Trajectory prediction method and apparatus, and storage medium

Abstract

The present application relates to a trajectory prediction method and apparatus, and a storage medium, relating to the technical field of automobiles, and particularly relating to the technical field of automobile trajectory prediction. The method is applied to a target vehicle, and comprises: acquiring motion state parameters of a vehicle to be predicted and first candidate road points; on the basis of the motion state parameters and a motion feasibility constraint condition of said vehicle, screening out second candidate road points from the first candidate road points; and, on the basis of the second candidate road points, generating a motion trajectory of said vehicle, wherein the first candidate road points are determined on the basis of surrounding road information of said vehicle. Thus, the present application can avoid the technical problem of low feasibility of predicted trajectories.

Description

Trajectory prediction methods, devices and storage media

[0001] This application claims priority to Chinese Patent Application No. 202510102803.6, filed on January 22, 2025, entitled “Trajectory Prediction Method, Apparatus and Storage Medium”, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of automotive technology, and more particularly to the field of automotive trajectory prediction technology, specifically to a trajectory prediction method, device, and storage medium. Background Technology

[0003] In recent years, the widespread application of deep neural networks in trajectory prediction has indeed brought significant progress to fields such as autonomous driving and intelligent transportation. Among them, target-driven trajectory prediction methods achieve more accurate predictions of future vehicle trajectories by decomposing the problem into two sub-tasks: predicting possible targets and estimating motion based on contextual features. However, while pursuing low prediction errors, this method often neglects the physical feasibility of the predicted trajectory.

[0004] In related technologies, it is proposed to determine the predicted trajectory of the target vehicle by using the historical trajectory information of the target vehicle and the historical trajectory information of the surrounding vehicles. However, there is a problem that the predicted trajectory is determined only by historical information, resulting in low feasibility of the predicted trajectory.

[0005] In related technologies, a neural network based on an attention mechanism has been proposed to predict the driving trajectory of a vehicle. However, this method has the problem of not being able to update the trajectory cyclically, resulting in low feasibility of trajectory prediction.

[0006] Therefore, more physical constraints and more complex trajectory representation methods need to be introduced into the prediction model to ensure that the generated predicted trajectory is both accurate and feasible. Summary of the Invention

[0007] This application provides a trajectory prediction method, apparatus, and storage medium to at least address the technical problem of low feasibility in trajectory prediction in related technologies. The technical solution of this application is as follows.

[0008] According to a first aspect provided in this application, a trajectory prediction method is provided, the method being applied to a target vehicle, the method comprising:

[0009] Obtain the motion state parameters of the vehicle to be predicted and the first candidate road point; based on the motion state parameters and the motion feasibility constraints of the vehicle to be predicted, select the second candidate road point from the first candidate road point; based on the second candidate road point, generate the motion trajectory of the vehicle to be predicted.

[0010] The first candidate road point is determined based on the surrounding road information of the vehicle to be predicted.

[0011] It should be understood that physical feasibility is crucial in vehicle trajectory prediction because it determines whether the prediction results can be practically applied to autonomous driving systems. If the predicted trajectory violates physical feasibility, even if its average displacement error is small, it may cause autonomous vehicles to encounter unfollowable trajectories in actual driving, thereby posing safety hazards.

[0012] Based on the above technical solution, by comprehensively considering the vehicle's motion state parameters and motion feasibility constraints, this invention can screen out candidate road points that are more consistent with the actual driving conditions of the vehicle, thereby generating a more accurate driving trajectory.

[0013] Meanwhile, this method also considers various factors such as motion feasibility constraints (road conditions, traffic rules), further improving the reliability and practicality of trajectory prediction.

[0014] In some embodiments, motion feasibility constraints are determined based on road driving rules and / or configuration information of the vehicle to be predicted; motion feasibility constraints include at least one of the following: maximum acceleration constraint of the vehicle to be predicted, maximum deceleration constraint of the vehicle to be predicted, maximum steering constraint of the vehicle to be predicted, and road speed limit constraint.

[0015] Based on the above technical solution, this method comprehensively considers the above constraints to ensure that the vehicle to be predicted can not only perform at its best during the movement, but also comply with road safety regulations, thereby achieving safe, efficient and feasible vehicle movement.

[0016] In some embodiments, obtaining a first candidate road point includes: determining a first road segment at a preset distance along the direction of movement of the vehicle to be predicted based on the location information of the vehicle to be predicted and surrounding road information; discretizing the centerline of the first road segment to obtain a third candidate road point; determining a first search area based on the current speed, heading angle, and road factor of the first road segment of the vehicle to be predicted; the road factor is determined based on the road segment type of the first road segment; the orientation of the centerline of the first search area is consistent with the heading angle; the search range of the first search area is positively correlated with the current speed; and determining the first candidate road point from the third candidate road point based on the first search area.

[0017] Based on the above technical solution, this method achieves dynamic and accurate determination of the first search area by comprehensively considering the current speed, heading angle and road factors under different scenarios of the vehicle to be predicted, which helps to improve the safety of vehicle driving and the accuracy of prediction.

[0018] In some embodiments, a second candidate road point is selected from the first candidate road points based on motion state parameters and motion feasibility constraints of the vehicle to be predicted. This includes: removing candidate points located within the near-end deceleration range and candidate points located outside the far-end acceleration range from the first candidate road points to obtain the second candidate road points; wherein the near-end deceleration range is determined based on the current speed and maximum deceleration constraints of the vehicle to be predicted; and the far-end acceleration range is determined based on the current speed, maximum acceleration constraints, and road speed limit constraints of the vehicle to be predicted.

[0019] Based on the above technical solution, this method, through the aforementioned screening mechanism, can effectively eliminate candidate road points that are theoretically impossible for the vehicle to reach or practically impossible to reach within the prediction time range, thereby ensuring the accuracy and practicality of the prediction results and further optimizing the accuracy and reliability of vehicle trajectory prediction.

[0020] In some embodiments, generating a motion trajectory of a vehicle to be predicted based on a second candidate road point includes: determining a target road point from the second candidate road points; generating a motion trajectory of the vehicle to be predicted based on the target road point; the motion trajectory is the trajectory between the current position of the vehicle to be predicted and the target road point.

[0021] Based on the above technical solution, this method generates a motion trajectory based on a target road point from the second candidate road points. The embodiments of this application can more accurately predict the future driving path of the vehicle, reduce prediction errors, and further improve the reliability and practicality of trajectory prediction.

[0022] In some embodiments, determining the target road point from the second candidate road points includes: inputting the location information of the vehicle to be predicted, motion state parameters, and location information of the candidate road points into a graph attention network model to obtain the target road point with the highest attention coefficient among the second candidate road points; the attention coefficient is used to characterize the confidence of the vehicle to be predicted in the candidate road points.

[0023] Based on the above technical solution, this method introduces a graph attention network model, enabling the embodiments of this application to more accurately capture the relationship between vehicles and road points, as well as the potential impact of different road points on the vehicle's driving path. Furthermore, the attention mechanism allows the model to adaptively focus on candidate road points that have the greatest impact on the vehicle's driving path, thereby improving the accuracy and robustness of trajectory prediction.

[0024] In some embodiments, generating a motion trajectory of a vehicle to be predicted based on a target road point includes: determining a reference path for the vehicle to be predicted based on the target road point; the reference path is a predicted path between the lane segment where the vehicle to be predicted is located and the lane segment where the target road point is located; determining multiple driving coordinates of the vehicle to be predicted based on the reference path and the location information of the vehicle to be predicted; generating a motion trajectory of the vehicle to be predicted based on the multiple driving coordinates; the motion trajectory is the trajectory between the current position of the vehicle to be predicted and the target road point.

[0025] Based on the above technical solution, this method comprehensively considers the location information, motion state, and location information of the target road point of the vehicle to be predicted, thereby enabling the embodiments of this application to generate a reference path and motion trajectory that are more consistent with the actual driving situation of the vehicle.

[0026] Furthermore, the determination of driving coordinates and the generation of motion trajectory fully consider the vehicle's dynamic characteristics and road conditions, further improving the accuracy and reliability of trajectory prediction.

[0027] In some embodiments, the reference path includes a plurality of pre-aiming points; the pre-aiming points are used to characterize the expected coordinates on the reference path; based on the reference path and the position information of the vehicle to be predicted, determining a plurality of driving coordinates of the vehicle to be predicted includes: for each of the plurality of pre-aiming points, determining the steering angle required for the vehicle to be predicted to travel to the pre-aiming point based on the motion state parameters of the vehicle to be predicted; determining the acceleration of the vehicle to be predicted based on the position information of the vehicle to be predicted and the target road point; and determining the driving coordinates of the vehicle to be predicted based on the acceleration and the steering angle, so as to obtain a plurality of driving coordinates of the vehicle to be predicted.

[0028] Based on the above technical solution, this method continuously updates the position of the pre-aiming point and repeatedly executes the pure tracking algorithm calculation and steering angle adjustment process. Through continuous iteration and optimization, the vehicle to be predicted can gradually approach and stably maintain its position on the reference path, thereby generating the driving coordinates of multiple vehicles to be predicted.

[0029] In some embodiments, determining a reference path for a vehicle to be predicted based on a target road point includes: obtaining a set of lane segments between a target lane segment and the lane segment where the vehicle to be predicted is located; the target lane segment is the lane segment where the target road point is located; and performing equidistant sampling along the lane centerline for each lane segment in the set of lane segments to generate a reference path between the target lane segment and the lane segment where the vehicle to be predicted is located.

[0030] Based on the above technical solution, this method obtains a set of lane segments between the target lane segment and the lane segment where the vehicle to be predicted is located, and performs equidistant sampling along the lane centerline. In this embodiment, a reference path that is more consistent with the actual driving situation of the vehicle can be generated, and the accuracy and reliability of the reference path are significantly improved, providing strong support for the autonomous driving and path planning of the vehicle.

[0031] According to a second aspect provided in this application, a trajectory prediction device is provided, the device comprising: a processing unit and an acquisition unit; the acquisition unit is configured to acquire motion state parameters of a vehicle to be predicted and a first candidate road point; the first candidate road point is determined based on surrounding road information of the vehicle to be predicted; the processing unit is configured to select a second candidate road point from the first candidate road point based on the motion state parameters and motion feasibility constraints of the vehicle to be predicted; the processing unit is further configured to generate a motion trajectory of the vehicle to be predicted based on the second candidate road point.

[0032] According to a third aspect provided in this application, a vehicle is provided, including a trajectory prediction device as described in the second aspect.

[0033] According to a fourth aspect provided in this application, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute instructions to implement the method of the first aspect described above and any possible implementation thereof.

[0034] According to a fifth aspect provided in this application, a computer-readable storage medium is provided that, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the methods described in the first aspect and any possible implementation thereof.

[0035] According to the sixth aspect provided in this application, a computer program product is provided, the computer program product including computer instructions, which, when executed on an electronic device, cause the electronic device to perform the method described in the first aspect and any possible implementation thereof.

[0036] Therefore, the above-mentioned technical features of this application have the following beneficial effects:

[0037] (1) By comprehensively considering the vehicle's motion state parameters and motion feasibility constraints, the present invention can screen out candidate road points that are more in line with the actual driving conditions of the vehicle, thereby generating a more accurate driving trajectory.

[0038] Meanwhile, the embodiments of this application also consider various factors such as motion feasibility constraints (road conditions, traffic rules), which further improves the reliability and practicality of trajectory prediction.

[0039] (2) By comprehensively considering the above constraints, this application embodiment ensures that the vehicle to be predicted can perform at its best while complying with road safety regulations during the movement process, thereby achieving safe, efficient and feasible vehicle movement.

[0040] (3) The embodiments of this application achieve dynamic and accurate determination of the first search area by comprehensively considering the current speed, heading angle and road factors under different scenarios of the vehicle to be predicted, which helps to improve the safety of vehicle driving and the accuracy of prediction.

[0041] (4) Through the above-mentioned screening mechanism, the embodiments of this application can effectively eliminate candidate road points that the vehicle to be predicted cannot reach in theory or are not feasible in actual driving within the prediction time range, thereby ensuring the accuracy and practicality of the prediction results and further optimizing the accuracy and reliability of vehicle trajectory prediction.

[0042] (5) By generating a motion trajectory based on a target road point from the second candidate road points, this embodiment of the application can more accurately predict the future driving path of the vehicle, reduce prediction error, and further improve the reliability and practicality of trajectory prediction.

[0043] (6) By introducing a graph attention network model, this embodiment of the application can more accurately capture the relationship between vehicles and road points, as well as the potential impact of different road points on the vehicle's driving path. The attention mechanism enables the model to adaptively focus on the candidate road points that have the greatest impact on the vehicle's driving path, thereby improving the accuracy and robustness of trajectory prediction.

[0044] (7) This application embodiment comprehensively considers the location information, motion state and target road point location information of the vehicle to be predicted, so that this application embodiment can generate a reference path and motion trajectory that are more in line with the actual driving situation of the vehicle.

[0045] Furthermore, the determination of driving coordinates and the generation of motion trajectory fully consider the vehicle's dynamic characteristics and road conditions, further improving the accuracy and reliability of trajectory prediction.

[0046] (8) In this embodiment of the application, the position of the pre-aiming point is continuously updated, and the above-described pure tracking algorithm calculation and steering angle adjustment process is repeatedly executed. Through continuous iteration and optimization, the vehicle to be predicted can gradually approach and stably maintain its position on the reference path, thereby generating the driving coordinates of multiple vehicles to be predicted.

[0047] (9) This application embodiment obtains the set of lane segments between the target lane segment and the lane segment where the vehicle to be predicted is located, and performs equidistant sampling along the lane centerline. In this way, this application embodiment can generate a reference path that is more in line with the actual driving situation of the vehicle, and the accuracy and reliability of the reference path are significantly improved, providing strong support for the autonomous driving and path planning of the vehicle.

[0048] It should be noted that the technical effects of any of the implementation methods in aspects two through six can be found in the technical effects of the corresponding implementation methods in aspect one, and will not be repeated here.

[0049] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0050] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application, and do not constitute an undue limitation of this application.

[0051] Figure 1 is a schematic diagram of the structure of a vehicle according to an exemplary embodiment;

[0052] Figure 2 is a block diagram illustrating a trajectory prediction device according to an exemplary embodiment;

[0053] Figure 3 is a flowchart illustrating a trajectory prediction method according to an exemplary embodiment;

[0054] Figure 4 is a scene diagram illustrating a trajectory prediction method according to an exemplary embodiment;

[0055] Figure 5 is a scenario diagram illustrating another trajectory prediction method according to an exemplary embodiment;

[0056] Figure 6 is a geometric diagram of a pre-aiming point according to an exemplary embodiment;

[0057] Figure 7 is a block diagram illustrating another trajectory prediction device according to an exemplary embodiment;

[0058] Figure 8 is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0059] To enable those skilled in the art to better understand the technical solutions of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0060] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0061] In the embodiments of this application, the words "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the words "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a specific manner.

[0062] First, the relevant technologies involved in this application will be explained to facilitate understanding by those skilled in the art.

[0063] In recent years, the widespread application of deep neural networks in trajectory prediction has indeed brought significant progress to fields such as autonomous driving and intelligent transportation. Among these, target-driven trajectory prediction methods, by decomposing the problem into two sub-tasks—predicting potential targets and estimating motion based on contextual features—have achieved more accurate predictions of future vehicle trajectories. However, while pursuing low prediction errors, these methods often neglect the physical feasibility of the predicted trajectory, a problem worthy of in-depth exploration.

[0064] To address the aforementioned issues, a more complex trajectory representation method can be employed to more comprehensively describe the vehicle's spatial occupancy, thereby generating a predicted trajectory and overcoming the problem of low feasibility. This method can more accurately reflect the actual space occupied by the vehicle, thus improving the physical feasibility of the prediction results.

[0065] In summary, while deep learning-based trajectory prediction methods have achieved significant results in reducing prediction errors, physical feasibility remains a key factor limiting their practical application. In the future, researchers need to incorporate more physical constraints and more complex trajectory representation methods into prediction models to ensure that the generated predicted trajectories are both accurate and feasible. This will provide a more reliable guarantee for the practical application of intelligent transportation systems such as autonomous driving.

[0066] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0067] The trajectory prediction method provided in this application can be applied to vehicles. Figure 1 shows a schematic diagram of a vehicle structure provided in this application embodiment. The vehicle 100 may include a chassis 110, a body 120, and wheels 130. It is understood that the vehicle 100 can be a gasoline-powered vehicle, an electric vehicle, a hybrid vehicle, a natural gas vehicle, a methanol vehicle, a solar-powered vehicle, etc.

[0068] For example, vehicle 100 can be a passenger car such as a sedan, sport utility vehicle (SUV), or multi-purpose vehicle (MPV), or a bus, truck, or semi-trailer. This application does not impose specific limitations in this regard.

[0069] It is understood that the above-mentioned components are only examples of some components of vehicle 100, and are not a limitation on the specific structure of vehicle 100.

[0070] Optionally, for vehicle control, vehicle 100 may also include a trajectory prediction system 140. The trajectory prediction system 140 can predict the movement trajectory of vehicles around vehicle 100.

[0071] As shown in Figure 2, this application embodiment provides a block diagram of a trajectory prediction system 140. The trajectory prediction system 140 includes a filtering module 210 and a trajectory prediction module 220.

[0072] The filtering module 210 is used to filter out a second candidate road point from the first candidate road point based on the feasible constraints of the motion of the vehicle to be predicted and the obtained motion state parameters of the vehicle to be predicted, and then send the second candidate road point to the trajectory prediction module 220.

[0073] The trajectory prediction module 220 is used to generate the motion trajectory of the vehicle to be predicted based on the second candidate road points.

[0074] It should be noted that the trajectory prediction system described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. Those skilled in the art will understand that with the evolution of electronic devices and the emergence of other electronic devices, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems. The methods in the following embodiments can all be implemented in the trajectory prediction system with the above-described hardware structure.

[0075] The methods described in the following embodiments can all be implemented in a trajectory prediction system with the above-described hardware structure.

[0076] The vehicle steering control method provided in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0077] The trajectory prediction method of this application embodiment can be applied to predict the motion trajectory of a vehicle to be predicted. As shown in Figure 3, the trajectory prediction method may include steps 301-303. Step 301 can also be referred to as the "acquiring motion state parameters of the vehicle to be predicted and the first candidate road point" process, and steps 302-303 can be referred to as the "generating the motion trajectory of the vehicle to be predicted" process. Steps 301-303 are described in detail below.

[0078] Step 301: Obtain the motion state parameters of the vehicle to be predicted and the first candidate road point.

[0079] In the embodiments related to this application, the first candidate road point is determined based on the surrounding road information of the vehicle to be predicted; the motion state parameters include, but are not limited to, the current speed, acceleration, position, heading angle, and other information of the vehicle to be predicted.

[0080] For example, the trajectory prediction system can obtain the motion state parameters of all vehicles to be predicted in the scene through the onboard perception sensors and positioning system of the target vehicle. The system obtains road information around the vehicle to be predicted through a high-precision map. This road information includes, but is not limited to, lane location and lane adjacency. The system then uses this road information to determine the first candidate road point, i.e., the initial candidate road point.

[0081] Where i = 0, 1, 2, ... represents the vehicle number, i = 0 represents the target vehicle, i = 1, 2, 3, ... represents the vehicles to be predicted around the target vehicle, t = -T, -T+1, ..., 0 represents the time step, and t = 0 is the current time. This represents the location information of the i-th vehicle to be predicted at time t. This represents the speed information of the i-th vehicle to be predicted at time t. This represents the acceleration information of the i-th vehicle to be predicted at time t. This represents the heading angle information of the i-th vehicle to be predicted at time t.

[0082] In some embodiments, the trajectory prediction system can determine a first road segment at a preset distance in the direction of movement of the vehicle to be predicted based on the location information of the vehicle to be predicted and the surrounding road information, discretize the centerline of the first road segment to obtain a third candidate road point, and then determine a first search area based on the current speed, heading angle and road factor of the first road segment of the vehicle to be predicted, and determine the first candidate road point from the third candidate road point based on the first search area.

[0083] Among them, the road factor is determined based on the road segment type of the first road segment and is used to characterize the degree of freedom of the vehicle to be predicted in the scene; the centerline orientation of the first search area is consistent with the heading angle; the search range of the first search area is positively correlated with the current speed.

[0084] For example, the trajectory prediction system performs uniform discretization sampling along the centerline of the first road segment at a sampling interval of 3M to obtain the third candidate road point. The first search region R can be obtained using the following formula 1. i Using the current location of the vehicle to be predicted as the center, filter vehicles located in the first search area R. i Candidate road points within the specified range are selected as the first candidate road points. For example, the first candidate road points are shown in Figure 4.

[0085] in, R0 represents the current speed of the vehicle to be predicted, R0 represents the threshold of the range (to prevent the search range from being too small due to low vehicle speed), and α represents the road factor (the value varies depending on the scene type, with a larger value for non-intersection scenes and a smaller value for intersection scenes). For example, in a non-intersection scene, the road factor can be assigned a relatively large value to reflect the relative freedom and larger safety space of vehicle movement in a non-intersection scene; in an intersection scene, the road factor can be assigned a relatively small value to reflect the restrictions and higher safety requirements of vehicle movement in an intersection scene.

[0086] Furthermore, the first search area can be a semi-circular search area, and the orientation of the angle bisector of the first search area is related to the current heading angle of the vehicle to be predicted. They are in the same direction.

[0087] It should be understood that the embodiments of this application, by comprehensively considering vehicle speed and road factors under different scenarios, achieve dynamic and accurate determination of the first search area, which helps to improve the safety of the predicted vehicle's driving and the accuracy of the prediction.

[0088] Step 302: Based on the motion state parameters and the motion feasibility constraints of the vehicle to be predicted, select the second candidate road point from the first candidate road point.

[0089] In the embodiments related to this application, the motion feasibility constraints are determined based on road driving rules and / or the configuration information of the vehicle to be predicted; the motion feasibility constraints include, but are not limited to, the maximum acceleration constraint of the vehicle to be predicted, the maximum deceleration constraint of the vehicle to be predicted, and the road speed limit constraint.

[0090] In some embodiments, the trajectory prediction system can filter out candidate points located within the near-end deceleration range and candidate points located outside the far-end acceleration range from the first candidate road points to obtain a second candidate road point.

[0091] The near-end deceleration range is determined based on the current speed and maximum deceleration constraints of the vehicle to be predicted; the far-end acceleration range is determined based on the current speed, maximum acceleration constraints, and road speed limit constraints of the vehicle to be predicted.

[0092] For example, the trajectory prediction system can determine the distance the vehicle to be predicted will decelerate within a preset time period based on the current speed and maximum deceleration of the vehicle to be predicted, and then use the deceleration distance as the near-end deceleration range, and filter out candidate points located within the near-end deceleration range from the first candidate road points.

[0093] The trajectory prediction system can also determine the distance the vehicle will accelerate within a preset time period based on its current speed, maximum acceleration, and road speed limit constraints. This accelerated distance is then used as the far-end acceleration range, and candidate points located within this range are eliminated from the first candidate road points. The trajectory prediction system then uses the remaining candidate points after eliminating the near-end deceleration range and the far-end acceleration range as the second candidate road points. For example, the second candidate road points are shown in Figure 4.

[0094] It should be understood that if the maximum acceleration of the vehicle to be predicted reaches the road speed limit constraint within a preset time period, the vehicle to be predicted will travel at the road speed limit constraint. Through the above-described screening mechanism, the embodiments of this application can effectively eliminate candidate road points that are theoretically impossible for the vehicle to reach or practically infeasible within the prediction time range, thereby ensuring the accuracy and practicality of the prediction results and further optimizing the accuracy and reliability of vehicle trajectory prediction.

[0095] Step 303: Based on the second candidate road points, generate the motion trajectory of the vehicle to be predicted.

[0096] In this embodiment of the application, the trajectory prediction system can determine the target road point from the second candidate road points and generate the motion trajectory of the vehicle to be predicted based on the target road point.

[0097] The motion trajectory is the path between the current position of the vehicle to be predicted and the target road point.

[0098] In some embodiments, the trajectory prediction system inputs the location information of the vehicle to be predicted, motion state parameters, and location information of candidate road points into a graph attention network model to obtain the target road point with the highest attention coefficient among the second candidate road points.

[0099] The attention coefficient is used to characterize the confidence of the vehicle to be predicted in the candidate road points.

[0100] For example, as shown in Figure 4, the trajectory prediction system can encode the motion state parameters and position information of the vehicle to be predicted by the Transformer module of the cross-attention mechanism to obtain the vehicle features of the vehicle to be predicted. The vehicle features include, but are not limited to, the current speed, acceleration, position, heading angle and other information of the vehicle to be predicted.

[0101] Furthermore, the trajectory prediction system can determine the positional features of the second candidate road point and the relative position vector r of the vehicle to be predicted. i,j This means obtaining the distance between the location of the second candidate road point and the location of the vehicle to be predicted. For example, in, p represents the current location information of the i-th vehicle to be predicted. j This represents the location information of the j-th second candidate road point.

[0102] The trajectory prediction system uses a multilayer perceptron to obtain the location information p of the j-th second candidate road point. j And the relative position vector r between the second candidate point and the vehicle to be predicted. i,j Encode the location features F of the j-th second candidate point. j The relative positional characteristics of the vehicle to be predicted.

[0103] F j =MLP(pj ) E i,j =MLP(r i,j )

[0104] Among them, Transformer(·) is the Transformer encoding module that encodes vehicle timing information. Let r be the motion state parameter of the vehicle to be predicted. i,j Let E be the relative position vector between the surrounding candidate points and the vehicle to be predicted, MLP(·) be a multilayer perceptron, and E be the relative position vector between the candidate points and the vehicle to be predicted. i,j This represents the relative positional characteristics between the second candidate point and the vehicle to be predicted.

[0105] Furthermore, the aforementioned vehicle characteristics F i F is the location feature of the vehicle node attribute and the j-th second candidate point. jE, as a candidate node attribute and relative position feature i,j As a directed edge attribute, a graph structure can be built based on multiple attributes.

[0106] Specifically, in this embodiment, the above graph structure can be constructed using the vehicle to be predicted as the starting node, while introducing multiple second candidate road points as other nodes. Based on this, a directed edge E is defined. i,j Each has a directed edge E i,j Starting from the vehicle to be predicted, point to each second candidate road point.

[0107] Next, to enhance the information representation capability of the graph structure, corresponding attributes can be assigned to the aforementioned nodes and edges. Specifically, the attributes of the vehicle to be predicted are based on its encoded vehicle features F. i The settings are made; while the attributes of the j-th second candidate point are based on its encoded positional features F. j The assignment is performed. For directed edges, the relative position feature E is used. i,j As its attribute.

[0108] To facilitate subsequent graph data processing and analysis, this embodiment also constructs graph data conforming to a specific format. The node feature matrix has a dimension of (N_nodes, feature_dim), where N_nodes equals 1 (representing vehicle nodes) + N (representing the total number of candidate nodes). Regarding edge representation, this embodiment uses an edge list or adjacency matrix. Given the advantages of GAT (Graph Attention Network) in processing graph data, this embodiment prefers to use an edge list to explicitly specify the neighboring nodes of each node. Specifically, in this embodiment, the neighboring nodes of vehicle node i are all candidate nodes j. Furthermore, this embodiment also constructs an edge feature matrix with a dimension of (N_edges, edge_feature_dim), where N_edges equals N because there is a directed edge from the vehicle node to each second candidate node.

[0109] Furthermore, the trajectory prediction system can input the constructed graph data into a graph attention network (GAT) model. The GAT model can calculate the multi-head attention coefficients of the vehicle to be predicted to the second candidate road points based on the graph data (node ​​attributes and edge attributes), sum them to determine the attention coefficients of multiple second candidate points, and take the candidate road point with the highest attention coefficient among the multiple second candidate road points as the target road point.

[0110] It should be understood that the attention coefficient can also be used to characterize the degree of attention paid to the second candidate road point of the vehicle to be predicted. The larger the attention coefficient, the closer and more important the relationship between the vehicle to be predicted and the corresponding candidate point, which helps the model to pay more attention to the candidate road points that have an important impact on the prediction results in subsequent processing.

[0111] In addition, the trajectory prediction system can also calculate the normalized attention coefficient after performing a flexible maximum value calculation on the attention coefficient, which is the confidence score of the second candidate road point, and use the second candidate road point with the highest confidence score as the target road point.

[0112] The confidence score can be used to characterize the degree of trust that the vehicle to be predicted has in each candidate road point.

[0113] For example, the attention coefficient can be obtained using the following formula 2.

[0114] α j =GAT(F i ,F j E i,j ) Formula 2

[0115] Where GAT(·) is a graph attention network, α j This represents the attention coefficient.

[0116] The normalized attention coefficient can also be obtained using the following formula 3.

[0117] α′ j =Softmax(α) j ) Formula 3

[0118] Where Softmax(·) is the softmax function for calculating the maximum flexibility, α′ j This is the normalized attention coefficient.

[0119] In some embodiments, the trajectory prediction system can determine a reference path for the vehicle to be predicted based on a target road point, and then determine multiple driving coordinates of the vehicle to be predicted based on the reference path and the location information of the vehicle to be predicted, and generate the motion trajectory of the vehicle to be predicted based on the multiple driving coordinates.

[0120] The reference path is the expected path between the lane segment where the vehicle to be predicted is located and the lane segment where the target road point is located; the trajectory is the trajectory between the current position of the vehicle to be predicted and the target road point.

[0121] For example, the trajectory prediction system can obtain a set of lane segments between the target lane segment and the lane segment where the vehicle to be predicted is located, and perform equidistant sampling of each lane segment in the set along the lane centerline to generate a reference path between the target lane segment and the lane segment where the vehicle to be predicted is located.

[0122] The target lane segment is the lane segment where the target road point is located.

[0123] In one scenario, as shown in Figure 5, starting from the target lane segment (lane segment A), the search traverses its longitudinally following lane segments and laterally adjacent lane segments until the lane segment where the vehicle to be predicted is located is found. This yields a set of lane segments between the target lane segment and the lane segment where the vehicle to be predicted is located. For example, lane segment A, lane segment B, and lane segment C.

[0124] Furthermore, each lane segment in the lane segment set is sampled at equal intervals along the lane centerline to obtain multiple lane center points, and these multiple lane center points are connected to generate a reference path.

[0125] In some embodiments, the trajectory prediction system determines the driving coordinates of the vehicle to be predicted based on the position information of the pre-aiming point and the pre-aiming point for each of a plurality of pre-aiming points, so as to obtain a plurality of driving coordinates of the vehicle to be predicted.

[0126] The reference path includes multiple aiming points; these aiming points are used to characterize the expected coordinates on the reference path.

[0127] For example, the trajectory prediction system can determine the steering angle required for the vehicle to travel to the target point based on the motion state parameters of the vehicle to be predicted, and then determine the acceleration of the vehicle to be predicted based on the position information of the vehicle to be predicted and the target road point, and determine the driving coordinates of the vehicle to be predicted based on the acceleration and steering angle, so as to obtain multiple driving coordinates of the vehicle to be predicted.

[0128] In one scenario, the trajectory prediction system can determine one or more aiming points as short-term targets based on the position information of the vehicle to be predicted and a reference path. Subsequently, a pure tracking algorithm is used to calculate and determine the steering angle (front wheel angle adjustment) of the vehicle to be predicted. Based on this steering angle, the system controls the vehicle to track the aiming points on the reference path and updates the current motion state parameters (such as speed, acceleration, etc.) and path geometry features of the vehicle to be predicted. Then, the calculated front wheel angle command is sent to the vehicle steering system to execute the steering operation. After the vehicle performs the steering, its state (position, speed, direction, etc.) is updated accordingly. This process is repeated continuously within a preset prediction time. Each iteration redetermines the aiming points and calculates the front wheel angle based on the latest vehicle state and reference path information to achieve continuous and accurate tracking of the reference path, and finally generates a smooth vehicle motion trajectory that meets the path requirements.

[0129] It should be understood that the aiming point is usually located a certain distance in front of the vehicle to be predicted. Its selection needs to take into account factors such as the vehicle's speed, path curvature, and system response time to ensure that the vehicle can adjust its driving direction in a timely and smooth manner.

[0130] Specifically, the trajectory prediction system uses the position of the vehicle to be predicted as the starting point and advances along the reference path at a preset forward look-ahead distance to obtain a pre-aiming point. The preset forward look-ahead distance... It can be determined using the following formula 4.

[0131] in, This represents the speed of the i-th vehicle to be predicted at time t; L represents the acceleration of the i-th vehicle to be predicted at time t; α and β represent constant coefficients, which can be adjusted according to vehicle type and road conditions; fc This represents the distance threshold, which is used to avoid the problem of insufficient forward sight distance for the vehicle being predicted when it is traveling at low speeds.

[0132] Furthermore, the trajectory prediction system calculates and determines the steering angle based on the position information of the vehicle to be predicted and the position information of the preview point. Then, based on this steering angle, it controls the vehicle to track the preview point on the reference path, thereby updating the pose of the vehicle to be predicted. The steering angle δ... f (Lateral control quantity) can be determined by the following formula 5.

[0133] in,

[0134] The geometric relationship between the aiming point and the vehicle to be predicted is shown in Figure 6, where L represents the wheelbase of the vehicle to be predicted, α is the angle between the vehicle to be predicted and the aiming point, and L... f δ represents the distance between the vehicle to be predicted and the aiming point. f Indicates the steering angle, R represents the radius; α t This represents the angle between the vehicle to be predicted and the aiming point at time t. δ represents the distance between the vehicle to be predicted at time t and the target point, δ0 represents the preset maximum steering angle, and -δ0 represents the preset minimum steering angle;

[0135] It should be understood that if the calculated steering angle is greater than the preset maximum steering angle, the maximum steering constraint in the motion feasibility constraint condition cannot be met, i.e., the preset maximum steering angle. Therefore, the preset maximum steering angle is used as the current steering angle of the vehicle to be predicted. If the calculated steering angle is less than the preset minimum steering angle, the preset minimum steering angle is used as the current steering angle of the vehicle to be predicted.

[0136] Furthermore, determine the distance d between the vehicle to be predicted and the target road point. i Based on the prediction duration, the speed and distance d of the vehicle to be predicted i Calculate the acceleration 'a' of the vehicle to be predicted during the prediction period. i,c As another control variable for controlling the movement of the vehicle (longitudinal control variable).

[0137] in, Represents the location information of the i-th vehicle to be predicted, (g i,x ,g i,y () represents the location coordinates of the target road point. The coordinates represent the location information of the i-th vehicle to be predicted. Let T be the speed of the i-th vehicle to be predicted, and T′ be the prediction duration.

[0138] Furthermore, the trajectory prediction system can predict the steering angle of the vehicle to be predicted. acceleration a i,c These lateral and longitudinal control variables are used to determine the vehicle's coordinates at the next moment. In other words, the vehicle's state (position, speed, heading angle) is updated, specifically using formulas 8-11.

[0139] in, This represents the location information of the i-th vehicle to be predicted at time t. Let represent the heading angle of the i-th vehicle to be predicted at time t. Let represent the speed of the i-th vehicle to be predicted at time t, where Δt is the time interval between adjacent time steps. This represents the position information of the i-th vehicle to be predicted at time t+1, i.e., the driving coordinates of the vehicle to be predicted at the next time step. This represents the heading angle of the i-th vehicle to be predicted at time t+1. Let represent the speed of the i-th vehicle to be predicted at time t+1.

[0140] After updating the state of the vehicle to be predicted at the current time t, the aiming point is obtained by repeatedly aiming forward along the reference path at time t+1. By tracking the aiming point through the steering angle and acceleration, the position coordinates of the vehicle to be predicted at the next time are obtained. Finally, a motion trajectory with a controllable rate of curvature change at each point is generated, as shown in Figure 5. The black solid line represents the generated motion trajectory within the future prediction time.

[0141] It should be understood that physical feasibility is crucial in vehicle trajectory prediction because it determines whether the prediction results can be practically applied to autonomous driving systems. If the predicted trajectory violates physical feasibility, even if its average displacement error is small, it may cause the autonomous vehicle to encounter an unfollowable trajectory during actual driving, thereby leading to safety hazards. The trajectory prediction system provided in this application comprehensively considers the vehicle's motion state parameters and motion feasibility constraints. This invention can filter out candidate road points that better match the actual driving conditions of the vehicle, thereby generating a more accurate driving trajectory.

[0142] Meanwhile, the embodiments of this application also consider various factors such as motion feasibility constraints (road conditions, traffic rules), which further improves the reliability and practicality of trajectory prediction.

[0143] The foregoing mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the image acquisition device or electronic device includes corresponding hardware structures and / or software modules for performing each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0144] This application embodiment can, according to the above method, exemplarily divide an image acquisition device or electronic device into functional modules. For example, the image acquisition device or electronic device may include functional modules corresponding to each functional division, or two or more functions may be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in this application embodiment is illustrative and only represents one logical functional division; in actual implementation, there may be other division methods.

[0145] Figure 7 is a block diagram illustrating another trajectory prediction device according to an exemplary embodiment. Referring to Figure 7, the trajectory prediction device includes a processing unit 701 and an acquisition unit 702.

[0146] The acquisition unit 702 is used to acquire the motion state parameters of the vehicle to be predicted and the first candidate road point; the first candidate road point is determined based on the surrounding road information of the vehicle to be predicted; the processing unit 701 is used to filter out the second candidate road point from the first candidate road point based on the motion state parameters and the motion feasibility constraints of the vehicle to be predicted; the processing unit 701 is also used to generate the motion trajectory of the vehicle to be predicted based on the second candidate road point.

[0147] In one possible implementation, the motion feasibility constraints are determined based on road driving rules and / or configuration information of the vehicle to be predicted; the motion feasibility constraints include at least one of the following: maximum acceleration constraint of the vehicle to be predicted, maximum deceleration constraint of the vehicle to be predicted, maximum steering constraint of the vehicle to be predicted, and road speed limit constraint.

[0148] In one possible implementation, the processing unit 701 is specifically configured to: determine a first road segment at a preset distance along the direction of movement of the vehicle to be predicted, based on the location information of the vehicle to be predicted and the surrounding road information; discretize and sample the centerline of the first road segment to obtain a third candidate road point; determine a first search area based on the current speed, heading angle, and road factor of the first road segment of the vehicle to be predicted; the road factor is determined based on the road segment type of the first road segment; the orientation of the centerline of the first search area is consistent with the heading angle; the search range of the first search area is positively correlated with the current speed; and determine a first candidate road point from the third candidate road point based on the first search area.

[0149] In one possible implementation, the processing unit 701 is specifically used to filter out candidate points located within the near-end deceleration range and candidate points located outside the far-end acceleration range from the first candidate road points to obtain a second candidate road point; wherein, the near-end deceleration range is determined based on the current speed and maximum deceleration constraint of the vehicle to be predicted; and the far-end acceleration range is determined based on the current speed, maximum acceleration constraint and road speed limit constraint of the vehicle to be predicted.

[0150] In one possible implementation, the processing unit 701 is specifically used to determine the target road point from the second candidate road points; generate the motion trajectory of the vehicle to be predicted based on the target road point; the motion trajectory is the trajectory between the current position of the vehicle to be predicted and the target road point.

[0151] In one possible implementation, the processing unit 701 is specifically used to input the location information of the vehicle to be predicted, motion state parameters, and location information of candidate road points into the graph attention network model to obtain the target road point with the highest attention coefficient among the second candidate road points; the attention coefficient is used to characterize the confidence of the vehicle to be predicted in the candidate road points.

[0152] In one possible implementation, the processing unit 701 is specifically used to determine a reference path for the vehicle to be predicted based on the target road point; the reference path is the expected path between the lane segment where the vehicle to be predicted is located and the lane segment where the target road point is located; based on the reference path and the location information of the vehicle to be predicted, determine multiple driving coordinates of the vehicle to be predicted; and generate the motion trajectory of the vehicle to be predicted based on the multiple driving coordinates; the motion trajectory is the trajectory between the current position of the vehicle to be predicted and the target road point.

[0153] In one possible implementation, the reference path includes multiple aiming points; the aiming points are used to characterize the expected coordinates on the reference path; the processing unit 701 is specifically used to: for each of the multiple aiming points, determine the steering angle required for the vehicle to travel to the aiming point based on the motion state parameters of the vehicle to be predicted; determine the acceleration of the vehicle to be predicted based on the position information of the vehicle to be predicted and the target road point; and determine the driving coordinates of the vehicle to be predicted based on the acceleration and the steering angle, so as to obtain multiple driving coordinates of the vehicle to be predicted.

[0154] In one possible implementation, the processing unit 701 is specifically used to: obtain a set of lane segments between the target lane segment and the lane segment where the vehicle to be predicted is located; the target lane segment is the lane segment where the target road point is located; and perform equidistant sampling along the lane centerline for each lane segment in the set of lane segments to generate a reference path between the target lane segment and the lane segment where the vehicle to be predicted is located.

[0155] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0156] Figure 8 is a block diagram of an electronic device according to an exemplary embodiment. As shown in Figure 8, the electronic device includes, but is not limited to, a processor 801 and a memory 802.

[0157] The memory 802 described above is used to store the executable instructions of the processor 801. It is understood that the processor 801 is configured to execute instructions to implement the image acquisition method in the above embodiments.

[0158] It should be noted that those skilled in the art will understand that the electronic device structure shown in FIG8 does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in FIG8, or combine certain components, or have different component arrangements.

[0159] The processor 801 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 802, and by calling data stored in the memory 802, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. The processor 801 may include one or more processing units. Optionally, the processor 801 may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 801.

[0160] The memory 802 can be used to store software programs and various data. The memory 802 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required by at least one functional module (such as a determination unit, processing unit, etc.), etc. Furthermore, the memory 802 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0161] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory 802 including instructions, which can be executed by a processor 801 of an electronic device to implement the methods in the above embodiments.

[0162] In actual implementation, the functions of the processing unit 701 and the acquisition unit 702 in Figure 7 can both be implemented by the processor 801 in Figure 8 calling the computer program stored in the memory 802. The specific execution process can be found in the description of the method section in the above embodiment, and will not be repeated here.

[0163] Optionally, the computer-readable storage medium may be a non-transitory computer-readable storage medium, such as a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0164] In an exemplary embodiment, this application also provides a computer program product including one or more instructions, which can be executed by a processor 801 of an electronic device to perform the methods described above.

[0165] It should be noted that when one or more instructions in the computer-readable storage medium or computer program product are executed by the processor of an electronic device, they implement the various processes of the above method embodiments and achieve the same technical effect as the above method. To avoid repetition, they will not be described again here.

[0166] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0167] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0168] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the classified units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0169] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0170] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially, or the parts that contribute to the prior art, or part or all of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0171] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A trajectory prediction method, characterized in that, Applied to a target vehicle, the method includes: Obtain the motion state parameters of the vehicle to be predicted and the first candidate road point; the first candidate road point is determined based on the surrounding road information of the vehicle to be predicted; Based on the motion state parameters and the motion feasibility constraints of the vehicle to be predicted, a second candidate road point is selected from the first candidate road point; Based on the second candidate road point, the motion trajectory of the vehicle to be predicted is generated.

2. The method according to claim 1, characterized in that, The motion feasibility constraints are determined based on road driving rules and / or the configuration information of the vehicle to be predicted; the motion feasibility constraints include at least one of the following: The maximum acceleration constraint of the vehicle to be predicted; The maximum deceleration constraint of the vehicle to be predicted; The maximum steering constraint of the vehicle to be predicted; Speed ​​limits on roads.

3. The method according to claim 1 or 2, characterized in that, Obtain the first candidate road point, including: Based on the location information of the vehicle to be predicted and the surrounding road information, a first road segment with a preset distance is determined in the direction of movement of the vehicle to be predicted; Discretize the centerline of the first road segment to obtain the third candidate road point; A first search area is determined based on the current speed and heading angle of the vehicle to be predicted and the road factor of the first road segment; the road factor is determined based on the road segment type of the first road segment; the centerline of the first search area is aligned with the heading angle; the search range of the first search area is positively correlated with the current speed. Based on the first search area, the first candidate road point is determined from the third candidate road point.

4. The method according to claim 2, characterized in that, The step of selecting a second candidate road point from the first candidate road point based on the motion state parameters and the motion feasibility constraints of the vehicle to be predicted includes: The candidate road points are filtered out from the first candidate road points, those located within the near-end deceleration range and those located outside the far-end acceleration range, to obtain the second candidate road points; The near-end deceleration range is determined based on the current speed of the vehicle to be predicted and the maximum deceleration constraint; the far-end acceleration range is determined based on the current speed of the vehicle to be predicted, the maximum acceleration constraint, and the road speed limit constraint.

5. The method according to any one of claims 1 to 4, characterized in that, The step of generating the motion trajectory of the vehicle to be predicted based on the second candidate road points includes: Determine the target road point from the second candidate road points; The motion trajectory of the vehicle to be predicted is generated based on the target road point; the motion trajectory is the trajectory between the current position of the vehicle to be predicted and the target road point.

6. The method according to claim 5, characterized in that, Determining the target road point from the second candidate road points includes: The location information of the vehicle to be predicted, the motion state parameters, and the location information of the candidate road points are input into the graph attention network model to obtain the target road point with the highest attention coefficient among the second candidate road points; the attention coefficient is used to characterize the confidence of the vehicle to be predicted in the candidate road points.

7. The method according to claim 5 or 6, characterized in that, The step of generating the trajectory of the vehicle to be predicted based on the target road point includes: Based on the target road point, a reference path for the vehicle to be predicted is determined; the reference path is the expected path between the lane segment where the vehicle to be predicted is located and the lane segment where the target road point is located. Based on the reference path and the location information of the vehicle to be predicted, multiple driving coordinates of the vehicle to be predicted are determined. Based on the multiple driving coordinates, the motion trajectory of the vehicle to be predicted is generated; the motion trajectory is the trajectory between the current position of the vehicle to be predicted and the target road point.

8. The method according to claim 7, characterized in that, The reference path includes multiple aiming points; the aiming points are used to characterize the expected coordinates on the reference path. The step of determining multiple driving coordinates of the vehicle to be predicted based on the reference path and the location information of the vehicle to be predicted includes: For each of the plurality of pre-aiming points, based on the motion state parameters of the vehicle to be predicted, the steering angle required for the vehicle to be predicted to travel to the pre-aiming point is determined; Based on the location information of the vehicle to be predicted and the target road point, the acceleration of the vehicle to be predicted is determined; Based on the acceleration and the steering angle, the driving coordinates of the vehicle to be predicted are determined to obtain multiple driving coordinates of the vehicle to be predicted.

9. The method according to claim 7 or 8, characterized in that, Determining the reference path of the vehicle to be predicted based on the target road point includes: Obtain the set of lane segments between the target lane segment and the lane segment where the vehicle to be predicted is located; the target lane segment is the lane segment where the target road point is located. Each lane segment in the set of lane segments is sampled at equal intervals along the lane centerline to generate the reference path between the target lane segment and the lane segment where the vehicle to be predicted is located.

10. A trajectory prediction device, characterized in that, The device includes: a processing unit and an acquisition unit; The acquisition unit is used to acquire motion state parameters of the vehicle to be predicted and a first candidate road point; the first candidate road point is determined based on the surrounding road information of the vehicle to be predicted. The processing unit is used to filter out a second candidate road point from the first candidate road point based on the motion state parameters and the motion feasibility constraints of the vehicle to be predicted. The processing unit is further configured to generate the motion trajectory of the vehicle to be predicted based on the second candidate road point.

11. A vehicle, characterized in that, The vehicle includes the device as described in claim 12.

12. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the method as described in any one of claims 1 to 9.

13. A computer-readable storage medium, characterized in that, When the computer-executable instructions stored in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is capable of performing the method as described in any one of claims 1 to 9.

14. A computer program product containing instructions, characterized in that, When the instructions are executed by a computer, the computer performs the method as described in any one of claims 1-9.

Images