Method, apparatus, device and storage medium for vehicle simulation test

Abstract

The present disclosure provides a vehicle simulation test method, device, equipment and storage medium, and relates to the field of automobiles.In the present disclosure, based on accident-related data, an accident scene is established, and then a test vehicle is used as a bystander vehicle in the accident scene, that is, a vehicle other than an accident vehicle, so as to verify the safety of an automatic driving program.In detail, simulated sensor data is generated based on the accident-related data and position data and motion state data of the test vehicle, and a decision is made based on the simulated sensor data.In this way, in the simulation scene, the test vehicle can be set as not only the accident vehicle but also any bystander vehicle, so that the simulation scene is fully utilized and the waste of processing resources is reduced.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of automobiles, and in particular, to a vehicle simulation test method, device, equipment and storage medium. BACKGROUND

[0002] With the development of computer technology, vehicle automatic driving technology emerges as the times require. Before the automatic driving technology is put into use, a large number of tests need to be carried out, so that when the vehicle uses the automatic driving technology, it can perceive and respond to various unexpected situations in advance, thereby ensuring the safety of the vehicle and passengers. The test of the automatic driving function can be completed on a computer, which can reduce the cost and risk of the test on the one hand, and freely simulate various scenes on the other hand, so that the test is more detailed and comprehensive.

[0003] In the related art, a technician can obtain a video, a photo and various sensor data of an accident vehicle of a traffic accident scene, and construct a simulation scene based on the data. In this scene, various sensor data of any accident vehicle can be input into an automatic driving vehicle (which can be referred to as a test vehicle) to verify the response of the test vehicle. Based on the response, the technician adjusts and improves the automatic driving function.

[0004] It takes a large amount of processing resources to construct a simulation scene based on an actual traffic accident. After the simulation scene is established, the application of the simulation scene is relatively narrow by only replacing the accident vehicle with the test vehicle, and the simulation scene is not fully utilized, resulting in waste of processing resources. SUMMARY

[0005] The embodiments of the present disclosure provide a vehicle simulation test method, device, equipment and storage medium, which can solve the technical problems in the related art. The technical solutions are as follows:

[0006] In a first aspect, the embodiments of the present disclosure provide a vehicle simulation test method, and the method comprises:

[0007] Obtain accident-related data in a time range and a position range corresponding to a target traffic accident, wherein the accident-related data includes camera video data and sensor data of at least one vehicle, and the at least one vehicle includes an accident vehicle;

[0008] Based on the accident-related data, establish accident scene data, wherein the accident scene data includes static environment data and dynamic object data, the dynamic object data includes position data and motion state data of at least one dynamic object in a target time period, and the at least one dynamic object includes the at least one vehicle, wherein the target time period is within the time range;

[0009] determine a first decision time point within the target time period and determine position data and motion state data of the test vehicle within a first specified time period before the first decision time point, wherein the test vehicle is a vehicle other than the accident vehicle;

[0010] generate simulated sensor data of the test vehicle within a second specified time period before the first decision time point based on position data of a fisheye monitoring device corresponding to the fisheye video data, accident-related data within the first specified time period before the first decision time point, and the position data and the motion state data of the test vehicle within the first specified time period before the first decision time point, wherein the simulated sensor data within the second specified time period before the first decision time point includes simulated sensor data corresponding to a plurality of time points within the second specified time period, the plurality of time points are uniformly distributed within the second specified time period and include the first decision time point, and the first specified time period is greater than the second specified time period;

[0011] process the simulated sensor data of the test vehicle within the second specified time period based on the automatic driving program of the test vehicle to obtain first decision information.

[0012] In a possible implementation, the sensor data includes vehicle video data, vehicle radar data, and motion state data.

[0013] In a possible implementation, the motion state data includes vehicle speed and vehicle steering angle.

[0014] In a possible implementation, the determining of the position data and the motion state data of the test vehicle within the first specified time period before the first decision time point includes:

[0015] receiving a test vehicle addition request;

[0016] determining the position data and the motion state data of the test vehicle within the first specified time period before the first decision time point carried in the test vehicle addition request.

[0017] In a possible implementation, the interval time length of adjacent two time points is equal to the decision cycle time length of the automatic driving program, and after the processing of the simulated sensor data of the test vehicle within the second specified time period based on the automatic driving program of the test vehicle to obtain the first decision information, the method further includes:

[0018] determining, based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, position data and motion state data of the test vehicle at a second decision time point, the second decision time point being after the first decision time point, and an interval time length between the first decision time point and the second decision time point being a decision cycle time length of the automatic driving program;

[0019] generating, based on the static environment data in the accident scene data, the position data and the motion state data of all dynamic objects in the accident scene data at the second decision time point, the position data and the motion state data of the test vehicle at the second decision time point, and the simulation sensor data of the test vehicle within a second specified time length before the first decision time point, simulation sensor data at the second decision time point;

[0020] processing, based on the automatic driving program of the test vehicle, the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length before the second decision time point, to obtain second decision information.

[0021] In a possible implementation, an interval time length between two adjacent time points is equal to a decision cycle time length of the automatic driving program, and after the simulation sensor data of the test vehicle within the second specified time length is processed based on the automatic driving program of the test vehicle to obtain the first decision information, the method further includes:

[0022] determining, based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, position data and motion state data of the test vehicle at a second decision time point, the second decision time point being after the first decision time point, and an interval time length between the first decision time point and the second decision time point being a decision cycle time length of the automatic driving program;

[0023] processing, based on simulation motion logic in the test program, the static environment data in the accident scene data and the dynamic object data within the second specified time length before the first decision time point, and the position data and the motion state data of the test vehicle within the second specified time length before the first decision time point, to obtain position data and motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point;

[0024] generate simulation sensor data of the test vehicle at the second decision time point based on static environment data in the accident scene data, position data and motion state data of the accident vehicle at the second decision time point in the accident scene data, position data and motion state data of the test vehicle at the second decision time point, position data and motion state data of other objects other than the test vehicle and the accident vehicle at the second decision time point, and simulation sensor data of the test vehicle within a second specified time length before the first decision time point;

[0025] process the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length before the second decision time point based on the automatic driving program of the test vehicle to obtain second decision information.

[0026] In a possible implementation, the at least one dynamic object further includes at least one non-accident vehicle, and the test vehicle is a vehicle in the at least one non-accident vehicle;

[0027] The determination of the first position data and the first motion state data of the test vehicle within the first specified time length before the first decision time point includes:

[0028] The first position data and the first motion state data of the test vehicle within the first specified time length before the first decision time point are determined based on accident-related data of the test vehicle within the first specified time length before the first decision time point.

[0029] In a possible implementation, the interval time length of adjacent two time points is equal to a decision cycle time length of the automatic driving program, and after the simulation sensor data of the test vehicle within the second specified time length is processed based on the automatic driving program of the test vehicle to obtain the first decision information, the method further includes:

[0030] The position data and the motion state data of the test vehicle at a second decision time point are determined based on the position data, the motion state data of the test vehicle at the first decision time point, and the first decision information, the second decision time point is after the first decision time point, and an interval time length between the first decision time point and the second decision time point is a decision cycle time length of the automatic driving program.

[0031] generate the simulation sensor data of the test vehicle at the second decision time point based on the static environment data in the accident scene data, the position data and the motion state data of the dynamic objects other than the test vehicle in the accident scene data at the second decision time point, the position data and the motion state data of the test vehicle at the second decision time point, and the simulation sensor data of the test vehicle within the second specified time length before the first decision time point;

[0032] process the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length based on the automatic driving program of the test vehicle to obtain second decision information.

[0033] In a possible implementation, the interval time length of two adjacent time points is equal to the decision cycle time length of the automatic driving program, and after the simulation sensor data of the test vehicle within the second specified time length is processed based on the automatic driving program of the test vehicle to obtain the first decision information, the method further includes:

[0034] determine the position data and the motion state data of the test vehicle at a second decision time point based on the position data, the motion state data of the test vehicle at the first decision time point, and the first decision information, the second decision time point being after the first decision time point, and the interval time length between the first decision time point and the second decision time point being the decision cycle time length of the automatic driving program;

[0035] process the static environment data in the accident scene data and the dynamic object data within the second specified time length before the first decision time point based on the simulation motion logic in the test program to obtain the position data and the motion state data of the other objects other than the test vehicle and the accident vehicle at the second decision time point;

[0036] generate the simulation sensor data of the test vehicle at the second decision time point based on the static environment data in the accident scene data, the position data and the motion state data of the accident vehicle in the accident scene data at the second decision time point, the position data and the motion state data of the test vehicle at the second decision time point, the position data and the motion state data of the other objects other than the test vehicle and the accident vehicle at the second decision time point, and the simulation sensor data of the test vehicle within the second specified time length before the first decision time point;

[0037] The processing module is configured to process the simulation sensor data of the test vehicle in the second specified time length before the first decision time point based on an automatic driving program of the test vehicle, to obtain first decision information.

[0038] In a second aspect, the embodiments of the present disclosure provide a device for vehicle simulation testing, and the device comprises:

[0039] The acquisition module is configured to acquire accident-related data in a time range and a location range corresponding to a target traffic accident, wherein the accident-related data comprises loop video data and sensor data of at least one vehicle, and the at least one vehicle comprises an accident vehicle.

[0040] The establishment module is configured to establish accident scene data based on the accident-related data, wherein the accident scene data comprises static environment data and dynamic object data, the dynamic object data comprises position data and motion state data of at least one dynamic object in a target time period, and the at least one dynamic object comprises the at least one vehicle, and the target time period is in the time range.

[0041] The determination module is configured to determine a first decision time point and determine position data and motion state data of a test vehicle in a first specified time length before the first decision time point, wherein the first decision time point is in the target time period, and the test vehicle is a vehicle other than the accident vehicle.

[0042] The generation module is configured to generate simulation sensor data of the test vehicle in a second specified time length before the first decision time point based on position data of a loop monitoring device corresponding to the loop video data, accident-related data in the first specified time length before the first decision time point, position data and motion state data of the test vehicle in the first specified time length before the first decision time point, wherein the simulation sensor data in the second specified time length before the first decision time point comprises simulation sensor data corresponding to a plurality of time points in the second specified time length, the plurality of time points are uniformly distributed in the second specified time length and comprise the first decision time point, and the first specified time length is greater than the second specified time length.

[0043] The processing module is configured to process the simulation sensor data of the test vehicle in the second specified time length before the first decision time point based on an automatic driving program of the test vehicle, to obtain first decision information.

[0044] In a possible implementation, the sensor data comprises vehicle video data, vehicle radar data, and motion state data.

[0045] In a possible implementation, the motion state data includes a vehicle speed and a vehicle steering angle.

[0046] In a possible implementation, the determining module is configured to:

[0047] receive a test vehicle addition request;

[0048] determine, from the test vehicle addition request, position data and motion state data of the test vehicle within a first specified time length before the first decision time point.

[0049] In a possible implementation, the interval length between adjacent time points is equal to a decision cycle length of the automatic driving program, and the processing module is further configured to:

[0050] determine, based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, position data and motion state data of the test vehicle at a second decision time point, the second decision time point being after the first decision time point, and the interval length between the first decision time point and the second decision time point being the decision cycle length of the automatic driving program;

[0051] generate, based on static environment data in the accident scene data, position data and motion state data of all dynamic objects in the accident scene data at the second decision time point, the position data and the motion state data of the test vehicle at the second decision time point, and simulation sensor data of the test vehicle within a second specified time length before the first decision time point, simulation sensor data at the second decision time point;

[0052] process, based on the automatic driving program of the test vehicle, the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length before the second decision time point, to obtain second decision information.

[0053] In a possible implementation, the interval length between adjacent time points is equal to a decision cycle length of the automatic driving program, and the processing module is further configured to:

[0054] determine, based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, position data and motion state data of the test vehicle at a second decision time point, the second decision time point being after the first decision time point, and the interval length between the first decision time point and the second decision time point being the decision cycle length of the automatic driving program;

[0055] based on the simulation motion logic in the test program, processing the static environment data in the accident scene data and the dynamic object data within a second specified time length before the first decision time point, and the position data and motion state data of the test vehicle within the second specified time length before the first decision time point, to obtain the position data and motion state data of the other objects except the test vehicle and the accident vehicle at the second decision time point;

[0056] based on the static environment data in the accident scene data, the position data and motion state data of the accident vehicle at the second decision time point in the accident scene data, the position data and motion state data of the test vehicle at the second decision time point, the position data and motion state data of the other objects except the test vehicle and the accident vehicle at the second decision time point, and the simulation sensor data of the test vehicle within the second specified time length before the first decision time point, generating the simulation sensor data of the test vehicle at the second decision time point;

[0057] based on the automatic driving program of the test vehicle, processing the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length before the second decision time point, to obtain the second decision information.

[0058] In a possible implementation, the at least one dynamic object further includes at least one non-accident vehicle, and the test vehicle is a vehicle in the at least one non-accident vehicle;

[0059] The determination module is configured to:

[0060] based on the accident related data of the test vehicle within the first specified time length before the first decision time point, determining the first position data and the first motion state data of the test vehicle within the first specified time length before the first decision time point.

[0061] In a possible implementation, the interval time length of the adjacent two time points is equal to the decision cycle time length of the automatic driving program, and the processing module is further configured to:

[0062] based on the position data, the motion state data of the test vehicle at the first decision time point and the first decision information, determining the position data and the motion state data of the test vehicle at the second decision time point, the second decision time point being after the first decision time point, and the interval time length of the first decision time point and the second decision time point being the decision cycle time length of the automatic driving program;

[0063] generate the simulation sensor data of the test vehicle at the second decision time point based on the static environment data in the accident scene data, the position data and the motion state data of the dynamic objects other than the test vehicle in the accident scene data at the second decision time point, the position data and the motion state data of the test vehicle at the second decision time point, and the simulation sensor data of the test vehicle within a second specified time length before the first decision time point;

[0064] process the simulation sensor data of the test vehicle at the second decision time point and within a second specified time length before the second decision time point based on the automatic driving program of the test vehicle, to obtain second decision information.

[0065] In a possible implementation, the interval time length of the two adjacent time points is equal to the decision cycle time length of the automatic driving program, and the processing module is further configured to:

[0066] determine the position data and the motion state data of the test vehicle at a second decision time point based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, the second decision time point being after the first decision time point, and the interval time length between the first decision time point and the second decision time point being the decision cycle time length of the automatic driving program;

[0067] process the static environment data in the accident scene data and the dynamic object data within a second specified time length before the first decision time point based on the simulation motion logic in the test program, to obtain the position data and the motion state data of the objects other than the test vehicle and the accident vehicle at the second decision time point;

[0068] generate the simulation sensor data of the test vehicle at the second decision time point based on the static environment data in the accident scene data, the position data and the motion state data of the accident vehicle in the accident scene data at the second decision time point, the position data and the motion state data of the test vehicle at the second decision time point, the position data and the motion state data of the objects other than the test vehicle and the accident vehicle at the second decision time point, and the simulation sensor data of the test vehicle within a second specified time length before the first decision time point;

[0069] process the simulation sensor data of the test vehicle at the second decision time point and within a second specified time length before the second decision time point based on the automatic driving program of the test vehicle, to obtain second decision information.

[0070] In a third aspect, an electronic device is provided, which includes a memory and a processor. The memory is configured to store computer instructions. The processor executes the computer instructions stored in the memory, so that the electronic device executes the method of the first aspect and possible implementation manners thereof.

[0071] In a fourth aspect, a computer readable storage medium is provided, which stores computer program codes. When the computer program codes are executed by an electronic device, the electronic device executes the method of the first aspect and possible implementation manners thereof.

[0072] In a fifth aspect, a computer program product is provided, which includes computer program codes. When the computer program codes are executed by an electronic device, the electronic device executes the method of the first aspect and possible implementation manners thereof.

[0073] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and are not limiting to the present disclosure.

[0074] In the present disclosure, after the accident scene is established based on the accident related data, the test vehicle can be set as a bystander vehicle in the accident scene, i.e., a vehicle other than the accident vehicle, so as to verify the safety of the automatic driving program. Specifically, the simulation sensor data is generated based on the accident related data and the position data and motion state data of the test vehicle, and a decision is made based on the simulation sensor data. In this way, in the simulation scene, the test vehicle can be set as not only the accident vehicle but also any bystander vehicle, so as to fully utilize the simulation scene and reduce the waste of processing resources. BRIEF DESCRIPTION OF DRAWINGS

[0075] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present disclosure, and other drawings can be obtained by those skilled in the art without creative labor.

[0076] Figure 1 is a structural schematic diagram of a computer device provided by an embodiment of the present disclosure;

[0077] Figure 2 is a flowchart of a vehicle simulation test provided by an embodiment of the present disclosure;

[0078] Figure 3 is a flowchart of a vehicle simulation test provided by an embodiment of the present disclosure;

[0079] Figure 4 is a flowchart of a vehicle simulation test provided by an embodiment of the present disclosure;

[0080] Figure 5 is a flowchart of a vehicle simulation test provided by an embodiment of the present disclosure;

[0081] Figure 6 is a flowchart of a vehicle simulation test provided by an embodiment of the present disclosure;

[0082] Figure 7 is a structural diagram of a device for vehicle simulation test provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0083] The present embodiment provides a vehicle simulation test method, which can be implemented by a computer device, which can be a terminal device or a server. The terminal device can be a mobile phone, an iPad, a desktop computer, a notebook computer, and the like. The server can be a server of an application program, a server of a cloud service, or a server for completing certain computing tasks, and the like. The present embodiment takes the terminal device as an example to describe the scheme in detail, and other cases are similar, which will not be repeated here.

[0084] As shown in Figure 1 , the terminal device can include a processor 110, a memory 120, and a communication component 130, and the like.

[0085] The processor 110 can be a central processing unit (CPU), which can be used to execute various operation instructions. For example, the processor can be used to establish an accident scene by analyzing accident-related data, and can also be used to generate simulated sensor data and make decisions based on the simulated sensor data, and the like.

[0086] The memory 120 can be various volatile memories or non-volatile memories, such as a solid state disk (SSD), a dynamic random access memory (DRAM) memory, and the like. The memory can be used to store pre-stored data, intermediate data, and result data related to the processing process, such as position data, motion state data, and simulated sensor data, and the like.

[0087] The communication component 130 can be a wired network connector, a wireless fidelity (WiFi) module, a Bluetooth module, a cellular network communication module, and the like. The communication component can be used for data transmission with other devices. For example, the communication component can be used to receive operation instructions of the processor, and can also be used to send decision information, position data and motion state data of each object to other computer devices, and the like.

[0088] Generally, after the accident scene data is established based on the real-life accident scene, the test vehicle is detected as the accident vehicle, that is, the test vehicle is configured based on the position data and various sensor data of the accident vehicle to test whether the test vehicle can respond to the emergency situation in time when the automatic driving program is used. In this way, the use method of the accident scene is more limited.

[0089] The method for vehicle simulation test provided by the embodiments of the present disclosure can use the test vehicle as the accident scene to obtain the bystander vehicle to detect the response of the test vehicle, and fully utilizes the accident scene when the vehicle is simulated and tested. As shown in the following figure, Figure 2 The method processing flow can include the following steps:

[0090] Step 201, obtaining accident related data in a time range and a position range corresponding to a target traffic accident.

[0091] The accident related data includes the card mouth video data and the sensor data of the accident vehicle. The card mouth video data is the monitoring video data of the accident road section, and one road section can correspond to multiple card mouth video data. The sensor data of the accident vehicle includes vehicle video data, vehicle radar data and motion state data. The vehicle video data can be a driving recorder, or video data shot by other cameras on the vehicle. The vehicle radar data can be a laser radar, or a millimeter wave radar, etc. The motion state data includes vehicle speed and vehicle steering angle. In addition, the accident related data can also include the accident scene report recorded by the traffic management department, the accident scene picture, etc. to obtain more perfect data to provide more detailed information for subsequent establishment of accident scene data.

[0092] The card mouth video data of the traffic accident road section and the sensor data of the accident vehicle in a time range (which can be called a time range) before and after the time point of the traffic accident are obtained. The time range can be set by the relevant technical personnel, for example, the time range can be one minute before the time point of the traffic accident to one minute after the time point of the traffic accident.

[0093] Step 202, establishing accident scene data based on the accident related data.

[0094] The accident scene data includes static environment data and dynamic object data. First, the static environment is introduced. The static environment can be divided into road conditions, road structures and weather. The road conditions can include the type of road surface (e.g. concrete, asphalt or epoxy floor), the humidity of the road surface, the flatness of the road surface (e.g. the pits, cracks, road camber, etc. of the road surface), the traffic signs on the road surface (e.g. speed limit signs, warning signs, etc.), other static objects (e.g. trees, etc.), the slope of the road surface, the friction coefficient of the road surface, etc. The road structure can include the type of road (e.g. straight road, curved road, lane-changing section, on-ramp, off-ramp, roundabout, parking lot, tunnel, S-bend, X / Y-shaped road or T-shaped road, etc.) and the lane information of the road (e.g. lane width, lane line type (e.g. dashed line, solid line, double solid line or double dashed line, etc.), number of lanes (e.g. single lane, double lane or triple lane in one direction, four lanes, six lanes or eight lanes in two directions, etc.)). The weather can include natural weather (e.g. rain, snow, fog or haze, etc.), light (e.g. sunlight, vehicle light, etc.). Accordingly, the static environment data can be roughly represented as point data of static objects, which can be traffic signs on the road surface (e.g. speed limit signs, warning signs, etc.) and other static objects (e.g. trees, stones, etc.), and the point data can be one point or multiple points, which can represent the outline of the static object, e.g. the four vertices of a traffic sign or many points on the outline of the traffic sign. The dynamic object data includes the position data and the motion state data of each dynamic object, which includes vehicles and pedestrians, and the vehicles can include passenger cars, trucks, vans, motorcycles, bicycles, etc. The position data can be point data, which can represent the outline of the dynamic object, e.g. the four vertices of a vehicle or many points on the outline of the vehicle. The motion state data of the vehicle includes the speed and the steering angle of the vehicle, and the motion state data of the pedestrian includes the speed and the direction of the pedestrian.

[0095] The dynamic objects can only include the objects related to the accident, i.e. the objects involved in the traffic accident. For example, when the accident is between vehicles, the dynamic objects can only include the accident vehicles, and when the accident is between a vehicle and a pedestrian, the dynamic objects include the accident vehicle and the pedestrian. In addition to the objects related to the accident, the dynamic objects can also include all vehicles and pedestrians that pass through the road section where the accident occurs, i.e. the objects that witness the traffic accident. The embodiments of the present disclosure take the case where the dynamic objects include the objects related to the accident and all vehicles and pedestrians that pass through the road section where the accident occurs as an example to explain the scheme in detail, and other cases are similar and will not be repeated here.

[0096] In step 203, the first decision time point is determined, and the position data and the motion state data of the test vehicle within a first specified time period before the first decision time point are determined.

[0097] The first decision time point is within the target time period. The first decision time point can be a point in time before the traffic accident, a point in time after the traffic accident, or the exact time the traffic accident occurs. The first decision time point can be set by relevant technical personnel. This disclosure does not limit this. The first decision time point indicates that after this time point, the test vehicle operates based on the autonomous driving program. The first specified duration is also set by relevant technical personnel, for example, it could be one minute.

[0098] The test vehicle can be any non-accident vehicle in the dynamic object, or it can be a vehicle newly added by relevant technicians in the accident scenario. This disclosure does not limit this.

[0099] Before the first decision point, for all dynamic objects, script data is created for each object based on its position and motion state data in the real world. After the first decision point, for objects involved in traffic accidents among the dynamic objects, script data is still created for each object based on its position and motion state data in the real world, so that the object's movement route is the same as its movement route in the real world. For objects witnessing traffic accidents among the dynamic objects, script data can be created for them based on their position and motion state data in the real world, or they can be made to move based on the simulated motion logic in the test program. This disclosure does not limit the specific implementation of the embodiments.

[0100] The specific operational procedures for this step differ depending on the setup methods for the test vehicle and the motion state settings of objects witnessing traffic accidents in dynamic scenarios. Please refer to the relevant documentation for details. Figure 3 , Figure 4 , Figure 5 and Figure 6 The processing flow is shown below.

[0101] Step 204: Based on the location data of the checkpoint monitoring equipment corresponding to the checkpoint video data, the accident-related data within the first specified time period before the first decision time point, and the location data and motion status data of the test vehicle within the first specified time period before the first decision time point, generate the simulated sensor data of the test vehicle within the second specified time period before the first decision time point.

[0102] The first specified time period is greater than the second specified time period. The second specified time period can be set by a person skilled in the art according to actual conditions. The simulated sensor data includes simulated vehicle video data, vehicle radar data, vehicle speed, and vehicle steering angle. The simulated sensor data within the first specified time period before the first decision time point includes simulated sensor data generated at multiple time points within the first specified time period, the multiple time points are uniformly distributed within the first specified time period and include the first decision time point, and the interval time period between adjacent two time points is equal to the decision cycle time period of the automatic driving program. The second specified time period and the interval time period can be set by a person skilled in the art, the longer the second specified time period, the more data that can be referred to by the automatic driving program when making decisions, but correspondingly, the processing resources consumed will also increase; the shorter the interval time period, the more frequently the simulated sensor data is generated, and the more frequently the automatic driving program makes decisions, which can better control the vehicle to respond to unexpected situations, but correspondingly, the processing resources consumed will also increase, for example, the second specified time period is 1s, and the interval time period between adjacent two time points is 0.01s, then 100 simulated sensor data is generated within 1s (including the end of 1s).

[0103] The simulated sensor data of the test vehicle can be generated based on the accident-related data within the first specified time period before the first decision time point, and the position data and motion state data of the test vehicle within the first specified time period before the first decision time point. The simulated sensor data includes simulated vehicle video data, simulated vehicle radar data, and motion state data, the motion state data is known in step 203, and the generation of simulated vehicle video data and simulated vehicle radar data is introduced below.

[0104] A machine learning model can be used to generate vehicle video data based on the fisheye video data, and the machine learning model used is different for different methods set by the test vehicle. The following are introduced respectively.

[0105] For the test vehicle being a non-accident vehicle in the dynamic object, the first machine learning model can be used to generate the simulation vehicle video data of the test vehicle. A large amount of training is performed on the first machine learning model in advance, each sample data including input data and benchmark data, the input data in the sample data being the turret video data photographed within a time range having the target vehicle and the position data of the turret monitoring device corresponding to the turret video data, the benchmark data being the vehicle video data of the target vehicle. The turret video data is the video data in which a traffic accident occurs within the position range in which the video data is photographed, and the target vehicle is a non-accident vehicle. The training method is that the input data of the sample data is input to the first machine learning model to obtain output data, the first machine learning model is adjusted based on the loss value of the output data and the benchmark data, and the first machine learning model is adjusted until the loss value of the first machine learning model is within a certain range. After a large amount of training, the first machine learning model can generate simulation vehicle video data of any vehicle photographed by the video data within the time range based on the turret video data and the position data of the turret monitoring device corresponding to the turret video data.

[0106] For the test vehicle being a newly added vehicle, the second machine learning model can be used to generate the simulation vehicle video data of the test vehicle. A large amount of training is performed on the second machine learning model in advance, each sample data including input data and benchmark data, the input data being the vehicle position data, the vehicle motion state data of the target vehicle, the turret video data photographed within a time range having the target vehicle (the processed turret video data is the turret video data in which the vehicle is erased from the turret video data, which can be operated by using image processing software), and the position data of the turret monitoring device corresponding to the turret video data, and the benchmark data being the vehicle video data of the target vehicle. The turret video data is the video data in which a traffic accident occurs within the position range in which the video data is photographed, and the target vehicle is a non-accident vehicle. The training method is that the input data of the sample data is input to the second machine learning model to obtain output data, the second machine learning model is adjusted based on the loss value of the output data and the benchmark data, and the second machine learning model is adjusted until the loss value of the second machine learning model is within a certain range. After a large amount of training, the second machine learning model can generate simulation vehicle video data based on the turret video data, the position data of the turret monitoring device corresponding to the turret video data, and the position data and motion state data of the vehicle within the position range in which the turret video data is photographed.

[0107] The generation method of vehicle radar data is introduced below. For vehicle radar data at a certain time point, static environment data and dynamic object data in the accident scene at the time point can be used to generate simulated vehicle radar data. Specifically, based on the position data of the test vehicle at the time point, the position data of the dynamic objects and the position data of the static objects, the relative positions of the test vehicle and each dynamic object and the relative positions of the test vehicle and each static object are obtained, and then the simulated vehicle radar data at the time point is generated. In addition, the vehicle radar data can be generated by using a machine learning model by replacing the vehicle video data with the vehicle radar data using a similar idea as described above, which will not be described in detail here.

[0108] The vehicle video data, vehicle radar data and motion state data of the test vehicle at the first decision time point constitute the simulated sensor data at the first decision time point.

[0109] In step 205, the simulated sensor data of the test vehicle within the second specified time period is processed based on the automatic driving program of the test vehicle to obtain first decision information.

[0110] The automatic driving program processes the simulated sensor data of the test vehicle generated within the second specified time period to obtain first decision information, which includes brake pedal angle, accelerator pedal angle, steering wheel angle, vehicle light, etc. Still taking the second specified time period as 1s and the interval time period between adjacent two time points as 0.01s as an example, at the end of 1s, the automatic driving program generates first decision information based on 100 simulated sensor data within 1s (including the end of 1s). The decision logic of the automatic driving program is set by the relevant technical personnel, and this embodiment of the present disclosure will not be described in detail. The following is described by way of example, for example, when the automatic driving program knows that the front vehicle is decelerating based on the simulated sensor data, and there is no vehicle driving in the left lane, it can decelerate and adjust the steering wheel angle to change lanes to the left side, and turn on the left turn signal.

[0111] After obtaining the first decision information, the test vehicle continues to drive based on the first decision information, and then generates simulated sensor data again after a time interval, and then generates second decision information, and the test vehicle continues to drive based on the second decision information, and then generates simulated sensor data and decision information, and so on until reaching the destination. For different setting methods of different test vehicles, and for different setting methods of objects witnessing traffic accidents in dynamic scenes, the specific operation process is different, and the related content is described in detail in the processing flow shown in Figure 3 、 Figure 4 、 Figure 5 and Figure 6 .

[0112] When the test vehicle is added by relevant technicians, and all objects other than the test vehicle and those involved in the accident move in the same manner after the first decision point, the vehicle simulation test method is as follows: Figure 3 As shown, it includes the following steps:

[0113] Step 301: Obtain accident-related data within the time and location range corresponding to the target traffic accident.

[0114] The processing flow for this step is similar to that of step 201, and you can refer to the description of step 201.

[0115] Step 302: Based on accident-related data, establish accident scenario data.

[0116] The processing flow for this step is similar to that of step 202, and you can refer to the description of step 202.

[0117] Step 303: Receive the test vehicle addition request, and determine the first decision time point, as well as the location data and motion status data of the test vehicle carried in the test vehicle addition request within the first specified time period before the first decision time point.

[0118] The first decision point is within the target time period. The first decision point is a point in time before the traffic accident occurs; after this point, the test vehicle operates based on the autonomous driving program. The first decision point can be set by relevant technical personnel. The first specified duration is also set by relevant technical personnel, for example, it could be 1 minute.

[0119] A test vehicle is added to the accident scenario, and its initial velocity and travel route are set. Based on the initial velocity and travel route, the position and motion data for a first specified time period before the first decision point can be determined. It can be assumed that before the first decision point, the vehicle maintains its initial velocity and the lane in which the test vehicle travels remains unchanged.

[0120] Step 304: Based on the location data of the checkpoint monitoring equipment corresponding to the checkpoint video data, the accident-related data within the first specified time period before the first decision time point, and the location data and motion status data of the test vehicle within the first specified time period before the first decision time point, generate the simulated sensor data of the test vehicle within the second specified time period before the first decision time point.

[0121] The processing flow for this step is similar to that of step 204, and you can refer to the description of step 204.

[0122] Step 305: Based on the autonomous driving program of the test vehicle, process the simulated sensor data of the test vehicle within the second specified time period to obtain the first decision information.

[0123] The processing flow of this step is similar to step 205, and please refer to the introduction of step 205.

[0124] In step 306, based on the position data and motion state data of the test vehicle at the first decision time point and the first decision information, the position data and motion state data of the test vehicle at the second decision time point are determined.

[0125] The second decision time point is after the first decision time point, and the interval time length between the first decision time point and the second decision time point is the decision cycle time length of the automatic driving program. For example, the second specified time length is 1s, the interval time length between adjacent two time points is 0.01s, the first decision time point is 00:01:00, and the automatic driving program generates the first decision information based on 100 simulation sensor data between 00:00:00 and 00:01:00 (including the time point 00:01:00 and excluding the time point 00:00:00). The second decision time point is 00:01:01. After obtaining the first decision information at the first decision time point, it can be approximately considered that the motion is performed according to the first decision information between the second decision time point and the first decision time point. Further, based on the position data and motion state data of the test vehicle at the first decision time point and the first decision information, the position data and motion state data of the test vehicle at the second decision time point are determined.

[0126] In step 307, based on the static environment data in the accident scene data, the position data and motion state data of all dynamic objects in the accident scene data at the second decision time point, the position data and motion state data of the test vehicle at the second decision time point, and the simulation sensor data of the test vehicle within the second specified time length before the first decision time point, the simulation sensor data at the second decision time point is generated.

[0127] First, the vehicle video data in the generated simulated sensor data is introduced. The third machine learning model can be used to generate the simulated vehicle video data of the test vehicle at the second decision time point. A large amount of training is performed on the third machine learning model in advance using a large amount of sample data, each of which includes input data and benchmark data. The input data is the vehicle video data of the target vehicle for a specified time length before the first time point, the position data of the target vehicle at the second time point, the position data of all dynamic objects within the position range at the second time point, and the position data of all static objects within the position range at the second time point. The position range can be a certain number of kilometers around the traffic accident or the range covered by the loop video data. The benchmark data is the vehicle video data of the target vehicle at the second time point. The training method is to input the input data of the sample data into the third machine learning model to obtain the output data, and based on the loss value of the output data and the benchmark data, the third machine learning model is adjusted until the loss value of the third machine learning model is within a certain range. After a large amount of training, the third machine learning model can generate the simulated vehicle video data of the test vehicle at the second decision time point based on the vehicle video data of the test vehicle for a second specified time length before the first decision time point, the position data of the test vehicle at the second decision time point, the position data of all dynamic objects at the second decision time point, and the position data of all static objects at the second decision time point.

[0128] Next, the vehicle radar data in the generated simulated sensor data is introduced. For vehicle radar data at a certain time point, it can be generated based on static environment data and dynamic object data in the accident scene at that time point. Specifically, based on the position data of the test vehicle, the position data of the dynamic objects and the position data of the static objects at the time point, the relative positions of the test vehicle and each dynamic object are obtained, and the relative positions of the test vehicle and each static object are obtained, and the simulated vehicle radar data at the time point is generated. In addition, similar to the above, the vehicle video data can be replaced by the vehicle radar data, and the machine learning model can be used to generate the vehicle radar data, which will not be described in detail here.

[0129] The vehicle video data, vehicle radar data and motion state data of the test vehicle corresponding to the second decision time point constitute the simulated sensor data at the second decision time point.

[0130] In step 308, based on the automatic driving program of the test vehicle, the simulated sensor data of the test vehicle at the second decision time point and within a second specified time length before the second decision time point is processed to obtain the second decision information.

[0131] At the second decision time point, based on the automatic driving program of the test vehicle, the simulated sensor data generated at the second decision time point, the simulated sensor data generated at the first decision time point, and the simulated sensor data generated at a plurality of time points before the first decision time point are processed to obtain second decision information, and the time period between the earliest time point in the plurality of time points before the first decision time point and the second decision time point is the second specified duration. For example, the second specified duration is 1s, the interval duration of adjacent two time points is 0.01s, the first decision time point is 00:01:00, and the second decision time point is 00:01:01. The automatic driving program generates the second decision information based on 100 simulated sensor data between 00:00:01 and 00:01:01 (including the time point 00:00:02 and excluding the time point 00:01:01).

[0132] The time point with an interval duration of 0.01s after the second decision time point can be taken as a third decision time point, and then the simulated sensor data of the test vehicle at the third decision time point is generated, and based on the automatic driving program of the test vehicle, the simulated sensor data of the test vehicle at the third decision time point and within the second specified duration before the third decision time point is processed to obtain third decision information. The processing here is similar to steps 307-308 described above, and then the cycle is repeated until the test vehicle reaches the specified position or the test vehicle is forced to stop due to collision.

[0133] When the test vehicle is added by a related technical person, in addition to the test vehicle and the object involved in the accident, other objects move based on the simulated motion logic in the test program after the first decision point, the vehicle simulation test method is as shown in Figure 4 The method comprises the following steps:

[0134] Step 401, acquiring accident-related data in the time range and position range corresponding to the target traffic accident.

[0135] The processing procedure of this step is similar to that of step 201, and can be referred to the introduction of step 201.

[0136] Step 402, establishing accident scene data based on the accident-related data.

[0137] The processing procedure of this step can be similar to that of step 202, and can be referred to the introduction of step 202.

[0138] Step 403, receiving a test vehicle addition request and determining a first decision time point and position data and motion state data of the test vehicle within a first specified duration before the first decision time point carried in the test vehicle addition request.

[0139] The processing procedure of this step is similar to that of step 303, and the description of step 303 can be referred to.

[0140] In step 404, the simulation sensor data of the test vehicle in the first specified time period before the first decision time point is generated based on the accident-related data in the first specified time period before the first decision time point, and the position data and the motion state data of the test vehicle in the first specified time period before the first decision time point.

[0141] The processing procedure of this step is similar to that of step 204, and the description of step 204 can be referred to.

[0142] In step 405, the simulation sensor data of the test vehicle in the second specified time period is processed based on the automatic driving program of the test vehicle to obtain the first decision information.

[0143] The processing procedure of this step is similar to that of step 205, and the description of step 205 can be referred to.

[0144] In step 406, the position data and the motion state data of the test vehicle at the second decision time point are determined based on the position data, the motion state data and the first decision information of the test vehicle at the first decision time point.

[0145] The processing procedure of this step is similar to that of step 306, and the description of step 306 can be referred to.

[0146] In step 407, the position data and the motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point are obtained by processing the static environment data in the accident scene data and the dynamic object data in the second specified time period before the first decision time point, and the position data and the motion state data of the test vehicle in the second specified time period before the first decision time point based on the simulation motion logic in the test program.

[0147] The simulation motion logic can be automatically generated by a simulation program, for example, the simulation program can be a virtual test drive (VTD) program.

[0148] In step 408, the simulation sensor data of the test vehicle at the second decision time point is generated based on the static environment data in the accident scene data, the position data and the motion state data of the accident vehicle at the second decision time point in the accident scene data, the position data and the motion state data of the test vehicle at the second decision time point, the position data and the motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point, and the simulation sensor data of the test vehicle in the second specified time period before the first decision time point.

[0149] The vehicle video data in the generated simulated sensor data is introduced first. The fourth machine learning model can be used to generate the simulated vehicle video data of the test vehicle at the second decision time point. The fourth machine learning model is trained in advance using a large amount of sample data, each of which includes input data and benchmark data. The input data includes the vehicle video data of the target vehicle within a specified time length before the first time point, the position data of the target vehicle at the second time point, the position data of the accident vehicle at the second decision time point, the position data of other objects in the position range at the second time point except the target vehicle and the accident vehicle, and the position data of all static objects in the position range at the second time point. The position range can be a certain number of kilometers around the traffic accident or the range covered by the loop video data. The benchmark data is the vehicle video data of the target vehicle at the second time point. The training method is to input the input data of the sample data into the fourth machine learning model to obtain output data, and to adjust the parameters of the fourth machine learning model based on the loss value of the output data and the benchmark data until the loss value of the fourth machine learning model is within a certain range. After a large amount of training, the fourth machine learning model can generate the simulated vehicle video data of the test vehicle at the second decision time point based on the vehicle video data of the test vehicle within a second specified time length before the first decision time point, the position data of the test vehicle at the second decision time point, the position data of the accident vehicle at the second decision time point, the position data of other objects except the test vehicle and the accident vehicle at the second decision time point, and the position data of all static objects at the second decision time point.

[0150] The generation of vehicle radar data in the simulated sensor data is introduced below. For vehicle radar data at a certain time point, it can be generated based on the static environment data and dynamic object data in the accident scene at the time point. Specifically, based on the position data of the test vehicle at the time point, the position data of the dynamic objects and the position data of the static objects, the relative positions of the test vehicle and each dynamic object, and the relative positions of the test vehicle and each static object are obtained, and the simulated vehicle radar data at the time point is generated. In addition, similar to the above, the vehicle video data can be replaced by the vehicle radar data, and the machine learning model can be used to generate the vehicle radar data, which will not be described in detail here.

[0151] At step 409, the simulated sensor data of the test vehicle at the second decision time point and within a second specified time length before the second decision time point is processed based on the automatic driving program of the test vehicle to obtain second decision information.

[0152] The processing procedure of this step is similar to that of step 308, and can be referred to the introduction of step 308.

[0153] When the test vehicle is a non-accident vehicle in a dynamic scenario, and all other objects except the test vehicle and those involved in the accident move in the same way after the first decision point, the vehicle simulation test method is shown in Figure 5, including the following steps:

[0154] Step 501: Obtain accident-related data within the time and location range corresponding to the target traffic accident.

[0155] The processing flow for this step is similar to that of step 201, and you can refer to the description of step 201.

[0156] Step 502: Based on accident-related data, establish accident scenario data.

[0157] The processing flow for this step is similar to that of step 202, and you can refer to the description of step 202.

[0158] Step 503: Determine the first decision time point, and based on the accident-related data of the test vehicle within the first specified time period before the first decision time point, determine the first position data and the first motion state data of the test vehicle within the first specified time period before the first decision time point.

[0159] The first decision point is within the target time period. The first decision point is a point in time before the traffic accident occurs; after this point, the test vehicle operates based on the autonomous driving program. The first decision point can be set by relevant technical personnel. The first specified duration is also set by relevant technical personnel, for example, it could be one minute.

[0160] The test vehicle can be a non-accident vehicle in the dynamic object. The accident-related data of the test vehicle within the first specified time period before the first decision time point are the location data and motion state data of the corresponding non-accident vehicle in the dynamic object.

[0161] Step 504: Based on the location data of the checkpoint monitoring equipment corresponding to the checkpoint video data, the accident-related data within the first specified time period before the first decision time point, and the location data and motion status data of the test vehicle within the first specified time period before the first decision time point, generate the simulated sensor data of the test vehicle within the second specified time period before the first decision time point.

[0162] The processing flow for this step is similar to that of step 204, and you can refer to the description of step 204.

[0163] Step 505: Based on the autonomous driving program of the test vehicle, process the simulated sensor data of the test vehicle within the second specified time period to obtain the first decision information.

[0164] The processing flow of this step can be similar to step 205, and the description of step 205 can be referred to.

[0165] In step 506, the position data and the motion state data of the test vehicle at the second decision time point are determined based on the position data, the motion state data of the test vehicle at the first decision time point, and the first decision information.

[0166] The processing flow of this step is similar to that of step 306, and the description of step 306 can be referred to.

[0167] In step 507, the simulated sensor data of the test vehicle at the second decision time point is generated based on the static environment data in the accident scene data, the position data and the motion state data of the dynamic objects other than the test vehicle in the accident scene data at the second decision time point, the position data and the motion state data of the test vehicle at the second decision time point, and the simulated sensor data of the test vehicle within the second specified time period before the first decision time point.

[0168] First, the vehicle video data in the generated simulated sensor data is introduced. The fifth machine learning model can be used to generate the simulated vehicle video data of the test vehicle at the second decision time point. A large amount of training is performed on the fifth machine learning model in advance using a large amount of sample data, and each sample data includes input data and benchmark data. The input data is the vehicle video data of the target vehicle within the specified time period before the first time point, the position data of the target vehicle at the second time point, the position data of all dynamic objects within the position range at the second time point, and the position data of all static objects within the position range at the second time point. The position range can be a certain number of kilometers from the traffic accident or the range covered by the loop video data. The benchmark data is the vehicle video data of the target vehicle at the second time point. The training method is to input the input data of the sample data into the fifth machine learning model to obtain output data, and to adjust the parameters of the fifth machine learning model based on the loss value of the output data and the benchmark data until the loss value of the fifth machine learning model is within a certain range. After a large amount of training, the fifth machine learning model can generate the simulated vehicle video data of the test vehicle at the second decision time point based on the vehicle video data of the test vehicle within the second specified time period before the first decision time point, the position data of the test vehicle at the second decision time point, the position data of all dynamic objects at the second decision time point, and the position data of all static objects at the second decision time point.

[0169] The vehicle radar data in the generated simulation sensor data is introduced below. For vehicle radar data at a time point, it can be generated based on static environment data and dynamic object data in the accident scene at the time point. Specifically, based on the position data of the test vehicle at the time point, the position data of the dynamic objects, and the position data of the static objects, the relative positions of the test vehicle and each dynamic object, and the relative positions of the test vehicle and each static object are known, and then the simulation vehicle radar data at the time point is generated. In addition, similar to the above, the vehicle video data is replaced by the vehicle radar data, and the vehicle radar data is generated using the machine learning model, which will not be described in detail here.

[0170] At step 508, based on the automatic driving program of the test vehicle, the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length before the second decision time point is processed to obtain second decision information.

[0171] The processing procedure of this step is similar to that of step 308, and the introduction of step 308 can be referred to.

[0172] When the test vehicle is a non-accident vehicle in a dynamic scene, in addition to the test vehicle and the object involved in the accident, other objects move based on the simulation motion logic in the test program after the first decision point, the vehicle simulation test method is as shown in Figure 6 The method comprises the following steps:

[0173] At step 601, accident-related data within the time range and the location range corresponding to the target traffic accident is obtained.

[0174] The processing procedure of this step can be similar to that of step 201, and the introduction of step 201 can be referred to.

[0175] At step 602, the accident scene data is established based on the accident-related data.

[0176] The processing procedure of this step can be similar to that of step 202, and the introduction of step 202 can be referred to.

[0177] At step 603, the first decision time point is determined, and the first position data and the first motion state data of the test vehicle within the first specified time length before the first decision time point are determined based on the accident-related data within the first specified time length before the first decision time point.

[0178] The processing procedure of this step can be similar to that of step 503, and the introduction of step 503 can be referred to.

[0179] At step 604, based on the location data of the fisheye monitoring device corresponding to the fisheye video data, the accident-related data within the first specified time period before the first decision time point, the location data and the motion state data of the test vehicle within the first specified time period before the first decision time point, the simulated sensor data of the test vehicle within the second specified time period before the first decision time point is generated.

[0180] The processing flow of this step can be similar to that of step 204, and the introduction of step 204 can be referred to.

[0181] At step 605, based on the automatic driving program of the test vehicle, the simulated sensor data of the test vehicle within the second specified time period is processed to obtain the first decision information.

[0182] The processing flow of this step can be similar to that of step 205, and the introduction of step 205 can be referred to.

[0183] At step 606, based on the location data and the motion state data of the test vehicle at the first decision time point and the first decision information, the location data and the motion state data of the test vehicle at the second decision time point are determined.

[0184] The processing flow of this step is similar to that of step 306, and the introduction of step 306 can be referred to.

[0185] At step 607, based on the simulated motion logic in the test program, the static environment data in the accident scene data and the dynamic object data within the second specified time period before the first decision time point are processed to obtain the location data and the motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point.

[0186] The simulated motion logic can be automatically generated by a simulation program, for example, the simulation program can be a virtual test drive (VTD) program.

[0187] At step 608, based on the static environment data in the accident scene data, the location data and the motion state data of the accident vehicle at the second decision time point in the accident scene data, the location data and the motion state data of the test vehicle at the second decision time point, the location data and the motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point, and the simulated sensor data of the test vehicle within the second specified time period before the first decision time point, the simulated sensor data of the test vehicle at the second decision time point is generated.

[0188] First, the vehicle video data in the generated simulated sensor data is introduced. The sixth machine learning model can be used to generate the simulated vehicle video data of the test vehicle at the second decision time point. A large amount of training is performed on the sixth machine learning model in advance using a large amount of sample data, each of which includes input data and benchmark data. The input data is the vehicle video data of the target vehicle for a specified time length before the first time point, the position data of the target vehicle at the second time point, the position data of the accident vehicle at the second decision time point, the position data of other objects in the position range at the second time point except the target vehicle and the accident vehicle, and the position data of all static objects in the position range at the second time point. The position range can be a certain number of kilometers from the traffic accident or the range covered by the loop video data. The benchmark data is the vehicle video data of the target vehicle at the second time point. The training method is to input the input data of the sample data into the sixth machine learning model to obtain output data, and to adjust the parameters of the sixth machine learning model based on the loss value of the output data and the benchmark data until the loss value of the sixth machine learning model is within a certain range. After a large amount of training, the sixth machine learning model can generate the simulated vehicle video data of the test vehicle at the second decision time point based on the vehicle video data of the test vehicle for a second specified time length before the first decision time point, the position data of the test vehicle at the second decision time point, the position data of the accident vehicle at the second decision time point, the position data of other objects at the second decision time point except the test vehicle and the accident vehicle, and the position data of all static objects at the second decision time point.

[0189] Next, the vehicle radar data in the simulated sensor data is introduced. For vehicle radar data at a certain time point, it can be generated based on the static environment data and dynamic object data in the accident scene at that time point. Specifically, based on the position data of the test vehicle, the position data of the dynamic objects and the position data of the static objects at that time point, the relative positions of the test vehicle and each dynamic object, and the relative positions of the test vehicle and each static object are obtained, and then the simulated vehicle radar data at that time point is generated. In addition, similar to the above, the vehicle video data can be replaced by the vehicle radar data, and the machine learning model can be used to generate the vehicle radar data, which will not be described in detail here.

[0190] At step 609, the simulated sensor data of the test vehicle at the second decision time point and within a second specified time length before the second decision time point is processed based on the automatic driving program of the test vehicle to obtain second decision information.

[0191] The processing procedure of this step is similar to that of step 308, and the introduction of step 308 can be referred to.

[0192] In the embodiments of the present disclosure, based on the accident-related data, the accident scene is established, and then the test vehicle can be used as a bystander vehicle in the accident scene, i.e., a vehicle other than the accident vehicle, to verify the safety of the automatic driving program. Specifically, the simulation sensor data of the test vehicle is generated based on the accident-related data and the position data and motion state data of the test vehicle, and a decision is made based on the simulation sensor data. In this way, in the simulation scene, the test vehicle can be set as not only the accident vehicle but also any bystander vehicle, which fully utilizes the simulation scene and reduces the waste of processing resources.

[0193] Based on the same technical concept, the embodiments of the present disclosure provide a device for vehicle simulation testing, which comprises:

[0194] The acquisition module 710 is configured to acquire accident-related data within a time range and a location range corresponding to a target traffic accident, wherein the accident-related data comprises loop video data and sensor data of at least one vehicle, and the at least one vehicle comprises an accident vehicle;

[0195] The establishment module 720 is configured to establish accident scene data based on the accident-related data, wherein the accident scene data comprises static environment data and dynamic object data, the dynamic object data comprises position data and motion state data of at least one dynamic object within a target time period, and the at least one dynamic object comprises the at least one vehicle, and the target time period is within the time range;

[0196] The determination module 730 is configured to determine a first decision time point and determine position data and motion state data of a test vehicle within a first specified time period before the first decision time point, wherein the first decision time point is within the target time period, and the test vehicle is a vehicle other than the accident vehicle;

[0197] The generation module 740 is configured to generate simulation sensor data of the test vehicle within a second specified time period before the first decision time point based on position data of a loop monitoring device corresponding to the loop video data, accident-related data within the first specified time period before the first decision time point, position data and motion state data of the test vehicle within the first specified time period before the first decision time point, wherein the simulation sensor data within the second specified time period before the first decision time point comprises simulation sensor data corresponding to a plurality of time points within the second specified time period, the plurality of time points are uniformly distributed within the second specified time period, and the plurality of time points include the first decision time point;

[0198] The processing module 750 is configured to process the simulation sensor data of the test vehicle in the first specified time period based on the automatic driving program of the test vehicle, to obtain first decision information.

[0199] In a possible implementation, the sensor data includes vehicle video data, vehicle radar data, and motion state data.

[0200] In a possible implementation, the motion state data includes vehicle speed and vehicle steering angle.

[0201] In a possible implementation, the determining module 730 is configured to:

[0202] receive a test vehicle addition request;

[0203] determine the position data and the motion state data of the test vehicle in a first specified time period before the first decision time point carried in the test vehicle addition request.

[0204] In a possible implementation, the processing module 750 is configured to:

[0205] determine the position data and the motion state data of the test vehicle at a second decision time point based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, the second decision time point being after the first decision time point, and an interval between the first decision time point and the second decision time point being a decision cycle length of the automatic driving program;

[0206] generate the simulation sensor data at the second decision time point based on the position data of the fisheye monitoring device corresponding to the fisheye video data, the static environment data in the accident scene data, the position data and the motion state data of all dynamic objects in the accident scene data at the second decision time point, the position data and the motion state data of the test vehicle at the second decision time point, and the simulation sensor data of the test vehicle in a second specified time period before the first decision time point;

[0207] process the simulation sensor data of the test vehicle at the second decision time point and in the second specified time period before the second decision time point based on the automatic driving program of the test vehicle, to obtain second decision information.

[0208] In a possible implementation, the processing module 750 is configured to:

[0209] determine, based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, position data and motion state data of the test vehicle at a second decision time point, the second decision time point being after the first decision time point, and an interval time length between the first decision time point and the second decision time point being a decision cycle time length of the automatic driving program;

[0210] process, based on simulation motion logic in a test program, the static environment data in the accident scene data and the dynamic object data within a second specified time length before the first decision time point, and the position data and the motion state data of the test vehicle within the second specified time length before the first decision time point, to obtain position data and motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point;

[0211] generate, based on position data of a fisheye monitoring device corresponding to the fisheye video data, the static environment data in the accident scene data, position data and motion state data of the accident vehicle at the second decision time point in the accident scene data, position data and motion state data of the test vehicle at the second decision time point, position data and motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point, and simulation sensor data of the test vehicle within the second specified time length before the first decision time point, simulation sensor data of the test vehicle at the second decision time point;

[0212] process, based on the automatic driving program of the test vehicle, the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length before the second decision time point, to obtain second decision information.

[0213] In a possible implementation, the at least one dynamic object further includes at least one non-accident vehicle, and the test vehicle is a vehicle in the at least one non-accident vehicle;

[0214] The determination module 730 is configured to:

[0215] determine, based on accident-related data of the test vehicle within a first specified time length before the first decision time point, first position data and first motion state data of the test vehicle within the first specified time length before the first decision time point.

[0216] In a possible implementation, an interval time length between two adjacent time points is equal to a decision cycle time length of the automatic driving program, and the processing module 750 is further configured to:

[0217] determine, based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, position data and motion state data of the test vehicle at a second decision time point, the second decision time point being after the first decision time point, and an interval time length between the first decision time point and the second decision time point being a decision cycle time length of the automatic driving program;

[0218] generate, based on position data of a fisheye monitoring device corresponding to the fisheye video data, static environment data in the accident scene data, position data and motion state data of dynamic objects other than the test vehicle in the accident scene data at the second decision time point, position data and motion state data of the test vehicle at the second decision time point, and simulation sensor data of the test vehicle within a second specified time length before the first decision time point, simulation sensor data of the test vehicle at the second decision time point;

[0219] process, based on the automatic driving program of the test vehicle, the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length before the second decision time point, to obtain second decision information.

[0220] In a possible implementation, the processing module 750, when the interval time length between two adjacent time points is equal to the decision cycle time length of the automatic driving program, is further configured to:

[0221] determine, based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, position data and motion state data of the test vehicle at a second decision time point, the second decision time point being after the first decision time point, and an interval time length between the first decision time point and the second decision time point being a decision cycle time length of the automatic driving program;

[0222] process, based on simulation motion logic in the test program, the static environment data in the accident scene data and the dynamic object data within the second specified time length before the first decision time point, to obtain position data and motion state data of other objects other than the test vehicle and the accident vehicle at the second decision time point;

[0223] generate the simulation sensor data of the test vehicle at the second decision time point based on the position data of the fisheye video data corresponding to the fisheye monitoring device, the static environment data in the accident scene data, the position data and the motion state data of the accident vehicle at the second decision time point in the accident scene data, the position data and the motion state data of the test vehicle at the second decision time point, the position data and the motion state data of the other objects except the test vehicle and the accident vehicle at the second decision time point, and the simulation sensor data of the test vehicle within the second specified time length before the first decision time point;

[0224] process the simulation sensor data of the test vehicle at the second decision time point and within the second specified time length before the second decision time point based on the automatic driving program of the test vehicle to obtain second decision information.

[0225] In the embodiments of the present disclosure, after the accident scene is established based on the accident-related data, the test vehicle can be set as a bystander vehicle in the accident scene, i.e., a vehicle other than the accident vehicle, to further verify the safety of the automatic driving program. Specifically, the simulation sensor data is generated based on the accident-related data and the position data and the motion state data of the test vehicle, and the decision is made based on the simulation sensor data. In this way, in the simulation scene, the test vehicle can be set as not only the accident vehicle but also any bystander vehicle, which fully utilizes the simulation scene and reduces the waste of processing resources.

[0226] It should be noted that the vehicle simulation test device provided in the above embodiments is only used as an example to divide the above functional modules when performing the vehicle simulation test processing. In actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the vehicle simulation test device and the vehicle simulation test method provided in the above embodiments belong to the same concept, and the specific implementation process is described in detail in the method embodiments, which will not be repeated here.

[0227] In the above embodiments, all or part of the steps can be implemented by software, hardware, firmware or any combination thereof, when implemented by using software, all or part of the steps can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions loaded and executed on a device, which generates the processes or functions described in the embodiments of the present disclosure. The computer instructions can be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line) or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a device or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available media can be magnetic media (such as floppy disk, hard disk and magnetic tape, etc.), optical media (such as digital video disk (digital video disk, DVD), etc.), or semiconductor media (such as solid state disk, etc.).

[0228] Those skilled in the art can understand that all or part of the steps of the above embodiments can be completed by hardware, or by program instructing relevant hardware, and the program can be stored in a computer readable storage medium, and the storage medium mentioned above can be read only memory, disk or optical disk, etc.

[0229] The above only describes one embodiment of the present disclosure, and does not limit the present disclosure, any modification, equivalent replacement, improvement, etc. within the principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method of vehicle simulation testing, characterized by, The method comprises: acquiring accident-related data in a time range and a location range corresponding to a target traffic accident, wherein the accident-related data comprises fisheye video data and sensor data of at least one vehicle, and the at least one vehicle comprises an accident vehicle; based on the accident-related data, establishing accident scene data, wherein the accident scene data comprises static environment data and dynamic object data, the dynamic object data comprises position data and motion state data of at least one dynamic object in a target time period, and the at least one dynamic object comprises the at least one vehicle, wherein the target time period is within the time range; determining a first decision time point and determining position data and motion state data of a test vehicle within a first specified time period before the first decision time point, wherein the first decision time point is within the target time period, and the test vehicle is a vehicle other than the accident vehicle; based on position data of a fisheye monitoring device corresponding to the fisheye video data, accident-related data within a first specified time period before the first decision time point, position data and motion state data of the test vehicle within the first specified time period before the first decision time point, generating simulated sensor data of the test vehicle within a second specified time period before the first decision time point, wherein the simulated sensor data within the second specified time period comprises simulated sensor data corresponding to a plurality of time points within the second specified time period, the plurality of time points are uniformly distributed within the second specified time period and include the first decision time point, and the first specified time period is greater than the second specified time period; processing the simulated sensor data of the test vehicle within the second specified time period based on an automatic driving program of the test vehicle to obtain first decision information; the interval time length of adjacent two time points is equal to the decision cycle time length of the automatic driving program; after the simulated sensor data of the test vehicle within the second specified time period is processed based on the automatic driving program of the test vehicle to obtain the first decision information, the method further comprises: based on the position data and motion state data of the test vehicle at the first decision time point and the first decision information, determining position data and motion state data of the test vehicle at a second decision time point, wherein the second decision time point is after the first decision time point, and the interval time length between the first decision time point and the second decision time point is the decision cycle time length of the automatic driving program; based on the static environment data in the accident scene data, the position data and motion state data of all dynamic objects in the accident scene data at the second decision time point, the position data and motion state data of the test vehicle at the second decision time point, and the simulated sensor data of the test vehicle within the second specified time period before the first decision time point, generating simulated sensor data at the second decision time point; The simulation sensor data of the test vehicle within the second specified time length is processed based on the automatic driving program of the test vehicle to obtain second decision information.

2. The method of claim 1, wherein, The sensor data includes vehicle video data, vehicle radar data, and motion state data.

3. The method of claim 2, wherein, The motion state data includes vehicle speed and vehicle steering angle.

4. The method of claim 3, wherein, The position data and motion state data of the test vehicle within the first specified time length before the first decision time point are determined, including: receiving a test vehicle addition request; determining the position data and motion state data of the test vehicle within the first specified time length before the first decision time point carried in the test vehicle addition request.

5. The method of claim 4, wherein, After the simulation sensor data of the test vehicle within the second specified time length is processed based on the automatic driving program of the test vehicle to obtain the first decision information, the second decision information is further obtained by the following steps: Based on the position data and motion state data of the test vehicle at the first decision time point and the first decision information, the position data and motion state data of the test vehicle at the second decision time point are determined, the second decision time point is after the first decision time point, and the interval time length between the first decision time point and the second decision time point is the decision cycle time length of the automatic driving program; Based on the simulation motion logic in the test program, the static environment data in the accident scene data and the dynamic object data within the second specified time length before the first decision time point, and the position data and motion state data of the test vehicle within the second specified time length before the first decision time point are processed to obtain the position data and motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point; Based on the static environment data in the accident scene data, the position data and motion state data of the accident vehicle at the second decision time point in the accident scene data, the position data and motion state data of the test vehicle at the second decision time point, the position data and motion state data of other objects except the test vehicle and the accident vehicle at the second decision time point, and the simulation sensor data of the test vehicle within the second specified time length before the first decision time point, the simulation sensor data of the test vehicle at the second decision time point is generated; The simulation sensor data of the test vehicle within the second specified time length before the second decision time point is processed based on the automatic driving program of the test vehicle to obtain second decision information.

6. An apparatus for vehicle simulation testing, characterized by The device comprises: An acquisition module is configured to acquire accident-related data within a time range and a location range corresponding to a target traffic accident, wherein the accident-related data includes fisheye video data and sensor data of at least one vehicle, and the at least one vehicle includes an accident vehicle; The establishment module is configured to establish accident scene data based on the accident-related data, wherein the accident scene data comprises static environment data and dynamic object data, the dynamic object data comprises position data and motion state data of at least one dynamic object in a target time period, the at least one dynamic object comprises the at least one vehicle, and the target time period is within the time range; The determination module is configured to determine a first decision time point and determine position data and motion state data of a test vehicle in a first specified time period before the first decision time point, wherein the first decision time point is within the target time period, and the test vehicle is a vehicle other than the accident vehicle; The generation module is configured to generate simulation sensor data of the test vehicle in a second specified time period before the first decision time point based on position data of a turret monitoring device corresponding to the turret video data, accident-related data in the first specified time period before the first decision time point, position data and motion state data of the test vehicle in the first specified time period before the first decision time point, wherein the simulation sensor data in the second specified time period before the first decision time point comprises simulation sensor data corresponding to a plurality of time points in the second specified time period, the plurality of time points are uniformly distributed in the second specified time period and comprise the first decision time point, and the first specified time period is greater than the second specified time period; The processing module is configured to process the simulation sensor data of the test vehicle in the second specified time period based on an automatic driving program of the test vehicle to obtain first decision information; The interval time length of adjacent two time points is equal to the decision cycle time length of the automatic driving program, and the processing module is further configured to: determine position data and motion state data of the test vehicle at a second decision time point based on the position data and the motion state data of the test vehicle at the first decision time point and the first decision information, wherein the second decision time point is after the first decision time point, and the interval time length between the first decision time point and the second decision time point is the decision cycle time length of the automatic driving program; generate simulation sensor data at the second decision time point based on the static environment data in the accident scene data, position data and motion state data of all dynamic objects in the accident scene data at the second decision time point, position data and motion state data of the test vehicle at the second decision time point, and simulation sensor data of the test vehicle in the second specified time period before the first decision time point; and process the simulation sensor data of the test vehicle at the second decision time point and in the second specified time period before the second decision time point based on the automatic driving program of the test vehicle to obtain second decision information.

7. An electronic device, comprising: The electronic device comprises a memory and a processor, and the memory is configured to store computer instructions. The processor executes the computer instructions stored in the memory, so that the electronic device executes the method in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer program code, and when the computer program code is executed by an electronic device, the electronic device executes the method in any one of claims 1-5.

9. A computer program product, characterised in that, The computer program product includes computer program code, and when the computer program code is executed by an electronic device, the electronic device executes the method in any one of claims 1-5.

Images