Trajectory prediction method, vehicle, computer equipment and program product

Abstract

The invention provides a trajectory prediction method, a vehicle, computer equipment and a program product applied to the technical field of automatic driving, and the method comprises the steps: when it is detected that a vehicle and a target obstacle vehicle are located at an intersection, obtaining a trajectory prediction result; determining a connection line direction of each target virtual lane connection line and a target distance between each target virtual lane connection line and the target obstacle vehicle; screening out a matched lane connection line matched with the target obstacle vehicle from the target virtual lane connection lines according to the connection line direction and the target distance corresponding to the target virtual lane connection lines; and determining a predicted trajectory of the target obstacle vehicle according to the vehicle information of the target obstacle vehicle, the road environment information and the matched lane connection line. The problem of inaccurate trajectory prediction of the obstacle vehicles around the vehicle at the intersection can be solved, trajectory prediction is performed by using multi-dimensional information such as the connecting line direction and the target distance, and the prediction accuracy of the driving trajectory of the obstacle vehicles in the intersection scene can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of autonomous driving technology, and in particular to a trajectory prediction method, a vehicle, a computer device, and a program product. Background Art

[0002] With the rapid development of artificial intelligence technology and new energy vehicles, the automatic driving function has become a standard feature of many new energy vehicles, and the automatic driving function of vehicles is becoming more and more powerful. As a very important functional module in automatic driving technology, the prediction module is mainly used to predict the driving trajectory of some obstacles around the vehicle in the future and input it into the planning module in automatic driving to assist the planning module in formulating a reasonable decision plan. After obtaining the decision plan, the control module in automatic driving can control the vehicle to achieve automatic driving based on the formulated decision plan. Therefore, the rationality and accuracy of the obstacle driving trajectory predicted by the prediction module will directly affect the safety and rationality of automatic driving.

[0003] However, predicting the trajectory of obstacles around the vehicle has always been a difficult problem, especially when the vehicle and the obstacle are at an intersection. Therefore, how to improve the rationality and accuracy of the obstacle trajectory predicted by the prediction module has become a technical difficulty. Summary of the invention

[0004] The embodiments of the present disclosure at least provide a trajectory prediction method, a vehicle, a computer device, and a program product.

[0005] In a first aspect, an embodiment of the present disclosure provides a trajectory prediction method, comprising: When it is detected that the ego vehicle and the target obstacle vehicle are located at the intersection, determine the connection direction of each target virtual lane connection line between the intersection entrance position and the intersection exit position, and the target distance between each target virtual lane connection line and the target obstacle vehicle; the target virtual lane connection line is constructed according to the topological relationship between the lanes at the intersection; According to the line direction and the target distance corresponding to the target virtual lane line, a matching lane line that matches the target obstacle vehicle is selected from each of the target virtual lane lines; The predicted trajectory of the target obstacle vehicle is determined according to the vehicle information of the target obstacle vehicle, the road environment information and the matching lane lines.

[0006] This implementation method, based on the topological relationship between lanes at the intersection, can fully consider various drivable road options and construct various target virtual lane lines for vehicles to travel. By combining the distance between the target virtual lane line and the target obstacle vehicle and the line direction corresponding to the target virtual lane line, it is possible to match the target obstacle vehicle with the target virtual lane line in two dimensions, namely, distance and end point direction, so as to more accurately judge the matching degree between the target obstacle vehicle and each target virtual lane line. Using the matching degree, the matching lane line that the target obstacle vehicle is most likely to travel can be accurately screened out, and then the driving trajectory of the target obstacle vehicle can be predicted using the matching lane line, which can improve the accuracy of the predicted driving trajectory. Especially when the self-vehicle and the target obstacle vehicle are at the intersection, using two-dimensional information for trajectory prediction can effectively improve the prediction accuracy and rationality of the driving trajectory of the target obstacle vehicle in the intersection scene, thereby improving the safety of autonomous driving.

[0007] In a second aspect, an embodiment of the present disclosure further provides a vehicle, including a controller, wherein the controller includes: a memory for storing machine-readable instructions; and A processor is used to call and execute the machine-readable instructions stored in the memory to implement the above-mentioned first aspect, or the steps in any optional implementation of the above-mentioned first aspect.

[0008] In a third aspect, an optional implementation of the present disclosure further provides a computer device, a processor, and a memory, wherein the memory stores machine-readable instructions executable by the processor, and the processor is used to execute the machine-readable instructions stored in the memory, and when the machine-readable instructions are executed by the processor, the steps of the above-mentioned first aspect, or any possible implementation of the first aspect are performed.

[0009] In a fourth aspect, an optional implementation of the present disclosure further provides a computer program product, which, when executed, executes the steps of the above-mentioned first aspect, or any possible implementation of the first aspect.

[0010] For a description of the effects of the above-mentioned vehicle, computer device, and computer program product, please refer to the description of the above-mentioned trajectory prediction method, which will not be repeated here.

[0011] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used together with the specification to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can also be obtained based on these drawings without creative work.

[0013] Figure 1 A flow chart of a trajectory prediction method provided by an embodiment of the present disclosure is shown; Figure 2 A schematic diagram showing a road topology connection at an intersection provided by an embodiment of the present disclosure; Figure 3 A schematic diagram of a right turn scenario provided by an embodiment of the present disclosure is shown; Figure 4 A schematic diagram of a straight-ahead right turn scenario provided by an embodiment of the present disclosure is shown; Figure 5 An architecture diagram of trajectory prediction provided by an embodiment of the present disclosure is shown; Figure 6 A schematic diagram of the structure of a computer device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0014] The terms "first", "second", etc. in the description and claims of the embodiments of the present disclosure and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that shown or described herein.

[0015] The "multiple or several" mentioned in this article refers to two or more. "And / or" describes the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0016] Research has found that in the field of autonomous driving, the Planning and Control (PNC) module plays the role of decision-making planning and control, and is the core brain of the entire autonomous driving system. The input of the PNC module is the predicted trajectory of obstacles around the vehicle predicted by the prediction module. The accuracy of the predicted trajectory input by the prediction module affects the accuracy and rationality of the decision-making of the PNC module. However, the prediction module has always been a difficult point for predicting the trajectory of obstacles, especially the trajectory prediction when obstacles pass through intersections. Due to the large number of obstacles at intersections and the complex lane topology, there will be frequent interactions between different obstacles, the vehicle and each obstacle, which greatly increases the difficulty of predicting the trajectory of obstacles. However, at the intersection, if a reasonable obstacle prediction trajectory cannot be given, the PNC module will frequently misjudge, affecting driving safety and the user's riding experience.

[0017] Based on the above research, the present disclosure provides a trajectory prediction method, a vehicle, a computer device and a program product. Based on the topological relationship between the lanes at the intersection, various drivable road options can be fully considered to construct various target virtual lane lines for vehicle travel. By combining the distance between the target virtual lane line and the target obstacle vehicle and the line direction corresponding to the target virtual lane line, it is possible to achieve the matching of the target obstacle vehicle and the target virtual lane line in terms of distance and terminal orientation, so as to more accurately judge the matching degree between the target obstacle vehicle and each target virtual lane line. Using the matching degree, the matching lane line that the target obstacle vehicle is most likely to travel can be accurately screened out, and then the matching lane line can be used to predict the driving trajectory of the target obstacle vehicle, which can improve the accuracy of the predicted driving trajectory. Especially when the self-vehicle and the target obstacle vehicle are at the intersection, the trajectory prediction using two-dimensional information can effectively improve the prediction accuracy and rationality of the driving trajectory of the target obstacle vehicle in the intersection scene, thereby improving the safety of autonomous driving.

[0018] The defects existing in the above solutions are the results obtained by the inventor after practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed by the present disclosure for the above problems below should be the contributions made by the inventor to the present disclosure during the disclosure process.

[0019] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0020] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0021] To facilitate understanding of the present embodiment, a trajectory prediction method disclosed in the embodiment of the present disclosure is first introduced in detail. The execution subject of the trajectory prediction method provided in the embodiment of the present disclosure is generally a terminal device or other processing device with certain computing capabilities, wherein the terminal device may be a user equipment (UE), a mobile device, a user terminal, a terminal, a personal digital assistant (PDA), a handheld device, a computer device, a vehicle server, a vehicle terminal, etc.; in some possible implementations, the trajectory prediction method may be implemented by a processor calling a computer-readable instruction stored in a memory.

[0022] The trajectory prediction method provided by the embodiment of the present disclosure is described below by taking a computer device as an example of an execution subject.

[0023] like Figure 1 As shown, it is a flowchart of a trajectory prediction method provided by an embodiment of the present disclosure, which may include the following steps: S101: When it is detected that the ego vehicle and the target obstacle vehicle are located at an intersection, determine the connection direction of each target virtual lane connection line between the intersection entrance position and the intersection exit position, and the target distance between each target virtual lane connection line and the target obstacle vehicle; the target virtual lane connection line is constructed according to the topological relationship between the lanes at the intersection.

[0024] First, the computer device used to execute the trajectory prediction algorithm provided by the embodiment of the present disclosure may be a vehicle computer server, a prediction module deployed with the trajectory prediction algorithm, etc. The following description will be made by taking the computer device as the prediction module as an example.

[0025] The ego vehicle can be any vehicle equipped with an autonomous driving function. The target obstacle vehicle can be other road vehicles detected by the ego vehicle that may have an impact on its own driving. In specific applications, the ego vehicle can use various sensors on the vehicle body to detect various road obstacles on the road and the obstacle information of each road obstacle. Among them, road obstacles mainly refer to various dynamic obstacles, such as obstacle vehicles and pedestrians, among which obstacle vehicles can include obstacle motor vehicles and obstacle non-motor vehicles. It is understandable that obstacle vehicles that are stationary on the road (for example, accident vehicles, abandoned vehicles, etc.) will also be identified as road obstacles, but the obstacles will be ignored during calculation.

[0026] The obstacle information may be information related to the road obstacle, for example, it may include the attribute information and kinematic posture information of the road obstacle, wherein the attribute information is used to indicate some static attributes of the road obstacle, for example, it may include the type, color, shape, length, width, height information, wheelbase, horsepower, etc. of the road obstacle. The kinematic posture information is used to indicate the kinematic information of the road obstacle, for example, the traveling direction, position, speed, acceleration, etc. of the road obstacle. For the obstacle vehicle among the road obstacles, the obstacle information of the obstacle vehicle is hereinafter referred to as the vehicle information of the obstacle vehicle.

[0027] In addition, when the vehicle is driving, various sensors installed on the vehicle body can be used to collect road environment information, among which the road environment information can specifically be static road information collected by various sensors, such as traffic light data at intersections, dynamic signs of tidal lanes, lane lines, various traffic signs, road poles and other information.

[0028] Since the vehicle is constantly moving, the identified road obstacles and road environment information can change in real time, so in specific implementation, the vehicle can collect road environment information and obstacle information of road obstacles in real time.

[0029] The target virtual lane line is a virtual lane line that matches the target obstacle vehicle. The virtual lane line can be pre-constructed based on the virtual lane lines corresponding to each real lane at the intersection in the high-precision map and the topological relationship between the lanes. The virtual lane line can be constructed based on the lane center lines of each real lane in the high-precision map, and is used to represent the route that the vehicle passes from entering the intersection to leaving the intersection. A real lane can correspond to a virtual lane line. Multiple virtual lane lines can be connected according to the topological relationship between lanes to form a virtual lane line. The set of virtual lane lines can be defined as a road topological line. The target virtual lane line can be a line that may be driven by the target obstacle vehicle, which is selected from the virtual lane lines at the intersection. The topological relationship between lanes is used to indicate whether there is a successor relationship between lanes at different positions. If there is a successor relationship, it means that the vehicle can drive from the predecessor lane to the successor lane; if there is no successor relationship, it means that the vehicle cannot drive from one lane to another. Especially at intersections, since vehicles have a variety of executable behaviors at intersections, such as U-turns, left / right turns, going straight, going straight and turning left, going straight and turning right, etc., the lane topology relationship between lanes at intersections is more complex, and the road topology connections constructed at intersections based on the topological relationship between lanes and lane center lines will also be more complex.

[0030] In the specific implementation, a topological relationship between lanes is formed by connecting the two ends of the two drivable lanes at the intersection. Under normal circumstances, the driving trajectory of a vehicle passing through an intersection generally follows a certain line of the intersection topology, that is, it follows a lane corresponding to a certain topological relationship. Therefore, in the specific implementation, the lanes on the high-precision map and the driving rules at the intersection can be used to predetermine the topological relationship between the lanes at the intersection. According to the lane centerline of each lane, a virtual lane line corresponding to each lane is constructed, and then the virtual lane lines corresponding to different lanes are connected according to the topological relationship between the lanes, that is, each virtual lane line that depends on the topological relationship of the intersection is obtained. It can be understood that at an intersection, there may be annotated auxiliary lines, such as left-turn area lines, right-turn area lines, tidal lane lines, etc. Therefore, when constructing virtual lane lines according to virtual lane lines and lane topological relationships, the auxiliary lines at the intersection can be combined to improve the rationality of the constructed virtual lane lines. For example, when connecting virtual lane lines in different directions of an intersection, an auxiliary line can be used as a connecting line.

[0031] like Figure 2 FIG. 1 is a schematic diagram of a road topology connection at an intersection provided by an embodiment of the present disclosure. Figure 2 In , each solid line is a constructed virtual lane line, and the position of the virtual lane line can be the position of the center line of the real lane. A virtual lane line can be obtained by connecting the virtual lane lines corresponding to different lanes at the intersection. A virtual lane line corresponding to a lane can appear in different virtual lane lines. For example, for a left-turn lane at an intersection, the virtual lane line corresponding to the lane when the vehicle enters the intersection can be connected with the virtual lane lines corresponding to each lane that the vehicle can drive when the vehicle exits the intersection, so as to obtain a plurality of virtual lane lines for left turns. For another example, for a straight lane at an intersection, the virtual lane line corresponding to the lane when the vehicle enters the intersection can be connected with the virtual lane line corresponding to the straight lane that the vehicle can drive when the vehicle exits the intersection, so as to obtain a straight virtual lane line. For another example, for a U-turn lane at an intersection, the virtual lane line corresponding to the lane at the intersection, the virtual lane lines corresponding to each lane that can be driven in the intersection, and the virtual lane lines corresponding to each lane that can be driven after the vehicle turns around can be connected, so as to obtain a plurality of virtual lane lines for U-turns.

[0032] It should be noted that the road topology connection can be pre-constructed by technicians based on the topological relationship between the lane centerline and the lanes in the high-precision map, and stored in the map database. When the vehicle is driving, the vehicle obtains the stored road topology connection from the map database to predict the driving trajectory of the target obstacle vehicle.

[0033] The intersection entrance position corresponding to the target virtual lane line at the intersection can be the intersection position of the target virtual lane line and the stop line at the intersection entrance. The intersection exit position corresponding to the target virtual lane line at the intersection can be the intersection position of the target virtual lane line and the stop line position at the intersection exit. It can be understood that the intersection entrance position and the intersection exit position are two relative concepts, which can be jointly determined according to the driving direction and driving position of the target obstacle vehicle. The target distance can be the closest distance between the target obstacle vehicle and the target virtual lane line.

[0034] In specific implementation, since the behavior of pedestrians in road obstacles is irregular, it is impossible to predict the trajectory of pedestrians. Therefore, in specific implementation, the embodiment of the present disclosure may only perform trajectory prediction for the obstacle vehicles in the road obstacles. Among them, trajectory prediction is not performed for the stationary vehicles in the obstacle vehicles.

[0035] In specific implementation, during the driving process of the vehicle, the prediction module can obtain the road environment information collected by the vehicle in real time, as well as the obstacle information of each road obstacle. At the same time, according to the position of the vehicle, a high-precision map and road topology connection matching the position of the vehicle can be obtained from the map database. Target obstacle vehicles that may affect the driving of the vehicle are screened out from the road obstacles. The prediction module first performs scene detection on the vehicle and the target obstacle vehicle, that is, detects whether the vehicle and the target obstacle vehicle are located at the intersection. If so, the trajectory prediction method provided by the embodiment of the present disclosure can be used to predict the trajectory of the target obstacle vehicle. If not, the trajectory prediction method in the prior art can be used to predict the trajectory of the target obstacle vehicle. Among them, whether the vehicle is located at the intersection can be determined based on the high-precision map or road topology connection matching the position of the vehicle. Whether the obstacle vehicle is located at the intersection can be determined based on the vehicle information and high-precision map of the target obstacle vehicle, for example, based on the position of the obstacle vehicle in the vehicle information, the lane information, the driving direction of the vehicle, etc.

[0036] When it is detected that the ego vehicle and the target obstacle vehicle are at an intersection and are about to pass through the intersection, for each target virtual lane line, the connection direction of the target virtual lane line between the intersection entrance position and the intersection exit position can be determined, and the target distance between the target obstacle vehicle and the target virtual lane line can be determined based on the real-time vehicle position in the vehicle information of the target obstacle vehicle and the actual position corresponding to the target virtual lane line in the high-precision map.

[0037] Optionally, when determining the target distance, the center position of the rear axle of the vehicle can be determined based on the real-time position of the target obstacle vehicle and the attribute information of the vehicle, and the target distance can be determined based on the center position and the actual position of the target virtual lane connection line.

[0038] It is understandable that the obstacle vehicles detected by the ego vehicle may include multiple ones, so in specific implementation, the target obstacle vehicles that are located at the same intersection as the ego vehicle and are about to pass through the intersection can be detected from the multiple obstacle vehicles. For these target obstacle vehicles, the target distance between each target obstacle vehicle and each target virtual lane line and the target discrete line segment corresponding to the target obstacle vehicle on each target virtual lane line are determined.

[0039] S102: According to the line direction and target distance corresponding to the target virtual lane line, a matching lane line that matches the target obstacle vehicle is selected from each target virtual lane line.

[0040] Here, the matching lane line may be a line among multiple target virtual lane lines that best matches the future driving trajectory of the target obstacle vehicle. After obtaining the matching lane line, the driving trajectory of the obstacle vehicle is predicted based on the matching lane line, which can effectively improve the prediction accuracy.

[0041] In specific implementation, the deviation angle between the target obstacle vehicle and each target virtual lane line can be determined based on the vehicle driving direction in the vehicle information of the target obstacle vehicle and the line direction corresponding to each target virtual lane line. Then, the matching degree of each target virtual lane line with the target obstacle vehicle is determined based on the target distance and deviation angle between the target obstacle vehicle and each target virtual lane line, and the target virtual lane line with the highest matching degree is used as the matching lane line.

[0042] Optionally, when determining the target distance and line direction corresponding to each target virtual lane line, the tangent direction of the position point where the target obstacle vehicle is closest to each target virtual lane line can also be determined, wherein the tangent direction is related to the driving direction of the target obstacle vehicle. Then, when screening the matching lane lines, the deviation angle between the target obstacle vehicle and each target virtual lane line can be determined based on the vehicle driving direction of the target obstacle vehicle and the line direction and tangent direction corresponding to each target virtual lane line. Then, based on the target distance and deviation angle between the target obstacle vehicle and each target virtual lane line, the matching degree of each target virtual lane line with the target obstacle vehicle is determined, and the target virtual lane line with the highest matching degree is used as the matching lane line.

[0043] S103: Determine the predicted trajectory of the target obstacle vehicle according to the vehicle information of the target obstacle vehicle, the road environment information and the matching lane connection.

[0044] In specific implementation, for each target obstacle vehicle, after obtaining the matching lane line corresponding to the target obstacle vehicle, the predicted trajectory of the target obstacle vehicle at the intersection can be calculated by using the trajectory prediction method of the prior art, taking into full consideration the traffic rule constraints, the interaction between different traffic objects at the intersection, etc., based on the kinematic posture information in the vehicle information of the target obstacle vehicle (for example, the current driving direction, current speed, current acceleration, vehicle type of the target obstacle vehicle), historical kinematic posture information, road environment information (for example, lane rule constraints, traffic light constraints, tidal lane constraints, etc.), and the real lane information corresponding to the matching lane line in the high-precision map. For example, the polynomial curve generation method can be used to combine the vehicle information of the target obstacle vehicle, the road environment information and the real lane information corresponding to the matching lane line to determine the predicted trajectory corresponding to the target obstacle vehicle.

[0045] After obtaining the predicted trajectory of the target obstacle vehicle, the predicted trajectory can be input into the PNC module for decision planning.

[0046] In this way, based on the topological relationship between lanes at the intersection, various drivable road options can be fully considered to construct various target virtual lane lines for vehicle driving. By combining the distance between the target virtual lane line and the target obstacle vehicle and the line direction corresponding to the target virtual lane line, the target obstacle vehicle and the target virtual lane line can be matched in two dimensions, namely, distance and end point direction, so as to more accurately judge the matching degree between the target obstacle vehicle and each target virtual lane line. Using the matching degree, the matching lane line that the target obstacle vehicle is most likely to drive can be accurately screened, and then the driving trajectory of the target obstacle vehicle can be predicted using the matching lane line, which can improve the accuracy of the predicted driving trajectory. Especially when the ego vehicle and the target obstacle vehicle are at the intersection, using two-dimensional information for trajectory prediction can effectively improve the prediction accuracy and rationality of the driving trajectory of the target obstacle vehicle in the intersection scene, thereby improving the safety of autonomous driving.

[0047] In one embodiment, in order to further reduce the amount of calculation during trajectory prediction, after the target virtual lane lines are obtained, they can be discretized first, and then the discretized lines can be used for trajectory prediction. Specifically, the target virtual lane lines can be discretized to obtain multiple discrete line segments corresponding to each target virtual lane line; then the line segment direction of the target discrete line segment closest to the target obstacle vehicle among the multiple discrete line segments corresponding to each target virtual lane line is determined.

[0048] Here, a target virtual lane line can be discretized into multiple discrete line segments, each discrete line segment is a line segment, and the starting point of the first discrete line segment among the multiple discrete line segments can be the starting point of the target virtual lane line, and the end point can be a position point on the target virtual lane line that is a set distance from the starting point. The starting point of the next discrete line segment is the end point of the previous discrete line segment.

[0049] Exemplarily, after obtaining the road topology connection, the prediction module can use the line segment discretization technology to discretize each target virtual lane connection in the road topology connection to obtain multiple virtual line segments corresponding to the target virtual lane connection. Then, for each target virtual lane connection, the target discrete line segment closest to the target obstacle vehicle can be determined from multiple discrete line segments according to the real positions corresponding to the multiple virtual line segments corresponding to the target virtual lane connection in the high-precision map and the real-time position of the target obstacle vehicle, and the direction from the line segment start point to the line segment end point of the target discrete line segment is used as the line segment direction of the target discrete line segment. Among them, the line segment start point to the line segment end point of the target discrete line segment are two relative concepts, which can be jointly determined according to the driving direction and driving position of the target obstacle vehicle. In addition, the distance between the target obstacle vehicle and the target discrete line segment can be used as the target distance between the target obstacle vehicle and the target virtual lane connection.

[0050] Optionally, when determining the target discrete line segments and target distances, the center position of the rear axle of the vehicle can be determined based on the real-time position of the obstacle vehicle and the attribute information of the vehicle, and the target discrete line segments and target distances can be determined based on the center position and the actual position corresponding to the discrete line segments in the high-precision map.

[0051] like Figure 3 , which is a schematic diagram of a right turn scenario provided by an embodiment of the present disclosure, wherein each dotted line segment in the target virtual lane line 1 and the target virtual lane line 2 represents a discrete line segment in the target virtual lane line. The rectangular box represents the target obstacle vehicle, h represents the vehicle driving direction of the target obstacle vehicle (that is, the vehicle speed direction), a represents the line direction of the target virtual lane line 1, and Figure 3 In the figure, the direction of the target virtual lane line 1 between the intersection entrance and exit. b represents the direction of the target discrete line segment. Figure 3 where is the line direction of the target discrete line segment closest to the target obstacle vehicle in the target virtual lane line 1. d represents the target distance. Figure 3 represents the target discrete line segment in the target virtual lane line 1 and the target distance between the target obstacle vehicle. Figure 3In FIG, only the line direction corresponding to the target virtual lane line 1, and the line segment direction and target distance of the target discrete line segment corresponding to the target virtual lane line 1 and the target obstacle vehicle are shown. In a specific implementation, the line direction corresponding to the target virtual lane line 2, and the line segment direction and target distance of the target discrete line segment corresponding to the target virtual lane line 2 and the target obstacle vehicle can also be determined.

[0052] Understandably, assuming that there are n target virtual lane lines at the intersection, and multiple target obstacle vehicles are detected at the intersection, then for each target obstacle vehicle, the line segment direction b and target distance d of the target discrete line segment corresponding to the target obstacle vehicle and each of the n target virtual lane lines can be determined. That is, there are n corresponding a, b, and d between the target obstacle vehicle and the n target virtual lane lines, which can be expressed as a1~an, b1~bn, and d1~dn respectively.

[0053] like Figure 4 , which is a schematic diagram of a straight-through right turn scenario provided by an embodiment of the present disclosure, wherein each dotted line segment in the target virtual lane line 3 and the target virtual lane line 4 represents a discrete line segment in the target virtual lane line. The rectangular box represents the target obstacle vehicle, h represents the vehicle driving direction of the target obstacle vehicle, and Figure 4 In , a represents the direction of the target virtual lane line 4 between the intersection entrance position and the intersection exit position, b represents the direction of the target discrete line segment in the target virtual lane line 4 that is closest to the target obstacle vehicle, and d represents the target distance between the target discrete line segment in the target virtual lane line 4 and the target obstacle vehicle. Figure 4 In FIG, only the line direction corresponding to the target virtual lane line 4, and the line segment direction and target distance of the target discrete line segment corresponding to the target virtual lane line 4 and the target obstacle vehicle are shown. In a specific implementation, the line direction corresponding to the target virtual lane line 3, and the line segment direction and target distance of the target discrete line segment corresponding to the target virtual lane line 3 and the target obstacle vehicle can also be determined.

[0054] Furthermore, after determining the line segment direction of each target discrete line segment in the target virtual lane line, the above S102 may be implemented according to the following steps: According to the line segment direction of the target discrete line segment in the target virtual lane line, the line direction corresponding to the target virtual lane line, and the target distance, a matching lane line is screened out from each target virtual lane line.

[0055] In specific implementation, for the target obstacle vehicle and any target virtual lane line, the deviation angle between the target obstacle vehicle and the target virtual lane line can be determined according to the vehicle driving direction of the target obstacle vehicle, the line direction corresponding to the target virtual lane line, and the line direction of the target discrete line segment in the target virtual lane line. Then, according to the target distance and deviation angle between the target obstacle vehicle and the target virtual lane line, the matching degree between the target virtual lane line and the target obstacle vehicle is determined. Among the target virtual lane lines, the target virtual lane line with the highest matching degree with the target obstacle vehicle is used as the matching lane line.

[0056] In this way, by discretizing the target virtual lane line, the smooth and possibly curved target virtual lane line can be discretized into multiple continuous line segments. Then, the discrete line segments are used for subsequent trajectory prediction, without the need to use all the position points on the target virtual lane line, thereby effectively reducing the amount of calculation. By combining the distance between the target discrete line segment and the target obstacle vehicle, the line direction corresponding to the target virtual lane line, and the line direction of the target discrete line segment, the target obstacle vehicle and the target virtual lane line can be matched in three dimensions: distance, instantaneous orientation, and end point orientation, thereby more accurately judging the matching degree between the target obstacle vehicle and each target virtual lane line.

[0057] In one embodiment, the target virtual lane line may be discretized according to the following steps: For any target virtual lane line, sampling is performed in the target virtual lane line according to a preset sampling distance to obtain multiple sampling points; and the line segments between adjacent sampling points are used as discrete line segments.

[0058] Here, the preset sampling distance can be set based on experience and is not specifically limited here.

[0059] In specific implementation, for any target virtual lane line, starting from the starting point of the target lane corresponding to the target virtual lane line, sampling can be performed on the target lane corresponding to the target virtual lane line according to the preset sampling distance to obtain the corresponding lane sampling points on the target lane. Among them, the starting point of the target lane is used as the first lane sampling point. Then, the position of each lane sampling point corresponding to the target virtual lane line is used as the corresponding sampling point on the target virtual lane line. Then, the line segment between every two adjacent adjacent sampling points can be used as a discrete line segment, so as to obtain the corresponding discrete line segments of the target virtual lane line.

[0060] That is, in the embodiment of the present disclosure, when discretizing the target virtual lane line, the continuous target virtual lane line can be cut into discrete line segments of fixed length and connected end to end. The purpose of the discretization process is to reduce the data level of the lane at the intersection and increase the speed of the intersection topology matching while improving the robustness of the intersection topology matching.

[0061] In one embodiment, before executing S103, since there are a large number of target virtual lane lines at the intersection, not all lanes corresponding to the target virtual lane lines are lanes that the target obstacle vehicle may travel. For example, virtual lane lines of the opposite lane, virtual lane lines at a longer distance, virtual lane lines that do not comply with traffic rules, etc. are all virtual lane lines that the obstacle vehicle will not travel on. Therefore, in order to avoid the influence of such virtual lane lines on trajectory prediction, the following steps can be used to preliminarily screen the virtual lane lines to eliminate the virtual lane lines that the obstacle vehicle will not travel on: Step 1: Determine the road feature information of each virtual lane line according to the road environment information and the lane information corresponding to each virtual lane line.

[0062] Here, each virtual lane line constructed in advance according to the virtual lane lines corresponding to each real lane at the intersection in the high-precision map and the topological relationship between each lane can be used as a virtual lane line. The lane information corresponding to the virtual lane line is the lane information of the real target lane corresponding to the virtual lane line in the high-precision map. Since a virtual lane line may be obtained by connecting multiple virtual lane lines, the target lane corresponding to a virtual lane line in the high-precision map may include multiple lanes, namely, lane 1 where the vehicle is located before arriving at the intersection, virtual lane 2 where the vehicle travels from the entrance of the intersection to the exit of the intersection, and lane 3 where the vehicle leaves the intersection. Therefore, the lane information corresponding to a virtual lane line may include the lane information of the multiple lanes.

[0063] Lane information may include the lane type, lane width, lane turning type, lane boundary information, etc. of the target lane, wherein the lane type may include motor vehicle lanes, non-motor vehicle lanes, sidewalks, bus lanes, etc.; the lane turning type may include left turn type, right turn type, straight type, one-way lane type, non-U-turn type, U-turn type, straight and left turn type, left turn and U-turn type, straight and right turn type, etc.

[0064] Road feature information may include, for example, the lane position, lane type, lane width, lane turn type, lane road boundary, whether the lane is drivable, whether there is an accident or fault in the lane, lane congestion, etc. of the target lane corresponding to the virtual lane line. High-precision maps include various prior lane information, such as tidal lane rules, turn types, whether the lane is occupied by construction, whether the road is maintained, etc. Based on the prior information in high-precision maps, more accurate road feature information can be determined.

[0065] In specific implementation, for each virtual lane line, the preprocessing engine can be used to extract road features in combination with the road environment information and the lane information of each target lane corresponding to the virtual lane line in the high-precision map, and obtain the road feature information of the virtual lane line. Alternatively, in order to further improve the division granularity of the road feature information, the preprocessing engine can also be used to discretize the virtual lane line to obtain multiple discrete line segments corresponding to each virtual lane line, and each discrete line segment is used as a lane node lane_node. At the same time, the road environment information and the lane information of each target lane corresponding to the virtual lane line in the high-precision map can be combined to extract road features, and the road feature information of each lane_node corresponding to the virtual lane line can be obtained and bound to the lane_node data structure.

[0066] For example, for the target lane corresponding to the virtual lane line in the high-precision map, when the target lane is a tidal lane, the road feature information of each lane_node corresponding to the virtual lane line can be determined based on the road environment information, such as different driving rules corresponding to the tidal lane in different time periods, restrictions on road traffic information during road construction, changes in road rules during temporary control, and the lane information of the target lane itself.

[0067] Step 2: Filter out each target virtual lane line from each virtual lane line according to the vehicle type and vehicle driving direction in the vehicle information, the lane type and lane position in the road feature information, and the target search range corresponding to the target obstacle vehicle.

[0068] Here, different types of target obstacle vehicles have different lane driving restrictions. For example, motor vehicles cannot drive on non-motorized lanes, non-motorized vehicles cannot drive on motorized lanes, and social vehicles cannot occupy bus lanes during a specific time period. The direction of vehicle travel can determine which reverse lanes are in the intersection. When predicting the trajectory, the target obstacle vehicle will definitely not drive towards the reverse lane, and the vehicle in the reverse lane will not affect the travel of the target obstacle vehicle. The lane position in the road feature information of the discrete line segment and the vehicle travel direction of the target obstacle vehicle affect which reverse lane is, that is, which reverse discrete line segment is.

[0069] The target search range may be a range determined according to the real-time position of the target obstacle vehicle and a preset search radius. The target virtual lane lines corresponding to the lanes within the range are lane lines on which the target obstacle vehicle may travel, and the target virtual lane lines corresponding to the lanes outside the range are likely to be lane lines on which the target obstacle vehicle may not travel.

[0070] In specific implementation, for each target obstacle vehicle, the target search range can be determined according to the real-time vehicle position and preset search radius in the vehicle information of the target obstacle vehicle, and the first virtual lane lines that match the target search range can be determined according to the target search range and the lane position in the road feature information of each lane_node. Then, according to the vehicle type and vehicle driving direction and other information in the vehicle information of the target obstacle vehicle, and the lane type and lane position and other information in the road feature information of the lane_node in each first virtual lane line, the second virtual lane lines whose lane type matches the vehicle type and does not belong to the opposite lane can be screened out from the first virtual lane lines, and these second virtual lane lines are used as the target virtual lane lines that the target obstacle vehicle can travel.

[0071] For example, the virtual lane lines at the intersection include virtual lane lines 1 to 10. The above steps 1 and 2 can be used to select the target virtual lane lines that the target obstacle vehicle can travel on from virtual lane lines 1 to 10. Assuming that the target virtual lane lines that the target obstacle vehicle can travel on are selected as virtual lane lines 3 to 7, then for the above S101, only the line directions corresponding to the virtual lane lines 3 to 7 can be calculated, and the line segment directions and target distances of the target discrete line segments corresponding to the target obstacle vehicle and the virtual lane lines 3 to 7 can be calculated. Virtual lane lines 1 to 2 and virtual lane lines 8 to 10 are not included in the calculation range.

[0072] In this way, by first screening multiple virtual lane lines and then calculating the corresponding dimensional information between the screened target virtual lane lines and the target obstacle vehicles in different dimensions, it is possible to avoid adding virtual lane lines that the target obstacle vehicles will not travel to the calculation, thereby effectively reducing the amount of calculation and improving the calculation speed.

[0073] Optionally, when the computing power of the prediction module is strong enough, the virtual lane lines may not be screened, and the line segment direction and target distance of the target discrete line segment corresponding to the target obstacle vehicle and each virtual lane line may be directly calculated.

[0074] In another embodiment, before executing S101, since there may be multiple obstacle vehicles detected at the intersection, not all of them will affect the driving of the vehicle at the intersection. For example, obstacle vehicles in the opposite lane and obstacle vehicles at a long distance are obstacle vehicles that do not affect the driving of the vehicle. Therefore, in order to avoid the influence of such obstacle vehicles on trajectory prediction, the following steps can be used to preliminarily screen the obstacle vehicles to eliminate the obstacle vehicles that will not affect the driving of the vehicle.

[0075] Specifically, when multiple obstructing vehicles are detected at the intersection, at least one target obstructing vehicle that affects the driving of the own vehicle is screened out from the multiple obstructing vehicles according to the vehicle information of each obstructing vehicle and the vehicle information of the own vehicle.

[0076] For example, after collecting the road environment information and the vehicle information of each obstacle vehicle, the two types of information can be preprocessed to adjust the unreasonable information and obtain the preprocessed road environment information and vehicle information. Then, the preprocessed road environment information and vehicle information are used to perform the above steps 1 and 2, as well as the step of screening obstacle vehicles.

[0077] The preprocessing of the road environment information can specifically be to extract information that affects the driving of the obstacle vehicle from the road environment information. For example, variable information in the road environment is extracted, such as driving rules of tidal lanes at different times, traffic light color information, temporary traffic sign information, etc.; road matching is performed based on the traffic light information on the road to determine the lanes where the obstacle vehicle can drive and the driving behaviors that can be performed on the lanes.

[0078] The preprocessing of vehicle information may specifically include current information feature processing and historical information feature processing. Among them, information feature processing may specifically include denoising vehicle information, adjusting unreasonable vehicle information (such as adjusting unreasonable acceleration information, etc.), etc.

[0079] After obtaining the pre-processed road environment information and vehicle information, if it is detected that there are multiple obstacle vehicles in the intersection scene, then according to the pre-processed vehicle information corresponding to each obstacle vehicle (such as the real-time position, orientation, lane, driving direction, speed, acceleration, historical kinematic posture information, etc. of the obstacle vehicle) and the vehicle information of the self-vehicle (such as the orientation of the self-vehicle, the position of the self-vehicle, the speed of the self-vehicle, the lane information of the lane where the self-vehicle is located, etc.), the target obstacle vehicles whose driving behavior complies with traffic rules and will affect the driving of the self-vehicle can be determined from the multiple obstacle vehicles. Among them, the number of the selected target obstacle vehicles can be one or more. For example, from multiple obstacle vehicles, each target obstacle vehicle that is within the preset range of the self-vehicle, is in the same driving direction as the self-vehicle, is in the associated lane with the self-vehicle, and may overlap with the self-vehicle in the future can be selected.

[0080] It is understandable that after obtaining the vehicle information of the own vehicle, the vehicle information of the own vehicle may be preprocessed first, and then the preprocessed vehicle information may be used.

[0081] Further, after screening out each target obstacle vehicle that affects the driving of the vehicle, the above S101 can be executed. Exemplarily, after determining each target obstacle vehicle that affects the driving of the vehicle, the line direction corresponding to each target virtual lane line can be determined respectively. At the same time, for any target obstacle vehicle, the line segment direction and target distance of the target discrete line segment corresponding to the target obstacle vehicle and each target virtual lane line can be determined.

[0082] For example, the obstacle vehicles at the intersection include vehicles 1-6. The above steps are used to screen out the target obstacle vehicles that affect the driving of the vehicle as vehicles 3-6 from vehicles 1-6. Then, the segment directions and target distances of the target discrete segments corresponding to vehicles 3-6 and the connecting lines of each target virtual lane can be determined respectively.

[0083] In this way, by first screening the multiple detected obstacle vehicles and then calculating the corresponding dimensional information between the screened target obstacle vehicles and the target virtual lane connection in different dimensions, it is possible to avoid adding obstacle vehicles that will not affect the driving of the vehicle into the calculation, thereby effectively reducing the amount of calculation and improving the calculation speed.

[0084] Optionally, in order to further reduce the amount of calculation, when there are multiple obstacle vehicles at the intersection, each target obstacle vehicle that affects the driving of the vehicle can be screened out from the multiple obstacle vehicles, and then, for each target obstacle vehicle, each target virtual lane line that the target obstacle vehicle can drive can be screened out from the multiple virtual lane lines at the intersection. Then, the line segment direction and target distance of the target discrete line segment corresponding to the target obstacle vehicle and each target virtual lane line are calculated.

[0085] In this way, by screening both obstacle vehicles and virtual lane connections, the influence of obstacle vehicles that will not affect the driving of the ego vehicle and virtual lane connections where the obstacle vehicles will not drive on the occupancy of computing resources and the computing speed can be avoided, thereby improving the speed of subsequent trajectory prediction.

[0086] like Figure 5 As shown, it is an architecture diagram of a trajectory prediction provided by an embodiment of the present disclosure, which may include an input part, a prediction module processing part and an output part, wherein the input part may include road environment information collected by the vehicle in real time, obstacle information of each road obstacle, high-precision map in the map database and road topology connection. The prediction module processing can be divided into three parts, one of which is to use the map preprocessing engine to perform road discretization (i.e., discretization processing of multiple virtual lane connections respectively), road feature extraction (specifically, refer to the content of step 1 above), and other processing, where the other processing is some conventional processing means in the prior art, such as denoising processing, optimization processing, etc. One part is to use the road environment information preprocessing engine to perform traffic light road matching, extract variable information in the road environment, and other processing, where the other processing is also some conventional processing means in the prior art, such as integration processing, optimization processing, and deletion processing of useless information of road environment information. The last part is to use the obstacle preprocessing engine to preprocess the obstacle information, which can specifically include current information feature processing and historical information feature processing, as well as other processing. The other processing here is also some conventional processing means in the prior art, such as feature alignment processing, feature adjustment processing, etc. After obtaining the preprocessed information corresponding to the three preprocessing engines, the scene recognition module can be used to determine the scene where the ego vehicle and the target obstacle vehicle are located. The scene here includes the intersection scene and other scenes. Other scenes are any scene other than the intersection scene, such as the high-speed ramp scene, the tunnel scene, etc. When the scene recognition module determines that the ego vehicle and the target obstacle vehicle are both located in the intersection scene, the intersection topology matching processing (that is, executing the above S101~S103) can be performed to determine the matching lane line that matches the target obstacle vehicle. Then, the trajectory generation module of the prior art can be used to predict the predicted trajectory corresponding to the target obstacle vehicle using the matching lane line, obstacle information (specifically the vehicle information of the target obstacle vehicle), road environment information, etc., and input it into the PNC module of the output part for decision planning.

[0087] In one embodiment, the above S102 may be performed according to the following voting and scoring mechanism: S102-1: Determine a first score of the target virtual lane line according to the target distance corresponding to the target virtual lane line, wherein the first score is used to represent the degree of proximity between the target virtual lane line and the target obstacle vehicle.

[0088] Here, since the distance between the target obstacle vehicle and the target virtual lane line affects whether the target obstacle vehicle will travel toward the target virtual lane line, scoring is performed using the target distance, and then the first score is used to screen the matching lane lines, which can improve the screening accuracy.

[0089] During specific implementation, a preset distance and score conversion formula may be used to convert the target distance into the first score.

[0090] In one embodiment, the smaller the first score is, the closer the distance between the target virtual lane line and the target obstacle vehicle is. The above S102-1 can be implemented according to the following steps: A first score is determined according to a ratio of a target distance corresponding to the target virtual lane line to a preset distance.

[0091] Here, the preset distance can be set based on experience and is not specifically limited here. Optionally, the preset distance can be consistent with the preset search radius used when determining the target search range above. The first score represents the scoring of the distance dimension, and the closer the target distance between the target obstacle vehicle and the target virtual lane line. In one embodiment, the lower the first score, the higher the possibility that the target obstacle vehicle matches the target virtual lane line.

[0092] by Figure 3 Take this as an example to illustrate. Figure 3 The target virtual lane line 1 in the image may determine a first score of the target virtual lane line 1 according to a target distance d and a preset distance D corresponding to the target obstacle vehicle and the target virtual lane line 1.

[0093] Specifically, for any target virtual lane connection, the first score can be determined using the following formula 1: ; (Formula 1) in, represents the first score between the target obstacle vehicle and the mth target virtual lane line, represents the target distance between the target obstacle vehicle and the mth target virtual lane line, and D represents the preset distance.

[0094] S102-2: Determine a second score of the target virtual lane line according to the line segment direction of the target discrete line segment in the target virtual lane line and the line direction corresponding to the target virtual lane line, wherein the second score is used to characterize the deviation angle of the target virtual lane line relative to the target obstacle vehicle.

[0095] Here, the second score is used to characterize the score under the deviation angle dimension. In one embodiment, the lower the second score, the smaller the deviation angle between the target obstacle vehicle and the target virtual lane connection, and the higher the possibility of matching between the target obstacle vehicle and the target virtual lane connection.

[0096] Exemplarily, for the target obstacle vehicle and any target virtual lane connection, the mathematical representation of the connection direction corresponding to the target virtual lane connection, the mathematical representation of the line segment direction corresponding to the target discrete line segment, and the mathematical representation of the vehicle driving direction of the target obstacle vehicle can be determined. Here, the direction can be represented by 0-2PI in mathematics, corresponding to 0-360 degrees, and addition and subtraction operations can be performed after the direction is represented using the mathematical representation. Then, the second score between the target obstacle vehicle and the target virtual lane connection is determined using the mathematical representation corresponding to the connection direction, the mathematical representation of the line segment direction, and the mathematical representation of the vehicle driving direction. For example, the mathematical representation corresponding to the connection direction and the mathematical representation of the line segment direction are determined, and the difference between them and the mathematical representation of the vehicle driving direction is determined, and then the second score is determined using the average value of the difference and the ratio of the set deviation angle.

[0097] S102-3: Determine a matching score between the target virtual lane line and the target obstacle vehicle according to the first score and the second score.

[0098] For example, the average of the first score and the second score can be used as the matching score between the target obstacle vehicle and the target virtual lane line. Alternatively, the first score and the second score can be weighted summed using the preset weights set for the distance dimension and the deviation angle dimension to obtain the matching score between the target obstacle vehicle and the target virtual lane line.

[0099] S102-4: Filter out matching lane lines from each target virtual lane line according to the matching score.

[0100] For example, since a smaller matching score indicates a closer match between the target obstacle vehicle and the target virtual lane line, for any target obstacle vehicle, the target virtual lane line with the smallest matching score may be used as the matching lane line that best matches the target obstacle vehicle.

[0101] In another alternative implementation, the higher the first score, the closer the distance between the target virtual lane line and the target obstacle vehicle, that is, the higher the possibility of the target obstacle vehicle matching the target virtual lane line; the higher the second score, the smaller the deviation angle between the target obstacle vehicle and the target virtual lane line, and the higher the possibility of the target obstacle vehicle matching the target virtual lane line. Therefore, the higher the matching score, the closer the target obstacle vehicle matches the target virtual lane line, so for any target obstacle vehicle, the target virtual lane line with the highest matching score can be used as the matching lane line that best matches the target obstacle vehicle.

[0102] In one embodiment, the second score may include a first deviation score and a second deviation score, wherein the first deviation score is used to characterize the deviation angle of the target virtual lane line relative to the target obstacle vehicle, and the second deviation score is used to characterize the deviation angle of the target discrete line segment relative to the target obstacle vehicle. The above S102-2 may be implemented according to the following steps: S102-2-1: Determine a first deviation score according to a set deviation angle, a connection direction corresponding to a target virtual lane connection, and a vehicle driving direction of a target obstacle vehicle.

[0103] Here, the first deviation score represents the score of the angle deviation value of the target obstacle vehicle relative to the entire target virtual lane connection line (lane) of the intersection. The lower the score, the more the vehicle driving direction of the target obstacle vehicle matches the direction of the entire lane. The deviation angle can be set based on experience and is not specifically limited in the implementation of this disclosure.

[0104] Exemplarily, the first deviation score may be determined according to the following formula 2: ; (Formula 2) in, represents the first deviation score between the target obstacle vehicle and the mth target virtual lane line, It represents the mathematical representation of the direction in which the m-th target virtual lane line is connected from end to end (i.e., the direction of the line corresponding to the m-th target virtual lane line), h represents the mathematical representation of the vehicle driving direction of the target obstacle vehicle, and H represents the maximum deviation angle of the target obstacle vehicle relative to the lane, which is the above-mentioned set deviation angle.

[0105] by Figure 3 Take this as an example to illustrate. Figure 3 For the target virtual lane line 1 in the figure, the mathematical representation of the vehicle driving direction h of the target obstacle vehicle, the mathematical representation of the line direction a corresponding to the target virtual lane line 1, and the set deviation angle H can be substituted into the above formula 2 to obtain the first deviation score corresponding to the target obstacle vehicle.

[0106] S102-2-2: Determine a second deviation score according to the set deviation angle, the line segment direction of the target discrete line segment and the vehicle driving direction.

[0107] Here, the second deviation score represents a score of the angle deviation value of the target obstacle vehicle relative to the target discrete line segment. The lower the score, the more the vehicle driving direction of the target obstacle vehicle matches the direction of the target discrete line segment.

[0108] Exemplarily, the second deviation score may be determined according to the following formula 3: ; (Formula 3) in, represents the second deviation score between the target obstacle vehicle and the mth target virtual lane line, represents the mathematical representation of the segment direction of the target discrete segment in the mth target virtual lane line, h represents the mathematical representation of the vehicle driving direction of the target obstacle vehicle, and H represents the set deviation angle.

[0109] by Figure 3 Take this as an example to illustrate. Figure 3 The target virtual lane line 1 in the target virtual lane line 1 can be substituted into the above formula 3 with the mathematical representation of the vehicle driving direction h of the target obstacle vehicle, the mathematical representation of the line segment direction b of the target discrete line segment in the target virtual lane line 1, and the set deviation angle H to obtain the second deviation score corresponding to the target obstacle vehicle.

[0110] Furthermore, in the case where the second score includes the first deviation score and the second deviation score, the matching score between the target obstacle vehicle and the target virtual lane line may be determined according to the following formula 4: ; (Formula 4) in, represents the matching score between the target obstacle vehicle and the mth target virtual lane line, Represents the preset weight set under the distance dimension, represents the preset weight under the dimension corresponding to the first deviation score, represents the preset weight under the dimension corresponding to the second deviation score, where m∈[1,n], n is the number of target virtual lane links.

[0111] Exemplarily, for a target obstacle vehicle and any target virtual lane line, the preset weights set in the three dimensions can be used to perform a weighted sum of the corresponding scores of the target obstacle vehicle and the target virtual lane line in the three dimensions (i.e., the first score, the first deviation score, and the second deviation score) to obtain a matching score between the target obstacle vehicle and the target virtual lane line.

[0112] In one embodiment, since the position and kinematic state of the target obstacle vehicle may change in real time, the matching lane line determined according to the real-time position of the vehicle may change. In this case, it is also necessary to determine whether the predicted trajectory of the target obstacle vehicle needs to be updated using the newly determined matching lane line. Specifically, after executing S103, the predicted trajectory can also be updated according to the following steps: S1: When the loop detection conditions are met, the target distance between each target virtual lane line and the target obstacle vehicle is re-determined.

[0113] S2: Determine a new matching lane line that matches the target obstacle vehicle according to the re-determined target distance and the line direction of the target virtual lane line.

[0114] Here, the cyclic detection condition is the condition for recalculating the matching lane connection that matches the target obstacle vehicle. Exemplarily, the cyclic detection condition can be a detection every N1 seconds, a detection every N2 meters of the target obstacle vehicle, and a detection every N3 meters of the vehicle. Among them, N1, N2, and N3 can all be set based on experience, and the embodiments of the present disclosure do not specifically limit them.

[0115] For the target obstacle vehicle, the new matching lane line can be regarded as the matching lane line determined by the real-time position of the vehicle at the new moment. For example, moment 1 is before moment 2. According to the real-time position of the target obstacle vehicle at moment 1, the matching lane line at moment 1 can be determined; according to the real-time position of the target obstacle vehicle at moment 2, the matching lane line at moment 2 can be determined. The matching lane line at moment 2 is relative to the matching lane line at moment 1, that is, the new matching lane line.

[0116] Exemplarily, after determining the predicted trajectory corresponding to the target obstacle vehicle at the previous moment by using the above steps S101 to S103, since the position of the target obstacle vehicle may be updated in real time, the above S101 can be returned to execute according to the real-time position of the target obstacle vehicle at the current moment, and the line segment direction and target distance of each target virtual lane line at the current moment are re-determined, and then the line segment direction of the re-determined target discrete line segment, the re-determined target distance, and the connection method of the target virtual lane line are used to determine the new matching lane line corresponding to the target obstacle vehicle at the current moment according to the above S102.

[0117] S3: When the new matching lane line is different from the matching lane line determined last time, if the matching score of the new matching lane line meets the trajectory update condition, a new predicted trajectory of the target obstacle vehicle is determined according to the vehicle information of the target obstacle vehicle, the road environment information and the new matching lane line.

[0118] Here, the new matching lane line can be regarded as the matching lane line selected according to the real-time position of the vehicle at the current moment. For example, if moment 2 is later than moment 1, and the real-time position of the vehicle at moment 2 and the real-time position of the vehicle at moment 1 meet the cycle detection condition, then the matching lane line determined last time can be the matching lane line determined at moment 1, and the new matching lane line can be the matching lane line determined at moment 2.

[0119] Based on the above embodiment, it can be known that when screening matching lane lines, it is necessary to use the matching score determined for the matching lane lines, so whether it is the matching lane lines determined last time or the new matching lane lines determined currently, they all have corresponding matching scores. Therefore, in the specific implementation, after obtaining the new matching lane lines, it can be determined whether the new matching lane lines are consistent with the matching lane lines determined last time. If they are consistent, the predicted trajectory determined last time can continue to be used, that is, it is determined that there is no need to update the predicted trajectory of the target obstacle vehicle. If they are inconsistent, since the predicted trajectory depends on the matching lane lines when determining, the predicted trajectory will definitely change after the matching lane lines change. Therefore, when they are inconsistent, it can be determined whether the matching score of the new matching lane lines meets the trajectory update conditions. In the case of meeting the trajectory update conditions, the new predicted trajectory of the target obstacle vehicle can be determined based on the latest vehicle information of the target obstacle vehicle, the latest road environment information and the new matching lane lines, and the new predicted trajectory is input into the PNC module for decision planning. The specific implementation process of the step of determining the new predicted trajectory is similar to the implementation process of S103 above, and will not be repeated here.

[0120] For example, when the matching lane line of the target obstacle vehicle is determined to be lane line 1 and the predicted trajectory is trajectory 1 according to the real-time vehicle position of the target obstacle vehicle at time 1, if the new matching lane line of the target obstacle vehicle is determined to be lane line 2 according to the real-time vehicle position of the target obstacle vehicle at time 2, it can be determined whether the matching score of lane line 1 meets the trajectory update condition. If so, the new predicted trajectory of the target obstacle vehicle, such as trajectory 2, can be determined using lane line 2, the vehicle information of the target obstacle vehicle, and the road environment information. If not, it can be determined that the predicted trajectory of the target obstacle vehicle is still trajectory 1.

[0121] The trajectory update condition may include, for example, condition 1 and condition 2. Condition 1 is that the matching score of the new matching lane connection is less than the matching score of the matching lane connection determined last time; condition 2 is that the matching score of the new matching lane connection is less than the target ratio of the matching score of the matching lane connection determined last time. The target ratio can be set based on experience and is not specifically limited here.

[0122] For example, the matched lane line at time 1 is lane line 1, the predicted trajectory is trajectory 1, the matched lane line at time 2 is lane line 2, and the trajectory update condition is condition 2. Then, the matching score of lane line 1 and the target ratio are multiplied to obtain the ratio score. Then, it is determined whether the matching score of lane line 2 is less than the ratio score. If so, it is determined that the matching score of lane line 2 meets the trajectory update condition; if not, it is determined that the matching score of lane line 2 does not meet the trajectory update condition. Through the relationship between the matching score of the new matching lane line and the target ratio of the matching score obtained last time, it is possible to timely update the predicted trajectory of the target obstacle vehicle when the matching score of the new matching lane line is significantly different from the matching score obtained last time.

[0123] It is understandable that when the target obstacle vehicle is continuously driving at the intersection, S1 to S3 can be continuously executed under the condition of satisfying the loop detection conditions until the target obstacle vehicle or the self-vehicle leaves the intersection. In this way, the predicted trajectory of the target obstacle vehicle can be updated in a timely manner, thereby ensuring the rationality of the decision-making planning made by the PNC module and reducing the abnormal deceleration and braking caused by PNC misjudgment.

[0124] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0125] In addition, an embodiment of the present disclosure further provides a vehicle, wherein the vehicle includes a controller, wherein the controller includes: a memory for storing machine-readable instructions; and The processor is used to call and execute the machine-readable instructions stored in the memory to implement the steps of the trajectory prediction method as described in the above embodiments.

[0126] Based on the same technical concept, the embodiment of the present application also provides a computer device. Figure 6 FIG. 1 is a schematic diagram of a computer device provided in an embodiment of the present application, including: Processor 601, memory 602 and bus 603. The memory 602 stores machine-readable instructions executable by the processor 601, and the processor 601 is used to execute the machine-readable instructions stored in the memory 602. When the machine-readable instructions are executed by the processor 601, the processor 601 executes the steps in the above trajectory prediction method embodiment.

[0127] The above-mentioned memory 602 includes a memory 6021 and an external memory 6022; the memory 6021 here is also called an internal memory, which is used to temporarily store the calculation data in the processor 601, as well as the data exchanged with the external memory 6022 such as a hard disk. The processor 601 exchanges data with the external memory 6022 through the memory 6021. When the computer device is running, the processor 601 communicates with the memory 602 through the bus 603, so that the processor 601 executes the execution instructions mentioned in the above method embodiment.

[0128] The embodiment of the present disclosure also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the trajectory prediction method described in the above method embodiment are executed. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0129] The computer program product of the trajectory prediction method provided in the embodiments of the present disclosure includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the steps of the trajectory prediction method described in the above method embodiment. For details, please refer to the above method embodiment, which will not be repeated here.

[0130] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK).

[0131] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0132] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0133] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0134] If the technical solution of this application involves personal information, the product using the technical solution of this application has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using the technical solution of this application has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, the personal information processing rules are notified by obvious signs / information, and the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

[0135] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the aforementioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be based on the protection scope of the claims.

Claims

1. A trajectory prediction method, characterized in that: include: When it is detected that the ego vehicle and the target obstacle vehicle are located at the intersection, determine the connection direction of each target virtual lane connection line between the intersection entrance position and the intersection exit position, and the target distance between each target virtual lane connection line and the target obstacle vehicle; the target virtual lane connection line is constructed according to the topological relationship between the lanes at the intersection; According to the line direction and the target distance corresponding to the target virtual lane line, a matching lane line that matches the target obstacle vehicle is selected from each of the target virtual lane lines; The predicted trajectory of the target obstacle vehicle is determined according to the vehicle information of the target obstacle vehicle, the road environment information and the matching lane lines.

2. The method according to claim 1, characterized in that: The method further comprises: Discretize the target virtual lane lines to obtain a plurality of discrete line segments corresponding to each of the target virtual lane lines; Determine the line segment direction of the target discrete line segment closest to the target obstacle vehicle among the multiple discrete line segments corresponding to each of the target virtual lane lines; The step of selecting a matching lane line that matches the target obstacle vehicle from each of the target virtual lane lines according to the line direction and the target distance corresponding to the target virtual lane line includes: The matching lane line is screened out from each of the target virtual lane lines according to the line segment direction of the target discrete line segment in the target virtual lane line, the line direction corresponding to the target virtual lane line, and the target distance.

3. The method according to claim 1 or 2, characterized in that: The target virtual lane line is a virtual lane line that matches the target obstacle vehicle, and the method further includes: Determine the road feature information of each virtual lane line according to the road environment information and the lane information corresponding to each virtual lane line; The target virtual lane line is screened out from the plurality of virtual lane lines according to the vehicle type and vehicle driving direction in the vehicle information, the lane type and lane position in the road feature information, and the target search range corresponding to the target obstacle vehicle.

4. The method according to claim 1 or 2, characterized in that: The method further comprises: When a plurality of obstructing vehicles are detected at the intersection, at least one target obstructing vehicle that affects the driving of the own vehicle is screened out from the plurality of obstructing vehicles according to the vehicle information of each obstructing vehicle and the vehicle information of the own vehicle.

5. The method according to claim 2, characterized in that: The step of selecting the matching lane line from each of the target virtual lane lines according to the line direction of the target discrete line segment in the target virtual lane line, the line direction corresponding to the target virtual lane line, and the target distance includes: Determine a first score of the target virtual lane line according to the target distance corresponding to the target virtual lane line, wherein the first score is used to represent the degree of proximity between the target virtual lane line and the target obstacle vehicle; Determine a second score of the target virtual lane line according to the line segment direction of the target discrete line segment in the target virtual lane line and the line direction corresponding to the target virtual lane line, wherein the second score is used to represent the deviation angle of the target virtual lane line relative to the target obstacle vehicle; Determining a matching score between the target virtual lane line and the target obstacle vehicle according to the first score and the second score; According to the matching score, the matching lane line is screened out from each of the target virtual lane lines.

6. The method according to claim 5, characterized in that The smaller the first score is, the closer the distance between the target virtual lane line and the target obstacle vehicle is; The step of determining a first score of the target virtual lane line according to the target distance corresponding to the target virtual lane line comprises: The first score is determined according to a ratio of the target distance corresponding to the target virtual lane line to a preset distance.

7. The method according to claim 5, characterized in that The second score includes a first deviation score and a second deviation score, wherein the first deviation score is used to represent the deviation angle of the target virtual lane line relative to the target obstacle vehicle, and the second deviation score is used to represent the deviation angle of the target discrete line segment relative to the target obstacle vehicle; The determining, according to the line segment direction of the target discrete line segment in the target virtual lane line and the line direction corresponding to the target virtual lane line, a second score of the target virtual lane line comprises: Determining the first deviation score according to a set deviation angle, a connection direction corresponding to the target virtual lane connection line, and a vehicle driving direction of the target obstacle vehicle; The second deviation score is determined according to the set deviation angle, the line segment direction of the target discrete line segment, and the vehicle driving direction.

8. The method according to claim 5, characterized in that The method further comprises: When the cyclic detection condition is met, re-determine the target distance between each of the target virtual lane lines and the target obstacle vehicle; Determine a new matching lane line that matches the target obstacle vehicle according to the re-determined target distance and the line direction of the target virtual lane line; In the case where the new matching lane line is different from the matching lane line determined last time, if the matching score of the new matching lane line meets the trajectory update condition, a new predicted trajectory of the target obstacle vehicle is determined according to the vehicle information of the target obstacle vehicle, the road environment information and the new matching lane line.

9. The method according to claim 8, characterized in that The trajectory update condition includes: the matching score of the new matching lane line is less than the target ratio of the matching score of the matching lane line determined last time.

10. The method according to claim 2, characterized in that The discretization process of the target virtual lane line to obtain a plurality of discrete line segments corresponding to each of the target virtual lane lines includes: For any of the target virtual lane lines, sampling is performed in the target virtual lane line according to a preset sampling distance to obtain a plurality of sampling points; The line segments between adjacent sampling points are taken as the discrete line segments.

11. A vehicle, characterized in that: A controller is included, the controller comprising: a memory for storing machine-readable instructions; and A processor, configured to call and execute the machine-readable instructions stored in the memory to implement the steps of the trajectory prediction method according to any one of claims 1 to 10.

12. A computer device, characterized in that: include: A processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and the processor is used to execute the machine-readable instructions stored in the memory. When the machine-readable instructions are executed by the processor, the processor performs the steps of the trajectory prediction method according to any one of claims 1 to 10.

13. A computer program product, characterized in that When the computer program product is executed by a computer device, the computer device executes the steps of the trajectory prediction method according to any one of claims 1 to 10.

Images