. If the road test vehicle has a manual driving process during the road test, the computer device can obtain the time when the road test vehicle switches to the manual driving state according to the vehicle state information, and use this as the manual driving time. That is, the manual driving time is the time when the road test vehicle enters the manual driving state. In one example, the manual driving time can be the time when the safety officer takes over the road test vehicle.

[0064] S300, using the vehicle state information and map information corresponding to when the road test vehicle is in the automatic driving state and the target driving period as the first driving information, and using the vehicle state information and map information corresponding to when the road test vehicle is in the manual driving state as the second driving information; wherein the end time of the target driving period is the manual driving time.

[0065] In other words, if the manual driving time is t h , then the target driving period can be [t a , t h ], where t a Before t hIt is understood that the specific duration and / or starting time of the target driving period can be determined according to actual conditions, and this application does not impose any specific restrictions on this.

[0066] The computer device may use the vehicle state information and map information that simultaneously meet the following two conditions as the first driving information: (1) the vehicle state information corresponding to the road test vehicle when it is in the automatic driving state, and the map information corresponding to the road test vehicle when it is in the automatic driving state; (2) the road test vehicle is in the target driving period [t a , t h ], and the vehicle status information corresponding to the target driving period [t a , t h ] corresponding to the map information. That is, the first driving information includes the road test vehicle in the automatic driving state and the target driving period [t a , t h ], and the vehicle status information corresponding to the time when the road test vehicle is in the automatic driving state and the target driving period [t a , t h ] corresponding map information.

[0067] The computer device may also obtain the vehicle state information and map information corresponding to the road test vehicle in the manual driving state, and use this as the second driving information. In one embodiment, the computer device may use the manual driving moment as the starting moment of the manual driving period, and use the vehicle state information and map information corresponding to the manual driving period as the second driving information. It is understood that the specific duration of the manual driving period can be determined according to actual conditions, for example, it can be determined according to the actual manual driving time or a preset time threshold, and this application does not impose specific restrictions on this.

[0068] S400, judging whether an avoidance event occurs during the road test according to the first driving information, the second driving information and the movement information of the traffic participant; if so, extracting a data segment corresponding to the avoidance event from the accident road test data, and generating an avoidance simulation scenario based on the data segment.

[0069] The avoidance event refers to an event in which the driving parameters of the road test vehicle (such as driving direction, driving trajectory and / or driving speed) are adjusted so that the road test vehicle is further away from the same traffic participant after the adjustment, or the road test vehicle takes a longer time to reach the location of the same traffic participant after the adjustment. Figure 2 shown.

[0070] When the computer device obtains the driving information of the road test vehicle before entering the manual driving state (i.e., the first driving information), the driving information of the road test vehicle after entering the manual driving state (i.e., the second driving information), and the motion information of traffic participants, it can automatically determine whether the automatic driving algorithm has taken a behavior that is not satisfactory to the safety officer based on the driving information before and after entering the manual driving state and the motion information of traffic participants, so that the safety officer manually takes over the road test vehicle in order to improve driving safety, and controls the road test vehicle to avoid dangerous objects, that is, automatically determines whether an avoidance event occurs during the road test of the road test vehicle. If an avoidance event occurs during the road test, a data segment corresponding to the avoidance event can be extracted from the accident road test data, and an avoidance simulation scene can be generated based on the data segment.

[0071] In one embodiment, the step of extracting data segments corresponding to the avoidance event from the accident road test data and generating the avoidance simulation scene based on the data segments may include: extracting map information segments, vehicle status information segments, and traffic participant motion information segments corresponding to the avoidance event, and generating the avoidance simulation scene based on the aforementioned three information segments. The map information, vehicle status information, and traffic participant motion information not only retain the data necessary for making the avoidance simulation scene, but also compress the data volume. Therefore, generating the avoidance simulation scene based on the aforementioned three information segments can perform simulation testing more quickly to improve test efficiency.

[0072] In the embodiment of the present application, the automatic identification of avoidance events and the automatic generation of avoidance simulation scenes can be realized based on the motion information of traffic participants and the driving information corresponding to the road test vehicle before and after entering the manual driving state, thereby greatly reducing the degree of human participation in the scene generation and configuration process, thereby improving the scene conversion efficiency. At the same time, using a stable automated algorithm to construct the avoidance simulation scene can ensure that the recall rate of legal scenes is not affected by human factors and reduce missed judgments.

[0073] In one embodiment, Figure 3 As shown, the step of determining whether an avoidance event occurs during the road test according to the first driving information, the second driving information and the motion information of the traffic participant includes:

[0074] S410: acquiring a predicted driving trajectory of the road test vehicle based on the first driving information, and acquiring a predicted motion trajectory of the traffic participant based on the motion information of the traffic participant.

[0075] Since the first driving information is the vehicle state information and map information corresponding to the road test vehicle in the automatic driving state and the target driving period, the predicted driving trajectory of the road test vehicle obtained by the computer device based on the first driving information may be the expected driving trajectory from the current moment to a certain moment in the future planned by the automatic driving algorithm for the road test vehicle. It should be noted that the "sometime in the future" may be later than the manual driving moment. In other words, the predicted driving trajectory of the road test vehicle may include the predicted driving trajectory corresponding to the target driving period and the predicted driving trajectory corresponding to the manual driving period.

[0076] The predicted movement trajectory of the traffic participant may include the movement trajectory of the traffic participant from the current moment to a certain moment in the future, which is inferred. The aforementioned "a certain moment in the future" may be as described above, and will not be repeated here. The predicted movement trajectory of the traffic participant may include the predicted movement trajectory corresponding to the target driving period and the predicted movement trajectory corresponding to the manual driving period.

[0077] Specifically, the computer device can obtain the predicted motion trajectory of the traffic participant based on the motion information of the traffic participant. In one embodiment, the computer device can obtain the instantaneous speed and acceleration of the traffic participant at each moment based on the motion information of the traffic participant, and can predict the motion trajectory of the traffic participant based on the instantaneous speed and acceleration at multiple consecutive moments.

[0078] When there are multiple traffic participants, the computer device may obtain the predicted motion trajectory of each traffic participant. In one embodiment, in order to reduce the amount of data processing of the computer device and further improve the efficiency of scene conversion, the computer device may determine the key traffic participants from the multiple traffic participants based on the distance between each traffic participant and the road test vehicle or the motion information of each traffic participant, and obtain the predicted motion trajectory of each key traffic participant.

[0079] S420, determining whether the predicted motion trajectory of the traffic participant intersects with the predicted driving trajectory of the road test vehicle, and obtaining an intersection determination result.

[0080] The "intersection" mentioned herein can be understood as the distance between the road test vehicle and the traffic participant being less than a preset distance threshold, and the specific value of the preset distance threshold can be determined according to the actual situation, and this application does not impose specific restrictions on this. The computer device can determine whether the road test vehicle controlled by the autonomous driving algorithm will overlap with or be too close to the position of the traffic participant at some point in the future based on the predicted motion trajectory of the traffic participant and the predicted driving trajectory of the road test vehicle.

[0081] S430: Based on the intersection determination result and the second driving information, determine whether an avoidance event occurs during the road test.

[0082] In one of the embodiments, the computer device may obtain information about obstacles located ahead of the road test vehicle in the driving direction when the road test vehicle is in a manual driving state based on the second driving information (hereinafter referred to as the front obstacle information), and combine the front obstacle information and the intersection judgment result to determine whether an avoidance event occurs during the road test.

[0083] In this embodiment, by predicting the motion trajectory of the traffic participant and the driving trajectory of the road test vehicle, it is possible to determine whether the trajectories of the traffic participant and the road test vehicle have intersected, or whether the trajectories are likely to intersect in a short period of time, based on the predicted motion trajectory of the traffic participant and the predicted driving trajectory of the road test vehicle. Based on the intersection judgment result and the second driving information, it is determined whether an avoidance event occurs during the road test, thereby improving the recognition accuracy of the avoidance event.

[0084] In one embodiment, Figure 4 As shown, the step of determining whether the avoidance event occurs during the road test based on the intersection determination result and the second driving information includes:

[0085] S432: Based on the second driving information, determine whether the road test vehicle has an evasive behavior when it is in a manual driving state, and obtain an evasive behavior determination result.

[0086] Among them, avoidance behavior refers to behavior that can increase the distance between the predicted motion trajectory of the traffic participant and the driving trajectory of the road test vehicle, or behavior that can delay the intersection time of the predicted motion trajectory of the traffic participant and the predicted driving trajectory of the road test vehicle. That is, compared with the situation where the autonomous driving algorithm controls the driving of the road test vehicle, after the safety officer takes over the road test vehicle, he increases the distance between the predicted motion trajectory of the traffic participant and the driving trajectory of the road test vehicle, or delays the intersection time of the predicted motion trajectory of the traffic participant and the predicted driving trajectory of the road test vehicle by taking specific driving behaviors.

[0087] If the road test vehicle is continuously controlled by the autonomous driving algorithm, the road test vehicle will continue to travel according to the predicted driving trajectory. Therefore, the relative position relationship or intersection time between the road test vehicle controlled by the autonomous driving algorithm and the traffic participant can be estimated based on the predicted driving trajectory. At this time, the minimum distance between the predicted driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant is the first distance. If the predicted driving trajectory of the road test vehicle intersects with the predicted motion trajectory of the traffic participant, the intersection time is the first time.

[0088] After the safety officer takes over the road test vehicle, the road test vehicle will drive under the control of the safety officer. At this time, the minimum distance between the actual driving trajectory of the road test vehicle under manual driving and the predicted motion trajectory of the traffic participant is the second distance. If the actual driving trajectory intersects with the predicted motion trajectory of the traffic participant, the intersection time is the second time.

[0089] The evasive action refers to an action that can make the second distance greater than the first distance, or an action that can make the second moment later than the first moment. In one example, the evasive action includes but is not limited to braking and avoiding to the side.

[0090] S434: Based on the avoidance judgment result and the intersection judgment result, determine whether an avoidance event occurs during the road test.

[0091] In one of the embodiments, if the road test vehicle has evasive behavior when it is in a manual driving state, and the predicted motion trajectory of the traffic participant intersects with the predicted driving trajectory of the road test vehicle, it can be determined that an evasive event occurs during the road test.

[0092] In this embodiment, by judging whether there is any evasive behavior when the road test vehicle is in a manual driving state, and judging whether an avoidance event occurs during the road test according to the avoidance judgment result and the intersection judgment result, compared with judging according to the obstacle information ahead, this embodiment can reduce the amount of data processed to further improve the scene conversion efficiency.

[0093] In one embodiment, the step of determining whether an avoidance event occurs during a road test based on the first driving information, the second driving information, and the movement information of traffic participants further includes: obtaining a first driving safety score of the road test vehicle in an automatic driving state based on the first driving information; obtaining a second driving safety score of the road test vehicle in a manual driving state based on the second driving information, and comparing the second driving safety score with the first driving safety score to obtain a comparison result.

[0094] The step of judging whether an avoidance event occurs during the road test based on the avoidance judgment result and the intersection judgment result includes judging whether an avoidance event occurs during the road test based on the avoidance judgment result, the intersection judgment result and the comparison result.

[0095] The driving safety score can be used to reflect the degree of danger of the road test vehicle when driving on the road. The lower the driving safety score, the greater the collision probability. In one embodiment, the deep learning model can be trained with artificially labeled dangerous scene data, so that the trained deep learning model can comprehensively consider the position, speed direction and lane direction of the traffic participants, predict the collision probability between the road test vehicle and the traffic participants, and score different moments according to the collision probability to output the driving safety scores corresponding to different moments.

[0096] The computer device can comprehensively judge whether the road test vehicle has an avoidance event during the road test according to the avoidance judgment result, the intersection judgment result, and the comparison result of the driving safety score before and after the manual takeover. In one embodiment, the computer device can judge whether the following conditions are met according to the first driving information, the second driving information, and the movement information of the traffic participants:

[0097] (1) The road test vehicle is taken over by the safety officer (i.e., the road test vehicle switches from the autonomous driving state to the manual driving state). If the safety officer takes over the road test vehicle, it means that the autonomous driving algorithm has adopted a dangerous driving behavior that the safety officer is dissatisfied with, or the road test vehicle faces a lot of uncertainty in the scenario;

[0098] (2) Before the safety officer takes over, the predicted driving trajectory of the road test vehicle intersects with the predicted movement trajectory of at least one traffic participant, and at least one traffic participant is located on the predicted driving trajectory of the road test vehicle. If the vehicle is driven according to the originally planned predicted driving trajectory, a dangerous situation will occur in a short period of time;

[0099] (3) The safety officer completes at least one emergency evasive action after taking over, including but not limited to braking and swerving, so that the actual driving trajectory of the autonomous vehicle after the safety officer takes over is farther away from the traffic participants or intersects with the traffic participants at a later time than before the safety officer takes over, or the extended speed direction of the autonomous vehicle after the safety officer takes over is farther away from the traffic participants or intersects with the traffic participants at a later time than before the safety officer takes over;

[0100] (4) The driving safety score before the safety officer takes over (i.e., the first driving safety score) is lower than the driving safety score after the safety officer takes over (i.e., the second driving safety score).

[0101] If all of the above four conditions are met, it can be determined that an avoidance event occurred during the road test, that is, an accident occurred during the road test in which the avoidance of a suddenly appearing dangerous vehicle was not handled sufficiently.

[0102] In this embodiment, by comparing the driving scores before and after manual takeover, the rationality of the safety officer's driving behavior can be determined through the comparison result to improve the recognition accuracy of avoidance events.

[0103] In one embodiment, Figure 5 As shown, the steps of extracting data segments corresponding to avoidance events from accident road test data and generating avoidance simulation scenarios according to the data segments include:

[0104] S440, taking the time corresponding to the intersection of the predicted driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant as the first target time;

[0105] S450, obtaining a second target time according to the second driving information, and determining a scene generation period based on the first target time and the second target time; the second target time is the earliest time corresponding to when the traveling direction of the road test vehicle does not intersect with the position of the traffic participant;

[0106] S460, extracting data segments corresponding to the scenario generation period from the accident road test data, and generating an avoidance simulation scenario according to the data segments.

[0107] The computer device can mark the key moments, and obtain the scene generation period according to the marked key moments, so as to generate an avoidance simulation scene based on the scene generation period and the accident road test data. Specifically, the computer device uses the moment corresponding to the intersection of the predicted driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant as the first target moment, and can judge whether the trajectory of the road test vehicle has avoided the traffic participant according to the direction of travel of the road test vehicle and the position of the traffic participant. When the direction of travel of the road test vehicle does not intersect with the position of the traffic participant, it can be determined that the trajectory of the road test vehicle has avoided the traffic participant, so the moment is marked as the second target moment. It should be noted that since the second driving information includes the vehicle status information and map information corresponding to when the road test vehicle is in a manual driving state, the second target moment is the earliest moment corresponding to when the direction of travel of the road test vehicle does not intersect with the position of the traffic participant when it is in a manual driving state. For example, at Figure 2 In the example of , the manual driving time is the time corresponding to when the road test vehicle travels to the position L1, and the second target time may be the time corresponding to when the road test vehicle travels to the position L2.

[0108] After obtaining the first target moment and the second target moment, the computer device can determine the scene generation period according to the first target moment and the second target moment, and extract the data segment corresponding to the scene generation period from the accident road test data to generate an avoidance simulation scene according to the data segment. In this way, it can be ensured that the data segment covers the entire occurrence period of the avoidance event, and when testing through the avoidance simulation scene, the autonomous driving algorithm can be comprehensively and accurately tested. At the same time, standardized scenes can be generated using standardized algorithms and scene production processes, which can enable the autonomous driving algorithm to obtain more accurate and stable evaluation criteria.

[0109] In one embodiment, the step of determining a scene generation period based on a first target moment and a second target moment includes: calculating the difference between the first target moment and a first preset value to obtain a start moment; calculating the sum of the second target moment and a second preset value to obtain an end moment; and determining the period corresponding to the start moment to the end moment as the scene generation period.

[0110] Among them, the specific numerical value of the first preset value can be determined according to the actual situation, so as to adjust the starting time of the scene generation period to the latest time point earlier than the first target time when the road test vehicle is running smoothly according to the characteristics of the scene. The specific numerical value of the second preset value can be determined according to the actual situation to determine a suitable time required to eliminate the danger and resume normal driving. In this embodiment, the starting time of the scene generation period is earlier than the first target time, and the ending time of the scene generation period is later than the second target time, so as to reserve adjustment time for the automatic driving algorithm to be tested during the simulation test, so that it can be adjusted to a stable running state before the first target time, so as to improve the accuracy of the test results.

[0111] In one example, the computer device may record the moment T0 (i.e., the first target moment) when the traffic participant appears on the predicted driving trajectory of the road test vehicle, the moment T1 (i.e., the manual takeover moment) when the safety officer takes over the road test vehicle, and the moment T2 (i.e., the second target moment) when the safety officer takes action and makes the trajectory of the road test vehicle avoid the traffic participant. The above key moments can all be obtained in the process of the computer device determining whether an avoidance event occurs during the road test. In this example, the first preset value is 2s and the second preset value is 4s. Therefore, the starting moment Ts of the scene generation period can be (T0-2s), and the ending moment Te of the scene generation period can be (T2+4s).

[0112] In one embodiment, the simulation scenario generation method further includes configuring a main vehicle behavior indication to ensure that the avoidance event can be reproduced. In order to ensure that the driving route of the vehicle controlled by the autonomous driving algorithm is consistent with the driving route of the road test vehicle before the avoidance event occurs, during the simulation test, the driving direction of the vehicle before the key traffic participant appears can be locked through remote control instructions.

[0113] In one embodiment, the simulation scenario generation method further includes: using the period corresponding to the manual takeover time to the second target time as a simulation evaluation period, and configuring the avoidance simulation scenario according to the simulation evaluation period. The simulation evaluation period is used to test whether the avoidance reaction time of the automatic driving algorithm for key traffic participants is less than the manual reaction time.

[0114] In one embodiment, the computer device may also mark the expected driving behavior so as to configure a suitable evaluation index using the expected driving behavior. Specifically, the computer device may divide the vehicle driving behavior into two sub-behaviors according to the appearance time of the key traffic participant. One of the sub-behaviors is the driving behavior before the appearance of the key traffic participant, and the other sub-behavior is the driving behavior after the appearance of the key traffic participant. The computer device may evaluate whether the vehicle controlled by the automatic driving algorithm to be tested decelerates and avoids the suspicious key traffic participant in advance and whether the vehicle controlled by the automatic driving algorithm to be tested handles and resolves the danger in time when the dangerous situation occurs according to the two sub-behaviors. During the simulation test, the manual driving moment and the second target moment may be used as the target reaction time and the processing completion time respectively, and the vehicle controlled by the automatic driving algorithm to be tested may complete the avoidance within the period between the manual driving moment and the second target moment. Further, the computer device may configure a target safety score that is not lower than the second driving safety score for the avoidance simulation scenario. During the simulation test, if the driving safety scores of the vehicles controlled by the automatic driving algorithm to be tested are all higher than the target safety score, it can be confirmed that the automatic driving algorithm to be tested can safely deal with the avoidance simulation scenario.

[0115] In one embodiment, Figure 6 As shown, after completing the configuration of the avoidance simulation scene, the scene configuration can be verified. Specifically, the computer equipment can use a large-scale simulation platform to regularly batch run the newly generated simulation scene, use the same version of the autonomous driving algorithm as when the accident (i.e., the avoidance event) occurred, and use algorithms such as trajectory similarity to detect whether the problem in the original accident will be reproduced in the simulation, that is, the simulation result is consistent with the driving trajectory that occurred during the road test. If the scene can be reproduced, it is further verified whether the same version of the autonomous driving algorithm when the accident occurred can indeed not achieve the expected reasonable evaluation index. The above verification results automatically generate a report and enter the scene library. The scene library is used to simulate each version of the autonomous driving algorithm before releasing each version of the autonomous driving algorithm, so as to analyze and evaluate the processing ability of the autonomous driving algorithm to be tested for this type of scene based on the simulation results. In this way, by screening the accident road test data into the scene library and quickly performing simulation tests, the iteration of the autonomous driving algorithm can be more firmly guaranteed, avoiding the negative impact of untested new algorithms on performance, affecting the working environment of the safety officer or the development of new algorithms, and thus more efficiently developing autonomous driving algorithms with better performance and protecting the safe working environment of the safety officer.

[0116] The simulation scene generation device 500 provided in an embodiment of the present application is described below. The simulation scene generation device 500 described below and the simulation scene generation method described above can be referenced to each other.

[0117] In one embodiment, the present application provides a simulation scene generation device 500, such as Figure 7 As shown, the device includes an information extraction module 510, a manual driving time acquisition module 520, a driving information acquisition module 530 and a simulation scene generation module 540. Among them:

[0118] The information extraction module 510 is used to obtain the accident road test data, and respectively extract the vehicle status information, the map information and the movement information of the traffic participants from the accident road test data;

[0119] A manual driving time acquisition module 520 is used to acquire a manual driving time according to the vehicle state information, wherein the manual driving time is the time when the road test vehicle switches to a manual driving state;

[0120] The driving information acquisition module 530 is used to use the vehicle state information and map information corresponding to the road test vehicle in the automatic driving state and the target driving period as the first driving information, and use the vehicle state information and map information corresponding to the road test vehicle in the manual driving state as the second driving information; the end time of the target driving period is the manual driving time;

[0121] The simulation scenario generation module 540 is used to determine whether an avoidance event occurs during the road test based on the first driving information, the second driving information and the movement information of the traffic participant; if so, extract a data segment corresponding to the avoidance event from the accident road test data, and generate an avoidance simulation scenario based on the data segment.

[0122] In one embodiment, the simulation scene generation module 540 includes a trajectory prediction unit, an intersection judgment unit, and an event judgment unit. The trajectory prediction unit is used to obtain the predicted driving trajectory of the road test vehicle based on the first driving information, and to obtain the predicted motion trajectory of the traffic participant based on the motion information of the traffic participant. The intersection judgment unit is used to determine whether the predicted motion trajectory of the traffic participant intersects with the predicted driving trajectory of the road test vehicle, and obtain an intersection judgment result. The first event judgment unit is used to determine whether the avoidance event occurs during the road test based on the intersection judgment result and the second driving information.

[0123] In one embodiment, the event judgment unit includes an avoidance judgment unit. The avoidance judgment unit is used to judge whether there is an avoidance behavior when the road test vehicle is in the manual driving state according to the second driving information, and obtain an avoidance judgment result. The avoidance behavior is a behavior that can increase the distance between the driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant, or the avoidance behavior is a behavior that can delay the intersection time between the driving trajectory of the road test vehicle and the predicted motion trajectory. The avoidance judgment unit is also used to judge whether the avoidance event occurs during the road test process based on the avoidance judgment result and the intersection judgment result.

[0124] In one embodiment, the second event acquisition unit includes a score comparison unit. The score comparison unit is used to obtain a first driving safety score of the road test vehicle in the automatic driving state according to the first driving information, and to obtain a second driving safety score of the road test vehicle in the manual driving state according to the second driving information, and to compare the second driving safety score with the first driving safety score to obtain a comparison result. The score comparison unit is also used to determine whether the avoidance event occurs during the road test based on the avoidance judgment result, the intersection judgment result and the comparison result.

[0125] In one embodiment, the scenario generation module includes a first target moment acquisition unit, a second target moment acquisition unit and a scenario generation unit. The first target moment acquisition unit is used to take the moment corresponding to the intersection of the predicted driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant as the first target moment. The second target moment acquisition unit is used to acquire the second target moment according to the second driving information, and determine the scenario generation period based on the first target moment and the second target moment; the second target moment is the earliest moment corresponding to the time when the traveling direction of the road test vehicle does not intersect with the position of the traffic participant. The scenario generation unit is used to extract the data segment corresponding to the scenario generation period from the accident road test data, and generate an avoidance simulation scenario based on the data segment.

[0126] In one embodiment, the scene generation unit includes a start time calculation unit, an end time calculation unit and a scene generation period determination unit. The start time calculation unit is used to calculate the difference between the first target time and the first preset value to obtain the start time. The end time calculation unit is used to calculate the sum of the second target time and the second preset value to obtain the end time. The scene generation period determination unit is used to determine the period corresponding to the start time to the end time as the scene generation period.

[0127] In one embodiment, the device further comprises a configuration module. The configuration module is used to use the period corresponding to the manual takeover time to the second target time as a simulation evaluation period, and configure the avoidance simulation scenario according to the simulation evaluation period.

[0128] In one embodiment, the present application also provides a storage medium, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the simulation scene generation method described in any of the above embodiments.

[0129] In one embodiment, the present application also provides a computer device having computer-readable instructions stored therein. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the simulation scene generation method described in any one of the above embodiments.

[0130] Indicatively, Figure 8 As shown, Figure 8 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. The computer device 900 may be provided as a server. Figure 8 The computer device 900 includes a processing component 902, which further includes one or more processors, and a memory resource represented by a memory 901 for storing instructions executable by the processing component 902, such as an application. The application stored in the memory 901 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 902 is configured to execute the instructions to perform the text recognition method of any of the above embodiments.

[0131] The computer device 900 may further include a power supply component 903 configured to perform power management of the computer device 900, a wired or wireless network interface 904 configured to connect the computer device 900 to a network, and an input / output (I / O) interface 905. The computer device 900 may operate based on an operating system stored in the memory 901, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM or the like.

[0132] Those skilled in the art will understand that the internal structure of the computer device shown in the present application is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0133] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not clearly listed, or also includes elements inherent to such process, method, article or equipment. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements. Herein, "one", "one", "said", "the" and "it" may also include plural forms, unless the context clearly indicates another way. A plurality refers to at least two cases, such as 2, 3, 5 or 8, etc. "And / or" includes any and all combinations of the relevant listed items.

[0134] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

[0135] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A simulation scene generation method, characterized in that: The method comprises: Acquire accident road test data, and respectively extract vehicle status information, map information, and movement information of traffic participants from the accident road test data; Acquire a manual driving time according to the vehicle state information, where the manual driving time is the time when the road test vehicle switches to a manual driving state; The vehicle state information and map information corresponding to the road test vehicle in the automatic driving state and the target driving period are used as the first driving information, and the vehicle state information and map information corresponding to the road test vehicle in the manual driving state are used as the second driving information; the end time of the target driving period is the manual driving time; According to the first driving information, the second driving information and the motion information of the traffic participant, determining whether an avoidance event occurs during the road test; if so, extracting a data segment corresponding to the avoidance event from the accident road test data, and generating an avoidance simulation scene based on the data segment; The step of determining whether an avoidance event occurs during the road test according to the first driving information, the second driving information and the movement information of the traffic participant includes: Acquire a predicted driving trajectory of the road test vehicle based on the first driving information, and acquire a predicted motion trajectory of a traffic participant based on the motion information of the traffic participant; Determine whether the predicted motion trajectory of the traffic participant intersects with the predicted driving trajectory of the road test vehicle, and obtain an intersection determination result; Based on the second driving information, determining whether the road test vehicle has an evasive behavior when in the manual driving state, and obtaining an evasive behavior determination result; wherein the evasive behavior is a behavior that can increase the distance between the driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant, or the evasive behavior is a behavior that can delay the intersection time between the driving trajectory of the road test vehicle and the predicted motion trajectory; Based on the avoidance judgment result and the intersection judgment result, it is determined whether the avoidance event occurs during the road test.

2. The simulation scene generation method according to claim 1, characterized in that: The step of determining whether an avoidance event occurs during the road test according to the first driving information, the second driving information, and the motion information of the traffic participant further includes: Acquire a first driving safety score of the road test vehicle in the automatic driving state according to the first driving information; Acquire a second driving safety score of the road test vehicle in the manual driving state according to the second driving information, and compare the second driving safety score with the first driving safety score to obtain a comparison result; The step of determining whether the avoidance event occurs during the drive test based on the avoidance determination result and the intersection determination result includes: Based on the avoidance judgment result, the intersection judgment result and the comparison result, it is judged whether the avoidance event occurs during the road test.

3. The simulation scene generation method according to any one of claims 1 to 2, characterized in that: The step of extracting a data segment corresponding to the avoidance event from the accident road test data and generating an avoidance simulation scenario according to the data segment comprises: The time corresponding to the intersection of the predicted driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant is taken as the first target time; Acquire a second target time according to the second driving information, and determine a scenario generation period based on the first target time and the second target time; the second target time is the earliest time corresponding to when the traveling direction of the road test vehicle does not intersect with the position of the traffic participant; A data segment corresponding to the scenario generation period is extracted from the accident road test data, and an avoidance simulation scenario is generated according to the data segment.

4. The simulation scene generation method according to claim 3, characterized in that: The step of determining a scene generation period based on the first target moment and the second target moment includes: Calculating the difference between the first target time and a first preset value to obtain a starting time; Calculating the sum of the second target time and the second preset value to obtain an end time; The time period corresponding to the start time to the end time is determined as the scene generation time period.

5. The simulation scene generation method according to claim 3, characterized in that: The method further comprises: The time period corresponding to the manual takeover time to the second target time is used as a simulation evaluation time period, and the avoidance simulation scenario is configured according to the simulation evaluation time period.

6. A simulation scene generation device, characterized in that: The device comprises: An information extraction module, used to obtain accident road test data, and respectively extract vehicle status information, map information and movement information of traffic participants from the accident road test data; A manual driving time acquisition module, used to acquire the manual driving time according to the vehicle state information, wherein the manual driving time is the time when the road test vehicle switches to the manual driving state; A driving information acquisition module, used to use the vehicle state information and map information corresponding to the road test vehicle in the automatic driving state and the target driving period as the first driving information, and use the vehicle state information and map information corresponding to the road test vehicle in the manual driving state as the second driving information; the end time of the target driving period is the manual driving time; a simulation scenario generating module, configured to determine whether an avoidance event occurs during the road test according to the first driving information, the second driving information and the motion information of the traffic participant; if so, extracting a data segment corresponding to the avoidance event from the accident road test data, and generating an avoidance simulation scenario based on the data segment; Wherein, the simulation scene generation module includes: a trajectory prediction unit, configured to obtain a predicted driving trajectory of the road test vehicle based on the first driving information, and to obtain a predicted motion trajectory of a traffic participant based on the motion information of the traffic participant; An intersection judgment unit, used to judge whether the predicted motion trajectory of the traffic participant intersects with the predicted driving trajectory of the road test vehicle, and obtain an intersection judgment result; A first event determination unit, configured to determine whether the avoidance event occurs during the road test based on the intersection determination result and the second driving information; An avoidance judgment unit is used to judge whether there is an avoidance behavior when the road test vehicle is in the manual driving state according to the second driving information, and obtain an avoidance judgment result; and is also used to judge whether the avoidance event occurs during the road test based on the avoidance judgment result and the intersection judgment result; wherein the avoidance behavior is a behavior that can increase the distance between the driving trajectory of the road test vehicle and the predicted motion trajectory of the traffic participant, or the avoidance behavior is a behavior that can delay the intersection time between the driving trajectory of the road test vehicle and the predicted motion trajectory.

7. A storage medium, characterized in that: The storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the simulation scene generation method according to any one of claims 1 to 5.

8. A computer device, characterized in that: include: one or more processors, and memory; The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the simulation scene generation method according to any one of claims 1 to 5 are performed.

Images