Scene construction method and device, computer equipment and readable storage medium

Abstract

The invention relates to a scene construction method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a road structure data set and a traffic participant motion data set of a crossroad; the traffic participant motion data set comprises a motion data set of each vehicle and each reference object passing through the crossroad; based on the road structure data set and the traffic participant motion data set, extracting driving features of each vehicle passing through the crossroad to obtain a driving feature set of the vehicle, and based on the driving feature set and an agent model, determining an estimated offset trajectory sequence set of the vehicle; the estimated offset track sequence set is used for dynamically adjusting the driving track of the vehicle; and obtaining a dynamic simulation scene of the crossroad based on each estimated offset trajectory sequence set, the traffic participant motion data set and the road structure data set. By adopting the method, the authenticity of a dynamic simulation scene and the test accuracy of automatic driving can be improved.

Description

Technical Field

[0001] The present application relates to the field of vehicle testing technology, and in particular to a scenario construction method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the rapid development of autonomous driving technology, it is necessary to use scenario construction methods to construct simulation scenarios when vehicles pass through intersections. Through this simulation scenario, the behavior of the vehicle passing through the intersection when in the autonomous driving stage is tested on a computer device.

[0003] The current scene construction method collects the road structure dataset of the real intersection and the natural driving data of each vehicle passing through the intersection within a preset time period. The natural driving data only contains the fixed driving trajectory of the vehicle passing through the intersection. The road structure dataset and each fixed driving trajectory are combined into a simulated scene of the intersection.

[0004] However, the current scenario construction method, based on fixed driving trajectories and fixed road structure data sets, constructs a simulation scenario that is too idealized. This simulation scenario has great limitations, resulting in low accuracy in the vehicle's autonomous driving test. Summary of the invention

[0005] Based on this, it is necessary to provide a scene construction method, device, computer equipment, computer-readable storage medium and computer program product to address the above technical problems.

[0006] In a first aspect, the present application provides a scene construction method, comprising:

[0007] Acquire a road structure dataset and a traffic participant motion dataset of an intersection; the traffic participant motion dataset includes a motion dataset of each vehicle and each reference object passing through the intersection;

[0008] Extracting driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant motion dataset to obtain a driving characteristic set of the vehicle, and determining an estimated offset trajectory sequence set of the vehicle based on the driving characteristic set and an agent model; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle;

[0009] Based on each of the estimated offset trajectory sequence sets, the traffic participant motion data set and the road structure data set, a dynamic simulation scene of the intersection is obtained.

[0010] In one embodiment, obtaining a road structure dataset of an intersection includes:

[0011] Acquire an initial road structure data set of an intersection; the initial road structure data set includes lane lines and boundary lines of the intersection;

[0012] Determine each lane of the intersection based on each lane line and each boundary line, and mark each lane;

[0013] Traversing each lane, establishing a reference driving trajectory between the lane and other lanes with the lane as a starting point, and obtaining a reference driving trajectory set;

[0014] The initial road structure data set is updated according to the reference driving trajectory set and each of the lanes to obtain a road structure data set.

[0015] In one embodiment, the road structure dataset includes a reference driving trajectory set, the driving features include a traffic intention feature, an interaction feature, and an offset feature, and the driving features of each vehicle passing through the intersection are extracted based on the road structure dataset and the traffic participant motion dataset to obtain the driving feature set of the vehicle, including:

[0016] For each motion data set of the vehicle, determining the motion data set as a main motion data set;

[0017] Screening each reference vehicle and each reference object that appear simultaneously with the vehicle in the traffic participant motion data set, and determining the motion data set of each reference vehicle and the motion data set of each reference object as each secondary motion data set;

[0018] Extracting the traffic intention feature of the vehicle according to the traffic intention rule and the main motion data set, and determining the reference driving trajectory corresponding to the traffic intention feature in the reference driving trajectory set;

[0019] According to the road structure data set, the reference driving trajectory, the main motion data set and each of the secondary motion data sets, the interaction features and the offset features of the vehicle are extracted, and the driving feature set of the vehicle is constructed according to the traffic intention features, the interaction features and the offset features.

[0020] In one embodiment, the road structure data set includes each boundary line and stop line, the main motion data set includes a main motion data subset at each acquisition time, and the extraction of the interaction feature and the offset feature of the vehicle according to the road structure data set, the reference driving trajectory, the main motion data set and each of the secondary motion data sets includes:

[0021] For each of the main motion data subsets, performing an offset operation on the stop line, the reference driving trajectory and the actual position data in the main motion data subset according to a lateral and longitudinal offset algorithm to obtain an offset feature;

[0022] Determining, in each of the secondary motion data sets, a secondary motion data subset corresponding to the primary motion data subset;

[0023] Based on the interactive feature algorithm, feature extraction is performed on the main motion data subset, each of the secondary motion data subsets and each boundary line to obtain the interactive feature of the vehicle.

[0024] In one embodiment, constructing the driving feature set of the vehicle according to the traffic intention feature, the interaction feature and the offset feature includes:

[0025] Constructing a driving feature subset of the vehicle at the collection time based on each of the interaction features, the offset feature corresponding to the interaction feature, and the traffic intention feature;

[0026] A driving feature set of the vehicle is constructed according to each of the driving feature subsets.

[0027] In one embodiment, determining the estimated deviation trajectory sequence set of the vehicle based on the driving feature set and the agent model includes:

[0028] dividing each driving feature subset in the driving feature set into driving feature subset groups;

[0029] Performing prediction processing on each of the driving feature subset groups based on the proxy model to obtain an initial estimated deviation trajectory sequence;

[0030] Coordinate transformation processing is performed on the initial estimated offset trajectory sequence to obtain an estimated offset trajectory sequence, and an estimated offset trajectory sequence set is constructed according to each of the estimated offset trajectory sequences.

[0031] In a second aspect, the present application also provides a scene construction device, including:

[0032] An acquisition module, used to acquire a road structure data set and a traffic participant motion data set of an intersection; the traffic participant motion data set includes a motion data set of each vehicle and each reference object passing through the intersection;

[0033] an extraction module, configured to extract driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant motion dataset, obtain a driving characteristic set of the vehicle, and determine an estimated offset trajectory sequence set of the vehicle based on the driving characteristic set and an agent model; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle;

[0034] A combination module is used to obtain a dynamic simulation scene of the intersection based on each of the estimated offset trajectory sequence sets, the traffic participant motion data set and the road structure data set.

[0035] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0036] Acquire a road structure dataset and a traffic participant motion dataset of an intersection; the traffic participant motion dataset includes a motion dataset of each vehicle and each reference object passing through the intersection;

[0037] Extracting driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant motion dataset to obtain a driving characteristic set of the vehicle, and determining an estimated offset trajectory sequence set of the vehicle based on the driving characteristic set and an agent model; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle;

[0038] Based on each of the estimated offset trajectory sequence sets, the traffic participant motion data set and the road structure data set, a dynamic simulation scene of the intersection is obtained.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0040] Acquire a road structure dataset and a traffic participant motion dataset of an intersection; the traffic participant motion dataset includes a motion dataset of each vehicle and each reference object passing through the intersection;

[0041] Extracting driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant motion dataset to obtain a driving characteristic set of the vehicle, and determining an estimated offset trajectory sequence set of the vehicle based on the driving characteristic set and an agent model; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle;

[0042] Based on each of the estimated offset trajectory sequence sets, the traffic participant motion data set and the road structure data set, a dynamic simulation scene of the intersection is obtained.

[0043] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0044] Acquire a road structure dataset and a traffic participant motion dataset of an intersection; the traffic participant motion dataset includes a motion dataset of each vehicle and each reference object passing through the intersection;

[0045] Extracting driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant motion dataset to obtain a driving characteristic set of the vehicle, and determining an estimated offset trajectory sequence set of the vehicle based on the driving characteristic set and an agent model; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle;

[0046] Based on each of the estimated offset trajectory sequence sets, the traffic participant motion data set and the road structure data set, a dynamic simulation scene of the intersection is obtained.

[0047] The above-mentioned scene construction method, device, computer equipment, computer-readable storage medium and computer program product obtain a road structure data set and a traffic participant motion data set of an intersection; the traffic participant motion data set includes motion data sets of each vehicle and each reference object passing through the intersection; based on the road structure data set and the traffic participant motion data set, the driving characteristics of each vehicle passing through the intersection are extracted to obtain the driving characteristics set of the vehicle, and the estimated offset trajectory sequence set of the vehicle is determined based on the driving characteristics set and the proxy model; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle; based on each of the estimated offset trajectory sequence sets, the traffic participant motion data set and the road structure data set, a dynamic simulation scene of the intersection is obtained. By adopting this method, by extracting the driving characteristics of the vehicle passing through the intersection, and determining the estimated offset trajectory sequence set of the vehicle through the proxy model and the driving characteristics set, the driving trajectory can be dynamically adjusted based on the estimated offset trajectory sequence set. Furthermore, a dynamic simulation scene can be obtained based on each offset trajectory sequence set, the traffic participant motion data set and the road structure data set, thereby improving the authenticity of the dynamic simulation scene and expanding the applicability of the dynamic simulation scene. Testing autonomous driving vehicles based on this dynamic simulation scenario can achieve two-way avoidance between autonomous driving vehicles and manually driven vehicles, improving the accuracy of autonomous driving testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 A schematic diagram of a flow chart of a scene construction method in an embodiment;

[0050] Figure 2 A schematic diagram of a process for obtaining a road structure dataset in one embodiment;

[0051] Figure 3 is a schematic diagram of a process of extracting driving features in one embodiment;

[0052] Figure 4 A schematic diagram of a process for extracting interaction features and offset features in one embodiment;

[0053] Figure 5 A schematic diagram of a process for constructing a driving feature set in one embodiment;

[0054] Figure 6 A schematic diagram of a process for determining an estimated offset trajectory sequence set in one embodiment;

[0055] Figure 7 A schematic diagram of a process for training an initial proxy model in one embodiment;

[0056] Figure 8 An architectural diagram of a scene construction device in one embodiment;

[0057] Fig. 9 A schematic diagram of a motor vehicle trajectory of a static simulation scenario and a motor vehicle trajectory of a dynamic simulation scenario in one embodiment;

[0058] Fig.10 is a schematic diagram of the architecture of a training module in one embodiment;

[0059] Fig.11 A structural block diagram of a scene construction device in one embodiment;

[0060] Fig.12 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] In one embodiment, Figure 1 As shown, a scene construction method is provided. The embodiment of the present application takes the method applied to a computer device as an example for explanation. The embodiment of the present application does not limit the execution device of the scene construction method, and includes the following steps 102 to 106:

[0063] Step 102: Obtain a road structure dataset and a traffic participant movement dataset at the intersection.

[0064] The traffic participant motion dataset includes the motion datasets of each vehicle and each reference object passing through the intersection.

[0065] In the implementation, the computer device obtains the initial road structure data set of the intersection, and preprocesses the initial road structure data set to obtain the road structure data set. At the same time, the computer device obtains the motion data set of each traffic participant passing through the intersection based on the preset acquisition conditions to obtain the traffic participant motion data set. Among them, the traffic participants include each vehicle and each reference object.

[0066] Specifically, the acquisition condition is a first time period. The computer device acquires all lane lines and all boundary lines of the intersection, and acquires the traffic light phase of the intersection. The computer device constructs an initial road structure data set of the intersection according to each lane line, each boundary line and the traffic light phase. The computer device determines each lane and a reference driving trajectory set based on each lane line and each boundary line, and updates the initial road structure data set according to each lane and the reference driving trajectory set to obtain a road structure data set. Then, the computer device acquires a motion data set of each traffic participant passing through the intersection within a preset period to obtain a traffic participant data set.

[0067] In an exemplary embodiment, the collection device collects the motion data set of each traffic participant passing through the intersection at each collection time according to a preset time interval and a first time period. The motion data set includes the position, speed, acceleration and attribute information of the traffic participant at each collection time. The driving trajectory of the traffic participant passing through the intersection can be constructed based on the actual position data of the traffic participant at each collection time. The collection device transmits the motion data set of each traffic participant at each collection time to the computer device. The computer device receives the motion data set of each traffic participant at each collection time to obtain the traffic participant data set.

[0068] Optionally, the reference object may be, but is not limited to, a non-motorized vehicle and a human being. For non-motorized vehicles such as bicycles and battery vehicles, the acquisition condition may be set to the first time period or the number of traffic participants. The embodiment of the present application does not limit the reference object and the acquisition condition.

[0069] Step 104 , extracting driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant motion dataset to obtain a driving characteristic set of the vehicle, and determining an estimated offset trajectory sequence set of the vehicle based on the driving characteristic set and the agent model.

[0070] The estimated deviation trajectory sequence set is used to dynamically adjust the vehicle's driving trajectory. Driving features include deviation features, driving intention features, and interaction features.

[0071] In the implementation, the computer device extracts the driving intention feature, deviation feature and interaction feature of each vehicle passing through the intersection based on the road structure data set and the traffic participant movement data set to obtain the driving feature set of the vehicle. Then, the computer device predicts the driving feature set based on the agent model to obtain the estimated deviation estimation sequence set of the vehicle.

[0072] Specifically, the computer device determines the motion data set as the main motion data set for each vehicle, and screens the sub-motion data sets corresponding to the main motion data set in the traffic participant data set. Then, the computer device extracts the traffic intention feature, offset feature and interaction feature of the vehicle passing through the intersection based on the active data set, each sub-motion data set and the road structure data set to obtain the vehicle's driving feature set. Among them, the traffic intention feature represents the traffic direction of the traffic participant. The interaction feature is the interaction feature between the traffic participant and the reference object or other vehicles or the intersection environment. The offset feature is the offset between the vehicle's driving trajectory and the reference driving trajectory.

[0073] Step 106 , based on each estimated offset trajectory sequence set, the traffic participant motion data set and the road structure data set, a dynamic simulation scene of the intersection is obtained.

[0074] Among them, the dynamic simulation scene of the intersection is used to test the vehicle's autonomous driving.

[0075] In implementation, a computer device combines each estimated offset trajectory sequence set, an interactive participant motion data set, and a road structure data set to obtain a dynamic simulation scene of an intersection.

[0076] In an exemplary embodiment, a computer device performs dynamic simulation based on a dynamic simulation scene of an intersection to obtain a simulation model. The computer device adds an autonomous driving vehicle to the simulation model. When the driving trajectory of the autonomous driving vehicle conflicts with that of a target vehicle among traffic participants, the target vehicle will adjust its driving trajectory based on the target estimated offset trajectory sequence in the target estimated offset trajectory sequence set corresponding to the target vehicle, thereby avoiding the autonomous driving vehicle. At the same time, the autonomous driving vehicle will also adjust the driving trajectory of the autonomous driving vehicle to avoid the target vehicle. After avoiding the autonomous driving vehicle, the target vehicle will adjust its driving trajectory based on the target estimated offset trajectory sequence in the target estimated offset trajectory sequence set, and adjust the driving trajectory of the target vehicle again, thereby achieving the target vehicle's traffic intention.

[0077] In the above scenario construction method, by extracting the driving characteristics of the vehicle passing through the intersection and determining the estimated offset trajectory sequence set of the vehicle through the proxy model and the driving characteristic set, the driving trajectory can be dynamically adjusted based on the estimated offset trajectory sequence set. Furthermore, a dynamic simulation scene can be obtained based on each offset trajectory sequence set, the traffic participant motion data set and the road structure data set, which improves the authenticity of the dynamic simulation scene and expands the applicability of the dynamic simulation scene. Testing the autonomous driving vehicle based on the dynamic simulation scene can achieve two-way avoidance of the autonomous driving vehicle and the manually driven vehicle, and improve the accuracy of the autonomous driving test.

[0078] In an exemplary embodiment, Figure 2 As shown, the specific processing process of obtaining the road structure data set of the intersection in step 102 includes steps 202 to 208. Among them:

[0079] Step 202: Acquire an initial road structure dataset of the intersection.

[0080] Among them, the initial road structure dataset contains lane lines and boundary lines at intersections.

[0081] During implementation, the computer device obtains each lane line, each boundary line and each stop line of the intersection. At the same time, the computer device obtains the traffic light phase in the intersection. The traffic light phase refers to the different display states of the traffic light, which is used to indicate when vehicles and pedestrians can pass through the intersection. The phase is generally divided into four stages, and each stage has a different indicator light display. The stop line is a solid white line located at the entrance of the intersection. The function of the stop line is to indicate the parking position of the vehicle when waiting for the traffic light. The vehicle must wait behind the stop line and cannot exceed or press the stop line. The computer device combines each lane line, each boundary line, each stop line and traffic light phase to obtain an initial road structure data set.

[0082] Specifically, the computer device draws each lane line, each boundary line and each stop line of the intersection, and obtains the phase of the traffic light. Specifically, the lane line, boundary line and stop line are all obtained by segmented fitting of discrete points. The computer device approximates the non-interactive area of ​​the intersection as a straight line y=kx+b, and the arc of the straight line connection in the interactive area is a standard circular arc line. Approximate fitting is performed to obtain the fitted lane lines, boundary lines and stop lines.

[0083] Step 204 , determining each lane of the intersection based on each lane line and each boundary line, and marking each lane.

[0084] In implementation, the computer device draws each road at the intersection based on the boundary line, and divides each road into lanes based on each lane line to obtain lanes corresponding to each road. Then, the computer device marks each lane based on the direction of the lane to obtain the name of the lane.

[0085] In an exemplary embodiment, the computer device draws four roads in the east, west, south and north of the intersection based on the boundary line, and draws a circular interactive area. Then, the computer device divides each road according to the lane line to obtain each lane. Among them, there are six lanes on the eastbound road and the westbound road, and there are four lanes on the southbound road and the northbound road. The computer device names the six lanes of the eastbound road as East Lane 1, East Lane 2, East Lane 3, East Lane 4, East Lane 5 and East Lane 6 in order from north to south, and names the lanes on the westbound road as West Lane 1, West Lane 2, West Lane 3, West Lane 4, West Lane 5 and West Lane 6. The computer device names the four lanes of the northbound road as North Lane 1, North Lane 2, North Lane 3 and North Lane 4 in order from north to west, and names the four lanes on the southbound road as South Lane 1, South Lane 2, South Lane 3 and South Lane 4.

[0086] Step 206 , traverse each lane, establish a reference driving trajectory between the lane and other lanes with the lane as the starting point, and obtain a reference driving trajectory set.

[0087] In implementation, the computer device determines the lane centerline for each lane based on the lane line and the boundary line. Then, starting from the lane centerline, the computer device connects the lane centerline with the lane centerlines of other lanes to obtain a reference driving trajectory from the lane to other lanes.

[0088] For example, if the lane is a middle lane, the computer device determines the lane centerline of the lane based on the lane line. If the lane is an edge lane, the computer device determines the lane centerline based on the lane line and the boundary line.

[0089] In an exemplary embodiment, the lanes include East Lane 1, East Lane 2, East Lane 3, East Lane 4, East Lane 5 and East Lane 6 of the eastbound road, West Lane 1, West Lane 2, West Lane 3, West Lane 4, West Lane 5 and West Lane 6 of the westbound road, North Lane 1, North Lane 2, North Lane 3 and North Lane 4 of the northbound road, and South Lane 1, South Lane 2, South Lane 3 and South Lane 4 of the southbound road. For East Lane 1, the computer device starts from the lane centerline of East Lane 1, connects the lane centerline of East Lane 1 and the lane centerline of West Lane 1, and obtains a reference driving trajectory of the vehicle from East Lane 1 to West Lane 1. Then, the computer device draws the reference driving trajectory from East Lane 1 to West Lane 2, West Lane 3, West Lane 4, West Lane 5, West Lane 6, South Lane 1, South Lane 2, South Lane 3, South Lane 4, North Lane 1, North Lane 2, North Lane 3 and North Lane 4 according to the above execution process.

[0090] Step 208 , updating the initial road structure data set according to the reference driving trajectory set and each lane to obtain a road structure data set.

[0091] In implementation, the computer device adds the reference driving trajectory set and each lane to the initial road structure data set to obtain the road structure data set.

[0092] In an exemplary embodiment, the computer device constructs a static simulation scenario based on a reference driving trajectory set, each lane, and an initial road structure data set.

[0093] In this embodiment, the initial road structure dataset is preprocessed to obtain a reference driving trajectory set and each lane, which facilitates the subsequent extraction of the vehicle's traffic intention features and offset features. In addition, the initial road result dataset is updated based on the reference driving trajectory set and each lane, making the structure of the intersection clearer and improving the accuracy of the road result dataset.

[0094] In an exemplary embodiment, the road structure data set includes a reference driving trajectory set, and the driving features include a traffic intention feature, an interaction feature, and an offset feature, such as Figure 3 As shown, the specific processing process of extracting the driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant movement dataset in step 104 to obtain the vehicle driving characteristic set to obtain the road structure dataset of the intersection includes steps 302 to 308. Among them:

[0095] Step 302 : for each motion data set of the vehicle, determine the motion data set as a main motion data set.

[0096] In implementation, the computer device determines the motion data set of each vehicle as the main motion data set. By determining the motion data set of the vehicle as the main motion data set, the driving characteristics of the vehicle can be determined from the perspective of the vehicle.

[0097] Step 304 , screening each reference vehicle and each reference object that appear simultaneously with the vehicle in the traffic participant motion data set, and determining the motion data set of each reference vehicle and the motion data set of each reference object as each secondary motion data set.

[0098] In implementation, the computer device determines the second time period in which the vehicle appears, and selects reference vehicles and reference objects in the second time period from among the traffic participants in the traffic participant motion data set. Then, the computer device determines the motion data set of each reference lane and the motion data set of each reference object in the traffic participant motion data set as each secondary motion data set. By selecting the reference vehicles and reference objects that appear at the same time as the vehicle, the reference objects and reference vehicles that may interact with the vehicle are obtained.

[0099] In an exemplary embodiment, the computer device determines the start time of the vehicle's appearance and the end time of the lane's disappearance, and determines the second time period of the vehicle's appearance based on the start time and the end time. If the start time of the vehicle's appearance is 16:24:41 on October 16, 2024, and the end time of its disappearance is 16:25:56 on October 16, 2024. The computer device determines October 16, 2024, 16:24:41 to October 16, 2024, 16:25:56 as the second time period of the vehicle, and selects each reference vehicle and each reference object within the second time period from each traffic parameter in the traffic participant motion data set. The computer device determines the motion data sets of each reference vehicle and each reference object as each secondary motion data set.

[0100] Step 306 , extracting the vehicle's traffic intention features according to the traffic intention rule and the main motion data set, and determining a reference driving trajectory corresponding to the traffic intention features in the reference driving trajectory set.

[0101] The main motion data set includes the driving trajectory of the vehicle passing through the intersection. The driving trajectory includes the position of the vehicle at each collection time.

[0102] In implementation, the computer device determines the vehicle's entry point and exit point through the intersection based on the driving trajectory in the main motion data set. The entry point is the starting point for the vehicle to enter the intersection, and the exit point is the end point for the vehicle to leave the intersection. The computer device determines the primary traffic intention and the secondary traffic intention of the vehicle based on the relative direction between the entry point and the exit point. Then, the computer device determines the entry lane where the entry point is located, and determines the exit lane where the exit point is located. Then, the computer device constructs the vehicle's third-level traffic intention based on the entry lane and the exit lane. The computer device constructs the vehicle's traffic intention based on the primary traffic intention, the secondary traffic intention, and the third-level traffic intention. Then, the computer device determines the reference driving trajectory corresponding to the third-level traffic intention in the reference driving trajectory set.

[0103] Specifically, the first-level traffic intention includes turning left, going straight, and turning right. The second-level traffic intention is the driving direction of the vehicle. For example, from east to west or from south to east, etc. The third-level traffic intention is the vehicle's entry lane and exit lane. The computer device determines the vehicle's entry point and exit point in the driving trajectory. Then, the computer device determines the entry lane where the entry point is located based on the lane range of each lane, and determines the exit lane where the exit point is located. Then, the computer device determines the vehicle's first-level traffic intention and second-level traffic intention based on the relative positions of the entry lane and the exit lane. The computer device constructs the vehicle's third-level traffic intention based on the entry lane and the exit lane.

[0104] For example, if the entry lane is East Lane 1 and the exit lane is West Lane 1, the computer device determines that the primary traffic intention of the vehicle is to go straight, and the secondary traffic intention of the vehicle is from east to west. Then, the computer device determines East Lane 1-West Lane 1 as the third-level traffic intention of the vehicle. Then, the computer device determines the traffic intention of the vehicle as going straight, from east to west, and East Lane 1-West Lane 1. Then, the computer device determines the reference driving trajectory from East Lane 1 to West Lane 1 in the reference driving trajectory set.

[0105] Step 308, extracting the interaction features and offset features of the vehicle based on the road structure data set, the reference driving trajectory, the main motion data set and each secondary motion data set, and constructing the vehicle's driving feature set based on the traffic intention features, the interaction features and the offset features.

[0106] The road structure dataset includes boundary lines and stop lines.

[0107] In implementation, the computer device determines the offset of the vehicle relative to the reference driving trajectory according to the stop line, the reference driving trajectory and the main motion data subset, and obtains the offset feature of the vehicle. Then, the computer device extracts the interaction feature of the vehicle according to each boundary line, the main motion data set and each secondary motion data set. The computer device constructs the driving feature set of the vehicle according to the communication intention feature, the interaction feature, the offset feature and the attribute information of the vehicle.

[0108] In this embodiment, the driving purpose of the vehicle is clarified by extracting the driving intention features of the vehicle, and the relationship between the offset of the vehicle and the intersection environment is clarified by extracting the interaction features and offset features of the vehicle, thereby obtaining a driving feature set of the vehicle during the driving process.

[0109] In an exemplary embodiment, the road structure data set includes various boundary lines and stop lines, and the main motion data set includes a main motion data subset at each acquisition time, such as Figure 4 As shown, the specific processing process of extracting the interaction features and offset features of the vehicle according to the road structure data set, the reference driving trajectory, the main motion data set and each secondary motion data set in step 308 includes steps 402 to 406. Among them:

[0110] Step 402 : for each main motion data subset, an offset operation is performed on the stop line, the reference driving trajectory and the actual position data in the main motion data subset according to the lateral and longitudinal offset algorithms to obtain an offset feature.

[0111] The vehicle's driving trajectory is composed of the actual position data of the vehicle at each acquisition time. The offset feature includes the offset direction, offset distance, and the distance between the vehicle and the stop line.

[0112] In implementation, the computer device calculates the offset of the reference driving trajectory and the actual position data according to the lateral and longitudinal offset algorithms for each subset of the main motion data, and obtains the offset direction and offset distance of the vehicle. Then, the computer device determines the distance between the vehicle and the stop line according to the stop line offset algorithm. The computer device combines the distance between the vehicle and the stop line, the offset direction and the offset distance of the vehicle to obtain the offset characteristics of the vehicle.

[0113] Specifically, the actual position data is point P. The computer device locates the point N closest to P based on vector operations. Specifically, the computer device uses the reference driving trajectory as the center line of the Frenet coordinate system and performs coordinate system transformation to obtain the reference driving trajectory after coordinate transformation. The computer device connects point P with the starting point S of each segment on the reference driving trajectory to form a vector The computer device constructs two adjacent points S and E (starting point and end point) on the reference driving trajectory as a vector vector Identifies the local direction of the reference driving trajectory. The computer device calculates the vector exist The projection length on the reference trajectory is obtained by calculating the foot N of the perpendicular to the point P on the reference trajectory segment. The calculation formula of the projection length is shown in the following formula (1):

[0114]

[0115] In the above formula (1), l is the projection length ratio, · represents the vector dot product, is a vector The value of l is limited to the range of [0, 1] to ensure that point N falls on the line segment formed by S and E. Then, the computer device determines the point N closest to P based on a linear interpolation algorithm. The linear interpolation algorithm is shown in the following formula (2):

[0116]

[0117] In the above formula (2), N is the point closest to P. S refers to the starting point of each segment on the driving trajectory. is the value in formula (1) l is the projection length ratio.

[0118] Then, the computer device performs a cross product calculation on the vector to obtain a vector cross product result, and determines the offset direction of the vehicle according to the sign of the vector cross product result. If the sign of the vector cross product result is positive, the computer device determines that the offset direction of the vehicle is right. If the sign of the vector cross product result is negative, the computer device determines that the offset direction of the vehicle is left. The cross product calculation of the vector is shown in the following formula (3):

[0119]

[0120] In the above formula (3), For vector The modulus length. sign() is a sign function used to determine the positive or negative nature of a number. is the value in the above formula (1) is the value in the above formula (1)

[0121] The computer device performs Euclidean distance calculation on point P and point N to obtain the lateral offset. The Euclidean distance calculation process is shown in the following formula (4):

[0122]

[0123] In the above formula (4), is the horizontal coordinate of point P, is the ordinate of point P. px is the abscissa of point N. The vertical coordinate of the horizontal coordinate of N.

[0124] The computer device then calculates the distance d between point P and the stop line lane . The distance between point P and the stop line is shown in the following formula group (5):

[0125]

[0126] Among them, in the above formula group (5), flagfa l se means the vehicle has not passed the stop line and has not entered the intersection interaction area. flagtr ue It means that the vehicle has passed the stop line and is located in the intersection interaction area. The distance between the actual position data and the stop line also represents the interaction characteristics between the vehicle and the traffic light phase.

[0127] Step 404: Determine in each secondary motion data set a secondary motion data subset corresponding to the primary motion data subset.

[0128] The secondary motion data set includes various secondary motion data subsets.

[0129] In implementation, the computer device determines, for each secondary motion data set, the secondary motion data subset at the acquisition time in the secondary motion data set according to the acquisition time of the primary motion data subset, and obtains each secondary motion data subset corresponding to the primary motion data subset.

[0130] In an exemplary embodiment, the primary motion data subset is collected at 16:24:41 on October 16, 2024. For each secondary motion data set, the computer device determines a secondary motion data subset collected at 16:24:41 on October 16, 2024 in the secondary motion data set.

[0131] Step 406 , extracting features from the primary motion data subset, each secondary motion data subset and each boundary line based on an interactive feature algorithm to obtain interactive features of the vehicle.

[0132] The primary motion data subset includes the actual position data of the vehicle at the time of collection, and the secondary motion data subset includes the reference position data of the reference object or reference vehicle.

[0133] In implementation, the computer device constructs a circular radar perception area with the actual position data as the center. The computer device selects the target reference object from each reference object and the target reference vehicle from the reference vehicle based on the reference position data of each reference object, the reference position data of each reference vehicle and the radar perception area. If there is no target reference object and no target reference vehicle, the computer device determines the interaction characteristics of the vehicle based on each boundary line and the actual position data. If there is a target reference object or a target reference vehicle or both a target reference vehicle and a target reference object, the computer device determines the interaction characteristics of the vehicle based on the reference position data of the target reference object, the reference position data of the target reference vehicle and the actual position data.

[0134] Specifically, the computer device transforms the states of other traffic participants to simulate vehicle perception. Specifically, the computer device establishes a circular radar perception area with a preset radius based on the actual position of the vehicle as the center. Then, the computer device evenly divides the radar perception area in a counterclockwise direction in front of the left long axis of the vehicle to obtain each perception area. Each perception area is a fan-shaped block with an inner angle of 10 degrees. The computer device determines the interaction characteristics of the vehicle based on each perception area, the main motion data subset, each secondary motion data subset and each boundary line. The interaction characteristics include two interaction variables, which are the interaction type of the object closest to the vehicle hit by the virtual laser The second is the interaction distance between the object and the vehicle Specifically, the interaction types are divided into entities and non-entities. Entities are target reference objects or target reference vehicles. Non-entities are environments, i.e., the boundaries of intersections.

[0135] The computer device determines, for each reference object, whether the reference object is located in the perception area based on the reference position data and the perception area of ​​the reference object. If the reference object is located in the perception area, the computer device determines the target reference object from the reference object. The computer device determines, for each reference vehicle, whether the reference vehicle is located in the perception area based on the reference position data and the perception area of ​​the reference vehicle. If the reference vehicle is located in the perception area, the computer device determines the target reference vehicle from the reference vehicle. If there is a target reference object or a target reference vehicle, the computer device If there is no target reference object and no target reference vehicle, the computer device will is determined to be 0. In the case where there is a target reference object or a target reference vehicle, the computer device calculates the distance between each target reference object and the vehicle, and calculates the distance between each target reference vehicle and the vehicle. Then, the computer device determines the closest distance among the distances as If there is no target reference vehicle and no target reference object, the actual position data of the vehicle is used as the center of the circle, and the starting angle and ending angle of the radar scanning angle are determined according to the orientation of the vehicle. Then, the computer device calculates the angle of each boundary point in the boundary point set relative to the vehicle for each boundary line boundary point set. Before calculating the angle of the boundary point relative to the vehicle, the computer device transforms the coordinates of the boundary point to the vehicle coordinate system and calculates the angle. The angle calculation formula is shown in the following formula (6):

[0136]

[0137] In the above formula (6), (x c ,y c ) is the actual position data of the vehicle, that is, the actual position coordinates. i is the starting angle θs Or end angle θ e .x i y i are the coordinates of the boundary points.

[0138] Then, the computer device calculates the s and θ e The angle condition is constructed based on the value of , and the boundary points within the specified angle range are filtered according to the angle condition. For the boundary points that meet the angle condition, the computer device calculates the square of the Euclidean distance between the boundary point and the vehicle. The calculation formula of Euclidean distance is shown in the following formula (7):

[0139]

[0140] In the above formula (7), (x c ,y c ) is the actual position data of the vehicle, that is, the actual position coordinates. i y i is the coordinate of the boundary point that meets the angle condition. Then, the computer device determines the minimum distance value among the distances between each boundary point and the vehicle, and determines the minimum distance value as

[0141] In this embodiment, by extracting the offset features and interaction features of the vehicle, the relationship between the position of the vehicle at each moment and the environment in which the vehicle is located is clarified, which facilitates the subsequent construction of the vehicle's driving feature set based on the offset features and interaction features.

[0142] In an exemplary embodiment, the road structure data set includes various boundary lines and stop lines, and the main motion data set includes a main motion data subset at each acquisition time, such as Figure 5 As shown, the specific processing process of constructing the vehicle driving feature set according to the traffic intention feature, the interaction feature and the offset feature in step 308 includes steps 502 to 504. Among them:

[0143] Step 502 : constructing a driving feature subset of the vehicle at the collection time based on each interaction feature, the offset feature corresponding to the interaction feature, and the traffic intention feature.

[0144] In implementation, the computer device preprocesses each interaction feature, offset feature, and communication intention feature to obtain the preprocessed interaction feature, offset feature, and traffic intention feature. Then, for each interaction feature, the computer device determines the offset feature corresponding to the interaction feature according to the collection time of the interaction feature. Then, the interaction feature, the offset feature corresponding to the interaction feature, the traffic intention feature, and the attribute information of the vehicle are constructed as a driving feature subset of the vehicle at the collection time.

[0145] Specifically, the attribute information of the vehicle includes the width and length of the vehicle. The computer device encodes and normalizes each interaction feature, offset feature and traffic intention feature according to the normalization algorithm and the encoding algorithm. At the same time, the computer device normalizes the attribute information of the vehicle according to the normalization algorithm. Specifically, if the feature (interaction feature, offset feature and traffic intention feature) is string data, the computer device encodes the feature according to the unique hot encoding algorithm to obtain the processed feature. If the feature is numerical data, the computer device normalizes the feature based on the Min-Max (Min-MaxNormalization, linear function normalization) algorithm to obtain the processed feature. The computer device normalizes the attribute information using the Min-Max algorithm to obtain the processed attribute information. Then, for each interaction feature, the computer device determines the offset feature corresponding to the interaction feature according to the collection time of the interaction feature. Then, the interaction feature, the offset feature corresponding to the interaction feature, the traffic intention feature and the attribute information of the vehicle are constructed as a driving feature subset of the vehicle at the collection time. The driving feature subset ξ includes dir1, dir2, dir3, dir4, dir5, dir6, dir7, dir8, dir9, dir10, dir11, dir12, dir13, dir14, dir15, dir16, dir17, dir18, dir19, dir20, dir210, dir221, dir31, dir32, dir33, dir34, dir35, dir36, dir37, dir38, dir39, dir39, dir311, dir312, dir313, dir314, dir315, dir316, dir317, dir318, dir329, dir33, dir34, dir35, dir36, dir37, dir38, dir39, dir319, dir311, dir lat , d lon , d lane , vehicle width l vehicle length w, interactive features Among them, dir1 is the first traffic intention feature, dir2 is the second traffic intention feature, dir3 is the third traffic intention feature, d lat is the offset direction, d lon is the offset distance, d lane The distance between the vehicle and the stop line. For interaction type, is the interaction distance.

[0146] Step 504 : constructing a driving feature set of the vehicle according to each driving feature subset.

[0147] In implementation, the computer device combines the driving characteristic subsets into a driving characteristic set for the vehicle.

[0148] In this embodiment, the vehicle's posture feature set is constructed through the vehicle's attribute information, traffic intention features, interaction features and offset features, which clarifies the relationship between the vehicle's driving trajectory and the vehicle's traffic intention features and interaction feature attributes, making it easier to determine the vehicle's estimated offset trajectory sequence set in the future.

[0149] In an exemplary embodiment, Figure 6 As shown, the specific processing process of determining the estimated deviation trajectory sequence set of the vehicle based on the driving feature set and the agent model in step 104 includes steps 602 to 606. Among them:

[0150] Step 602: Divide each driving feature subset in the driving feature set into driving feature subset groups.

[0151] In implementation, the computer device sorts each driving feature subset in the driving feature set in ascending order of collection time to obtain a driving feature subset sequence. The computer device divides the driving feature subset sequence according to a preset value to obtain each driving feature subset group.

[0152] In an exemplary embodiment, one acquisition moment is one frame. The value is 30 frames. The computer device sorts the driving feature subsets in the driving feature set in ascending order of acquisition moments to obtain a driving feature subset sequence. Then, the computer device divides the 30 driving feature subsets in the driving feature subset sequence into one driving feature subset group to obtain each driving feature subset group.

[0153] Step 604 , performing prediction processing on each driving feature subset group based on the proxy model to obtain an initial estimated deviation trajectory sequence.

[0154] Among them, the proxy model is obtained by training the initial proxy model based on the sample feature set.

[0155] In implementation, the computer device inputs the driving feature subset group into the proxy model for each driving feature subset group, and performs prediction processing on the driving feature subset group through the proxy model to obtain an initial estimated offset trajectory sequence.

[0156] In an exemplary embodiment, the driving feature subset group includes 30 frames of driving feature subsets. The computer device inputs the 30-frame driving feature subsets into the proxy model, and performs prediction processing on the 30-frame driving feature subsets through the proxy model to obtain an initial estimated offset estimation sequence for the next 10 frames. Specifically, the proxy model includes a generator and a discriminator. The generator is an LSTM network (Long Short-Term Memory) with an attention mechanism, and the received feature input dimensions are a three-dimensional feature matrix of n, t, and m. The features received by the generator are the feature input dimensions of the proxy model, including: n is the number of traffic participants in the batch (driving feature subset group), t is 1; t is the time frame of the training input, and m is the feature dimension. The discriminator is a feedforward neural network, and the received feature input dimensions are n, t+t f , n a , where t f is the feature of the traffic participants in the first 30 frames, n a The proxy model is fed with the d of the next 10 frames. lat , d lon , that is, the estimated offset direction and the estimated offset distance.

[0157] Step 606 , coordinate conversion processing is performed on the initial estimated offset trajectory sequence to obtain an estimated offset trajectory sequence, and an estimated offset trajectory sequence set is constructed according to each estimated offset trajectory sequence.

[0158] In the implementation, since the initial estimated offset estimation sequence output by the proxy model is the coordinate of the Frenet coordinate system (a coordinate system used to describe the motion and position of an object on a curve), it is necessary to convert the initial estimated offset sequence into coordinate points in the real coordinate system through coordinate transformation. The computer device performs coordinate transformation processing on the initial estimated offset trajectory sequence according to the transformation algorithm to obtain an estimated offset trajectory sequence. Then, the computer device constructs an estimated offset trajectory sequence set according to each estimated offset trajectory sequence. Among them, at the trajectory point (S ix , S iy ), the coordinate system conversion algorithm is shown in the following formula group (8):

[0159]

[0160] In the above formula (8), (S ix ,S iy ) is the trajectory point at the current moment (current frame), (S (i+1)x , S (i+1)y ) is the trajectory point at the next moment. x and p y is the estimated trajectory point after coordinate transformation, N x and N y is the reference coordinate. lat Estimated offset distance.

[0161] In this embodiment, the estimated offset trajectory sequence set of the vehicle is determined by the proxy model and the driving feature set, and the vehicle's driving trajectory can be dynamically adjusted based on the estimated offset trajectory sequence set. Furthermore, a dynamic simulation scene can be obtained based on each offset trajectory sequence set, the traffic participant motion data set, and the road structure data set, which improves the authenticity of the dynamic simulation scene and expands the applicability of the dynamic simulation scene. Testing the autonomous driving vehicle based on the dynamic simulation scene can achieve two-way avoidance between the autonomous driving vehicle and the manually driven vehicle, and improve the accuracy of the autonomous driving test.

[0162] In an exemplary embodiment, before applying the proxy model, it is necessary to first train the initial proxy model based on the sample data set to obtain the proxy model. Figure 7 As shown, before step 102 is executed, the specific execution process of the scene construction method also includes:

[0163] Step 702, obtaining a sample data set and a road structure data set of the intersection, and extracting a sample driving feature of each sample vehicle passing through the intersection based on the road structure data set and the sample data set to obtain a sample driving feature set of the vehicle.

[0164] During implementation, the computer device obtains a sample data set of the intersection and a road structure data set of the intersection. The sample data set is a SinD data set (Signalized Intersection Dataset, a signalized intersection data set, which is a drone aerial photography data set focusing on urban signalized intersections), and the sample data set contains a motion data set of each sample traffic participant. The sample traffic participants include sample vehicles and sample reference objects. The computer device extracts the sample driving features of each sample vehicle passing through the intersection based on the road structure data and the sample data set to obtain a sample driving feature set. The extraction process of the sample driving features is the same as the extraction process of the driving features. Please refer to the above steps 302 to 308 for the specific process, and the embodiments of the present application will not be repeated here.

[0165] In an optional embodiment, the computer device performs Monte Carlo sampling on the sample traffic intention features in each sample driving feature set to obtain each lane at the intersection and a reference driving trajectory set.

[0166] Step 704 : determining each sample driving feature subset group based on each sample driving feature set.

[0167] Each sample driving feature set includes each sample driving feature subset.

[0168] In implementation, the computer device divides each sample driving feature set into initial sample driving feature subset groups according to a preset value for each sample driving feature set. Then, the computer device combines the initial sample driving feature subset groups of each vehicle that appears at a similar time into a sample driving feature subset group according to a preset number of vehicles.

[0169] In an exemplary embodiment, a computer device contains a sample driving feature set A of vehicle A, a sample driving feature set B of vehicle B, and a sample driving feature set B of vehicle B. Sample driving feature set A includes sample driving feature subsets A1 to A90. Sample driving feature set B includes sample driving feature subsets B1 to B90. Sample driving feature set C includes sample driving feature subsets C1 to C90. The computer device divides sample driving feature set A into initial sample driving feature subset groups a1, a2, and a3 according to a preset value 30. Initial sample driving feature subset group a1 includes sample driving feature subsets A1 to A30. Initial sample driving feature subset group a1 includes sample driving feature subsets A31 to A60. Initial sample driving feature subset group a1 includes sample driving feature subsets A61 to A90. At the same time, the computer device divides the sample driving feature set B into initial sample driving feature subset groups b1, b2 and b3, and divides the sample driving feature set C into initial sample driving feature subset groups c1, c2 and c3. The computer device combines the initial sample driving feature subset groups a1, b1 and c1 into a sample driving feature subset group X according to the preset number of vehicles 3. At the same time, the computer device combines the initial sample driving feature subset groups a2, b2 and c2 into a sample driving feature subset group Y, and combines the initial sample driving feature subset groups a3, b3 and c3 into a sample driving feature subset group Z.

[0170] Optionally, the number of vehicles may be, but is not limited to, set to 3, and the value may be, but is not limited to, set to 30. The number of vehicles and the value are determined according to training requirements, and the embodiment of the present application does not limit the number of vehicles and the value.

[0171] Step 706 : Train an initial proxy model based on each sample driving feature subset group to obtain a proxy model.

[0172] In an embodiment, the computer device inputs each sample driving feature subset group into the initial proxy model, and performs prediction processing on the sample driving feature subset group through the proxy model to obtain a sample output result. The computer device calculates each sample output result and the real result corresponding to each sample output result according to a preset loss function algorithm to obtain a loss value. The computer device determines whether the loss value meets the preset training stop condition. If the loss value meets the preset stop condition, the computer device determines the initial proxy model as the proxy model. If the loss value does not meet the stop condition, the computer device updates the parameters of the initial proxy model based on the loss value, and continues to execute the step of the computer device inputting each sample driving feature subset group into the initial proxy model until the loss value reaches the training stop condition.

[0173] Specifically, the initial proxy model includes a generator and a discriminator. The computer device constructs a three-dimensional feature matrix n, t, m based on the sample driving feature subset group, where n is the number of traffic participants in the sample driving feature subset group, t is the time frame of the training input, and m is the feature dimension; the discriminator is a feedforward neural network, and the received feature input dimension is n, t+t f ,n a , where t f is the feature of the traffic participants in the first 30 frames, n a It is the dimension of two characteristic quantities: lateral offset and longitudinal offset of traffic participants.

[0174] The computer device inputs the three-dimensional feature matrix to the generator and the discriminator. The generator will output the sample estimated offset trajectory sequence according to the input features. The discriminator will give an action in the three-dimensional feature matrix, that is, the probability that the sample estimated offset trajectory sequence is "true". This probability will be compared with the true value and the discriminator will be scored to obtain the score. The loss of the true action and the false action is further calculated, which is explained as: the discriminator can correctly identify "true" as "true", and the discriminator identifies "false" as "true". The optimizer is Adam. The generator network is an LSTM network, and the solution of the generator is shown in the following formula (9):

[0175] h t , c t =LSTM(x t ,h t-1 , c t-1 ) (9)

[0176] In the above formula (9), h t and c t denote the hidden state and cell state at time step t, respectively, and x t is the input feature at time step t. The attention weight is calculated through a new fully connected layer, which maps the output of LSTM to a weight coefficient matrix, and then normalizes the weight coefficient matrix through the softmax function to ensure that the sum of all weights is 1. The attention weight is calculated as follows:

[0177]

[0178] In the above formula (10), w and b are the weight and bias of the fully connected layer, respectively, and α t is the attention weight at time step t, seq l is the feature sequence. The loss includes cross entropy loss, discontinuity loss and dissimilarity loss. The cross entropy of the discriminator is shown in the following formula (11), and the cross entropy of the generator is shown in the following formula (12):

[0179]

[0180] In the above formulas (11) and (12), N is the number of samples in the batch, x i is the input sequence, y i is a label, D(x i ) is the discriminator network for the input sequence x i The predicted probability of i ) is the generator according to the state input z i The generated samples, D(G(z i )) is the discriminator's predicted probability value of the authenticity of the generated sample. The overall loss function is:

[0181]

[0182] In the above formula (13), is the value of formula (11) is the value of formula (12) Coefficient λ C ,λ S is a weight factor that balances the importance of custom loss. and As shown in the following formula (14) and formula (15):

[0183]

[0184] In the above formulas (14) and (15), Δtraj g is the difference of the generated offset sequence, θ is the threshold set to evaluate whether the current difference value is acceptable, λ is the scaling factor for adjusting the penalty sensitivity; γ and μ are the exponents and multipliers for adjusting the penalty intensity; traj g and traj e are the generated offset sequence and the true offset sequence respectively, v is the weight factor, and i is the count value of the batch calculation.

[0185] Based on the above loss function, an optimization function of the initial proxy model is obtained. The computer device optimizes the model parameters of the initial dialing model based on the optimization function. The optimization function is shown in the following formula (16):

[0186]

[0187] In formula (16), Indicates expected value; traj e represents the trajectory from the expert data, x t represents the input features of the generator G at time step t; p daiais the actual data distribution, p G (x) is the output distribution of generator G.

[0188] In this embodiment, each sample driving feature subset group is determined through a sample data set, and then an initial proxy model is trained based on each sample driving feature subset group to obtain a proxy model, thereby improving the accuracy of the proxy model and further improving the accuracy of the estimated offset trajectory sequence set.

[0189] In an exemplary embodiment, Figure 8 The present invention is an architecture diagram of a scene construction device in an embodiment. An architecture diagram of a scene construction device corresponding to a scene construction method is provided. The scene construction device includes a data and feature processing module, an agent model construction module and a scene generator. The data and feature processing module is used to extract the sample driving features of each sample vehicle passing through an intersection based on a road structure data set and a sample data set to obtain a sample driving feature set of the vehicle. The agent model training module is used to train an initial agent model based on each sample driving feature set to obtain an agent model. The scene generator model is used to obtain a road structure data set and a traffic participant motion data set of an intersection; the traffic participant motion data set includes a motion data set of each vehicle and each reference object passing through an intersection; based on the road structure data set and the traffic participant motion data set, the driving features of each vehicle passing through an intersection are extracted to obtain a driving feature set of the vehicle, and based on the driving feature set and the agent model, an estimated offset trajectory sequence set of the vehicle is determined; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle; based on each estimated offset trajectory sequence set, the traffic participant motion data set and the road structure data set, a dynamic simulation scene of the intersection is obtained. By modeling and reconstructing the behavioral decisions of traffic participants, estimating the deviation trajectory sequence, and adding the estimated deviation trajectory sequence as a dynamic element to the simulation scene, the authenticity and interactive complexity of the test scene are ensured. Fig. 9 It is a schematic diagram of the trajectory of a motor vehicle in a static simulation scenario and the trajectory of a motor vehicle in a dynamic simulation scenario in one embodiment. Fig. 9 (a) Schematic diagram of motor vehicle trajectories in the static simulation scenario. Fig. 9 (b) Schematic diagram of motor vehicle trajectories in the dynamic simulation scenario.

[0190] In an exemplary embodiment, Fig.10 FIG. 1 is a schematic diagram of the architecture of a training module in an embodiment. Fig.10As shown in the figure, the upper layer is the feature processing stage. The idea of ​​constructing the traffic participant agent model comes from the way human drivers pass through intersections, which involves the following assumptions: drivers will not change the direction of travel at will when passing through intersections; drivers will try their best to drive along the expected trajectory of the direction under different travel directions unless other cars interfere with their own car; when drivers perceive other cars in the same direction, they only pay attention to the other cars closest to them and make interactive reactions. Based on the above assumptions, in the feature processing stage, this application designs the following Fig.10 The driving feature set shown includes attributes, temporal features, and perceptual features including perceptual type and distance. After encoding, normalizing, batching, and sequence extraction of the extracted features, all rounds, scene file segments, batches, and time frames are traversed in turn. The lower half is the implementation process of training, including generator network design, discriminator network design, loss design, and training. The initial proxy model includes a generator (LSTM) and a discriminator (feedforward network). The computer device inputs the three-dimensional feature matrix to the generator and the discriminator. The generator will output a sample estimated offset trajectory sequence based on the input features. The discriminator will give an action in the three-dimensional feature matrix, that is, the sample estimated offset trajectory sequence is "true". The probability is compared with the true value and the discriminator is scored to obtain a score, and the loss of the true action and the false action is further calculated, and the generator is optimized based on the loss and the optimizer.

[0191] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0192] Based on the same inventive concept, the embodiment of the present application also provides a scene construction device for implementing the scene construction method involved above. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above method, so the specific limitations in one or more scene construction device embodiments provided below can refer to the limitations of the scene construction method above, and will not be repeated here.

[0193] In an exemplary embodiment, Fig.11As shown, a scene construction device 1100 is provided, comprising: an acquisition module 1101, an extraction module 1102 and a combination module 1103, wherein:

[0194] The acquisition module 1101 is used to acquire a road structure dataset and a traffic participant motion dataset of an intersection; the traffic participant motion dataset includes a motion dataset of each vehicle and each reference object passing through the intersection.

[0195] The extraction module 1102 is used to extract the driving characteristics of each vehicle passing through the intersection based on the road structure data set and the traffic participant movement data set to obtain the vehicle's driving characteristic set, and determine the vehicle's estimated offset trajectory sequence set based on the driving characteristic set and the agent model; the estimated offset trajectory sequence set is used to dynamically adjust the vehicle's driving trajectory.

[0196] The combination module 1103 is used to obtain a dynamic simulation scene of the intersection based on each estimated offset trajectory sequence set, the traffic participant motion data set and the road structure data set.

[0197] In an exemplary embodiment, the acquisition module 1101 includes a first acquisition submodule and a second acquisition submodule. The first acquisition submodule includes:

[0198] The third acquisition submodule is used to acquire an initial road structure dataset of the intersection; the initial road structure dataset includes lane lines and boundary lines of the intersection.

[0199] The first determination submodule is used to determine each lane of the intersection based on each lane line and each boundary line, and mark each lane.

[0200] The first establishing submodule is used to traverse each lane, establish a reference driving trajectory between the lane and other lanes with the lane as a starting point, and obtain a reference driving trajectory set.

[0201] The first updating submodule is used to update the initial road structure data set according to the reference driving trajectory set and each lane to obtain a road structure data set.

[0202] In an exemplary embodiment, the road structure data set includes a reference driving trajectory set, the driving features include a traffic intention feature, an interaction feature, and an offset feature, and the extraction module 1102 includes a first extraction submodule and a second determination submodule. The first extraction submodule includes:

[0203] The second determining submodule is used to determine the motion data set as a main motion data set for each vehicle's motion data set.

[0204] The first screening submodule is used to screen reference vehicles and reference objects that appear simultaneously with the vehicle in the traffic participant motion data set, and determine the motion data sets of the reference vehicles and the motion data sets of the reference objects as the secondary motion data sets.

[0205] The second extraction submodule is used to extract the vehicle's passing intention features according to the passing intention rule and the main motion data set, and determine a reference driving trajectory corresponding to the passing intention features in the reference driving trajectory set.

[0206] The third extraction submodule is used to extract the interaction features and offset features of the vehicle according to the road structure dataset, the reference driving trajectory, the main motion dataset and each secondary motion dataset, and to construct a driving feature set of the vehicle according to the traffic intention features, the interaction features and the offset features.

[0207] In an exemplary embodiment, the road structure data set includes each boundary line and stop line, the main motion data set includes a main motion data subset at each acquisition time, and the third extraction submodule includes a fourth extraction submodule and a first construction submodule. Among them, the fourth extraction submodule includes:

[0208] The first operation submodule is used for performing an offset operation on the stop line, the reference driving trajectory and the actual position data in the main motion data subset according to the lateral and longitudinal offset algorithms for each main motion data subset to obtain an offset feature.

[0209] The third determination submodule is used to determine, in each secondary motion data set, a secondary motion data subset corresponding to the primary motion data subset.

[0210] The fifth extraction submodule is used to extract features from the primary motion data subset, each secondary motion data subset and each boundary line based on an interactive feature algorithm to obtain interactive features of the vehicle.

[0211] In an exemplary embodiment, the third extraction submodule includes a fourth extraction submodule and a first construction submodule. The first construction submodule includes:

[0212] The second construction submodule is used to construct a driving feature subset of the vehicle at the collection time based on each interaction feature, an offset feature corresponding to the interaction feature, and a traffic intention feature.

[0213] The third construction submodule is used to construct a driving feature set of the vehicle according to each driving feature subset.

[0214] In an exemplary embodiment, the road structure data set includes a reference driving trajectory set, the driving features include a traffic intention feature, an interaction feature, and an offset feature, and the extraction module 1102 includes a first extraction submodule and a second determination submodule. The second determination submodule includes:

[0215] The division submodule is used to divide each driving feature subset in the driving feature set into each driving feature subset group.

[0216] The first processing submodule is used to perform prediction processing on each driving feature subset group based on the proxy model to obtain an initial estimated deviation trajectory sequence.

[0217] The second processing submodule is used to perform coordinate conversion processing on the initial estimated offset trajectory sequence to obtain an estimated offset trajectory sequence, and to construct an estimated offset trajectory sequence set according to each estimated offset trajectory sequence.

[0218] Each module in the above scene construction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0219] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.12 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (Near Field Communication, NFC) or other technologies. When the computer program is executed by the processor, a scene construction method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0220] Those skilled in the art will understand that Fig.12The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0221] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0222] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0223] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0224] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0225] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0226] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A scene construction method, characterized in that: The method comprises: Acquire a road structure dataset and a traffic participant motion dataset of an intersection; the traffic participant motion dataset includes a motion dataset of each vehicle and each reference object passing through the intersection; Extracting driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant motion dataset to obtain a driving characteristic set of the vehicle, and determining an estimated offset trajectory sequence set of the vehicle based on the driving characteristic set and an agent model; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle; Based on each of the estimated offset trajectory sequence sets, the traffic participant motion data set and the road structure data set, a dynamic simulation scene of the intersection is obtained.

2. The method according to claim 1, characterized in that: The step of obtaining a road structure dataset of an intersection includes: Acquire an initial road structure data set of an intersection; the initial road structure data set includes lane lines and boundary lines of the intersection; Determine each lane of the intersection based on each lane line and each boundary line, and mark each lane; Traversing each lane, establishing a reference driving trajectory between the lane and other lanes with the lane as a starting point, and obtaining a reference driving trajectory set; The initial road structure data set is updated according to the reference driving trajectory set and each of the lanes to obtain a road structure data set.

3. The method according to claim 1, characterized in that The road structure dataset includes a reference driving trajectory set, the driving features include a traffic intention feature, an interaction feature, and an offset feature, and the driving features of each vehicle passing through the intersection are extracted based on the road structure dataset and the traffic participant movement dataset to obtain the driving feature set of the vehicle, including: For each motion data set of the vehicle, determining the motion data set as a main motion data set; Screening each reference vehicle and each reference object that appear simultaneously with the vehicle in the traffic participant motion data set, and determining the motion data set of each reference vehicle and the motion data set of each reference object as each secondary motion data set; Extracting the traffic intention feature of the vehicle according to the traffic intention rule and the main motion data set, and determining the reference driving trajectory corresponding to the traffic intention feature in the reference driving trajectory set; According to the road structure data set, the reference driving trajectory, the main motion data set and each of the secondary motion data sets, the interaction features and the offset features of the vehicle are extracted, and the driving feature set of the vehicle is constructed according to the traffic intention features, the interaction features and the offset features.

4. The method according to claim 3, characterized in that The road structure data set includes each boundary line and stop line, the main motion data set includes a main motion data subset at each acquisition time, and the interaction feature and offset feature of the vehicle are extracted according to the road structure data set, the reference driving trajectory, the main motion data set and each of the secondary motion data sets, including: For each of the main motion data subsets, performing an offset operation on the stop line, the reference driving trajectory and the actual position data in the main motion data subset according to a lateral and longitudinal offset algorithm to obtain an offset feature; Determining, in each of the secondary motion data sets, a secondary motion data subset corresponding to the primary motion data subset; Based on the interactive feature algorithm, feature extraction is performed on the main motion data subset, each of the secondary motion data subsets and each boundary line to obtain the interactive feature of the vehicle.

5. The method according to claim 4, characterized in that The step of constructing the driving feature set of the vehicle according to the traffic intention feature, the interaction feature and the offset feature comprises: Constructing a driving feature subset of the vehicle at the collection time based on each of the interaction features, the offset feature corresponding to the interaction feature, and the traffic intention feature; A driving feature set of the vehicle is constructed according to each of the driving feature subsets.

6. The method according to claim 1, characterized in that The step of determining the estimated deviation trajectory sequence set of the vehicle based on the driving feature set and the agent model includes: dividing each driving feature subset in the driving feature set into driving feature subset groups; Performing prediction processing on each of the driving feature subset groups based on the proxy model to obtain an initial estimated deviation trajectory sequence; Coordinate transformation processing is performed on the initial estimated offset trajectory sequence to obtain an estimated offset trajectory sequence, and an estimated offset trajectory sequence set is constructed according to each of the estimated offset trajectory sequences.

7. A scene construction device, characterized in that: The device comprises: An acquisition module, used to acquire a road structure data set and a traffic participant motion data set of an intersection; the traffic participant motion data set includes a motion data set of each vehicle and each reference object passing through the intersection; an extraction module, configured to extract driving characteristics of each vehicle passing through the intersection based on the road structure dataset and the traffic participant motion dataset, obtain a driving characteristic set of the vehicle, and determine an estimated offset trajectory sequence set of the vehicle based on the driving characteristic set and an agent model; the estimated offset trajectory sequence set is used to dynamically adjust the driving trajectory of the vehicle; A combination module is used to obtain a dynamic simulation scene of the intersection based on each of the estimated offset trajectory sequence sets, the traffic participant motion data set and the road structure data set.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images