Automatic driving track planning method and system based on sampling scoring strategy

Abstract

The invention relates to the technical field of automatic driving track planning, and particularly discloses an automatic driving track planning method and system based on a sampling scoring strategy, and the method comprises the steps: building a track cluster data set through the sampling of a front wheel rotation angle number and a vehicle speed based on a vehicle model; inputting the environment information, the vehicle state information, the vehicle historical track information, the time sequence feature information and the track cluster data as a vehicle model; evaluating each track generated by the vehicle model based on a cross attention mechanism network so as to predict a track suitable for the current vehicle and environment condition; the cross entropy is used as a loss function of the vehicle model to measure the difference between the prediction result and the actual performance. The invention aims to provide an efficient and safe driving track for an automatic driving vehicle, not only can adapt to variable road environments and social traffic rules, but also remarkably improves the safety and user experience of an automatic driving system.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving trajectory planning, and in particular to an autonomous driving trajectory planning method and system based on a sampling scoring strategy. Background Art

[0002] With the advancement of cutting-edge technologies such as sensor technology, computer vision, and machine learning, autonomous driving technology is gradually becoming an important part of the modern transportation system. As one of the core links, the research on trajectory planning algorithms not only needs to consider the physical characteristics and motion constraints of the vehicle itself, but also needs to be able to adapt to changing road conditions and social traffic rules to ensure the safety and comfort of driving in a human-machine co-driving environment. Traditional multi-module trajectory planning algorithms have achieved good results, but there are still problems with complex algorithm structure and poor results in some scenarios. In recent years, emerging technical means such as large models have shown broad application prospects. Using network models to assist in trajectory planning for autonomous driving is of great significance for reducing the incidence of traffic accidents and promoting the transformation of the automotive industry towards intelligence.

[0003] In the field of autonomous driving, trajectory planning algorithms, as one of the core links, must not only consider the physical characteristics and motion constraints of the vehicle itself, but also adapt to changing road conditions and social traffic rules to ensure driving safety and comfort in a human-machine co-driving environment. Traditional multi-module trajectory planning algorithms have achieved these goals to a certain extent and have achieved good results. However, this type of method has the following shortcomings:

[0004] (1) Complex algorithm structure: Traditional methods usually rely on the combination of multiple independent modules, such as perception, prediction, decision-making, and control modules. This not only increases the complexity of the system, but also may lead to error accumulation and coordination problems between modules, affecting the overall performance.

[0005] (2) Poor performance in some scenarios: Faced with complex and ever-changing actual traffic environments, especially in emergency situations or unstructured road conditions, traditional algorithms may exhibit problems such as slow response or incorrect decision-making, making it difficult to achieve generalization in new scenarios.

[0006] In recent years, emerging technologies such as large models have shown broad application prospects, using neural networks to assist in trajectory planning for autonomous driving. However, existing methods based on large models are time-consuming and cannot meet real-time requirements. Small models currently still perform poorly and are prone to problems such as mode collapse.

[0007] In order to solve the shortcomings of the above-mentioned prior art, the present invention proposes an end-to-end trajectory planning algorithm based on sampling scoring strategy. The algorithm aims to achieve effective planning and selection of driving trajectories for autonomous vehicles in complex dynamic environments through an efficient neural network model. Summary of the invention

[0008] To achieve the purpose of the present invention, the present application provides an autonomous driving trajectory planning method based on a sampling scoring strategy, comprising:

[0009] Step S1: Based on the vehicle model, a trajectory cluster dataset is constructed by sampling the front wheel turning angle and vehicle speed;

[0010] Step S2: using environmental information, vehicle state information, vehicle historical trajectory information, time series feature information, and trajectory cluster data as vehicle model input;

[0011] Step S3: Evaluate each trajectory generated by the vehicle model based on the cross-attention mechanism network to predict a trajectory suitable for the current vehicle and environment conditions;

[0012] Step S4: Use cross entropy as the loss function of the vehicle model to measure the gap between the predicted results and the actual performance.

[0013] In some specific embodiments, step S1 includes: the vehicle kinematic model is determined according to the following formula:

[0014]

[0015] Where x(t) and y(t) represent the lateral and longitudinal positions of the vehicle at time t, respectively; ψ(t) represents the heading angle of the vehicle at time t; v represents the speed of the vehicle; β represents the sideslip angle; L represents the wheelbase of the vehicle; θ represents the front wheel turning angle; and Δt represents the time step.

[0016] In some specific embodiments, step S2 includes: the environmental information is based on image data collected by cameras around the vehicle body, including: front left, front right, side left, side right, rear left and rear right, and each camera provides a three-channel RGB image and its corresponding parameter matrix.

[0017] In some specific embodiments, step S2 includes: the vehicle state information includes the current front wheel turning angle, speed and acceleration of the vehicle.

[0018] In some specific embodiments, step S2 includes: the vehicle historical trajectory information is the coordinates of each trajectory point on the optimal path selected from the trajectory cluster data set.

[0019] In some specific embodiments, step S2 includes: the time series feature information is an information carrier connecting the previous and next frames of the vehicle trajectory.

[0020] In some specific embodiments, step S3 includes: the cross attention mechanism is determined according to the following formula:

[0021]

[0022] In the formula, softmax() represents the activation function, Q represents the query output, and K T represents the transpose of the key output, while V represents the value output, d k Represents the scaling parameter.

[0023] In some specific embodiments, step S3 also includes: assigning a score to each trajectory generated by the vehicle kinematic model based on a cross-attention mechanism network, wherein the value of the score is determined based on the safety, comfort, and degree of match with the target path of the trajectory, and storing the result of the current frame of the vehicle trajectory to update and maintain the temporal feature information.

[0024] In some specific embodiments, step S4 includes: the loss function is determined according to the following formula:

[0025] Loss=-∑ k p k ·log(q k )

[0026] In the formula, p k represents the true score of the kth trajectory; q k Represents the score of the trajectory predicted by the vehicle kinematic model.

[0027] To achieve the same invention purpose, the present application also provides an autonomous driving trajectory planning system based on a sampling scoring strategy, including:

[0028] Model building module: used to build a trajectory cluster dataset based on the vehicle model by sampling the front wheel turning angle and vehicle speed;

[0029] Model training module: used to take environmental information, vehicle status information, vehicle historical trajectory information, time series feature information and trajectory cluster data as vehicle model input;

[0030] Result prediction module: used to evaluate each trajectory generated by the vehicle model based on a cross-attention mechanism network, so as to predict a trajectory suitable for the current vehicle and environment conditions;

[0031] Loss determination module: used to use cross entropy as the loss function of the vehicle model to measure the gap between the predicted results and the actual performance.

[0032] Beneficial effects of the above technical solution:

[0033] The present invention proposes an autonomous driving trajectory planning algorithm based on a sampling scoring strategy, which aims to provide an efficient and safe driving trajectory for autonomous driving vehicles. It can not only adapt to the changing road environment and social traffic rules, but also significantly improve the safety and user experience of the autonomous driving system. The present invention selects multiple possible driving trajectories that can meet specific conditions. These trajectories not only fully consider the physical limitations of the vehicle (such as the maximum front wheel turning angle and the maximum vehicle speed), but also combine safety considerations to ensure optimal control stability and safety at different speeds.

[0034] In order to effectively evaluate the quality of each candidate trajectory, the present invention adopts a cross-attention mechanism network. This network structure allows the model to simultaneously focus on the relationship between multiple input sources, and has a certain degree of interpretability and good known performance. In addition, the model comprehensively considers environmental information, vehicle status information, and historical trajectory information to effectively guide the model to learn how to correctly evaluate and select the optimal driving path, thereby ensuring the efficiency and reliability of the entire system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 A flowchart of an autonomous driving trajectory planning method based on a sampling scoring strategy provided by an embodiment of the present invention;

[0037] Figure 2 A schematic diagram of a sampling trajectory cluster data set of an autonomous driving trajectory planning method based on a sampling scoring strategy provided by an embodiment of the present invention;

[0038] Figure 3 A schematic diagram of a cross-attention mechanism network structure of an autonomous driving trajectory planning method based on a sampling scoring strategy provided by an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of the structure of an autonomous driving trajectory planning system based on a sampling scoring strategy is provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0041] Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar symbols throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0042] Embodiment 1

[0043] An embodiment of the present invention provides an automatic driving trajectory planning method based on a sampling scoring strategy, referring to Figure 1 As shown, including:

[0044] Step S1: Based on the vehicle model, a trajectory cluster dataset is constructed by sampling the front wheel turning angle and vehicle speed.

[0045] In a specific embodiment of the present invention, the present invention defines two key variables in the vehicle control space: the front wheel steering angle θ and the vehicle speed v. These two variables directly determine the range of actions that the vehicle can take. Then, according to the wheelbase L of the vehicle, these control variables are mapped to the position change {x, y} in the Cartesian coordinate system, thereby predicting the possible driving trajectory in the future. The vehicle kinematic model is determined according to the following formula:

[0046]

[0047] Where x(t) and y(t) represent the lateral and longitudinal positions of the vehicle at time t, respectively; ψ(t) represents the heading angle of the vehicle at time t; v represents the speed of the vehicle; β represents the sideslip angle; L represents the wheelbase of the vehicle; θ represents the front wheel turning angle; and Δt represents the time step.

[0048] Specifically, the functional relationship between the vehicle speed and the front wheel turning angle is determined according to the vehicle kinematic model, so that when the vehicle speed is higher, the corresponding maximum front wheel turning angle is smaller, and vice versa. This setting can be characterized as follows:

[0049]

[0050] In the formula, θ v，max represents the maximum front wheel turning angle at the current speed, k and m are auxiliary parameters, where m is 1.1, θ max It is the maximum front wheel turning angle limited by the vehicle hardware platform.

[0051] Specifically, dense sampling is performed in the vehicle control space within a predetermined time window of 5s to obtain a series of discrete {angle,v} corresponding points, which are converted into trajectory points {x,y} in the Cartesian coordinate system. In order to cover the needs of different scenarios, a total of 315 different trajectory cluster data are generated, starting from the initial position (0,0) and up to a distance of 120 meters. Figure 2 As shown, the characteristics of various road conditions such as urban areas, ramps, and highways are fully considered. Figure 2 The generated sample trajectory cluster is shown in , where the horizontal axis X represents the longitudinal direction of the vehicle and the vertical axis Y represents the lateral offset. Each trajectory starts from the origin and extends to a different end position, reflecting the possible driving path under different speed and steering combinations. These trajectories provide the basic data set for the subsequent scoring.

[0052] Step S2: taking environmental information, vehicle state information, vehicle historical trajectory information, time series feature information and trajectory cluster data as vehicle model input.

[0053] In a specific embodiment of the present invention, step S2 includes: the environmental information is based on image data collected by cameras around the vehicle body, including: front left, front right, side left, side right, rear left and rear right, and each camera provides a three-channel RGB image and its corresponding parameter matrix.

[0054] In a specific embodiment of the present invention, step S2 includes: the vehicle state information includes the front wheel turning angle, speed and acceleration of the current vehicle.

[0055] In a specific embodiment of the present invention, step S2 includes: the vehicle historical trajectory information is the coordinates of each trajectory point on the optimal path selected from the trajectory cluster data set. Recording the best trajectory selected at the last moment, that is, the {x, y} coordinates of each trajectory point on the optimal path selected from the trajectory cluster, helps to maintain continuity and consistency.

[0056] In a specific embodiment of the present invention, step S2 includes: the time series feature information is an information carrier connecting the previous and next frames of the vehicle trajectory, including a partial result of the output of the previous frame of the vehicle trajectory, which is used to transmit the time series feature and enhance the memory ability of the model.

[0057] Step S3: Evaluate each trajectory generated by the vehicle model based on the cross-attention mechanism network to predict a trajectory suitable for the current vehicle and environment conditions.

[0058] In a specific embodiment of the present invention, the network structure of the cross attention mechanism network is as follows Figure 3 As shown, the cross attention mechanism is determined according to the following formula:

[0059]

[0060] In the formula, softmax() represents the activation function, Q represents the query output, and K T represents the transpose of the key output, while V represents the value output, d k Represents the scaling parameter.

[0061] In a specific embodiment of the present invention, step S3 also includes: assigning a score to each trajectory generated by the vehicle kinematic model based on a cross-attention mechanism network, wherein the value of the score is determined according to the safety, comfort and degree of matching with the target path of the trajectory, and storing the result of the current frame of the vehicle trajectory to update and maintain the temporal feature information.

[0062] Specifically, the generated trajectory cluster is regarded as a query, and the environmental information processed by the BEV backbone is concatenated with other relevant information as the key and value. This architectural design enables the model to accurately evaluate the pros and cons of each trajectory from a global perspective, thereby selecting one or several trajectories that best suit the current vehicle and environment.

[0063] Step S4: Use cross entropy as the loss function of the vehicle model to measure the gap between the predicted results and the actual performance.

[0064] In a specific embodiment of the present invention, step S4 includes: the loss function is determined according to the following formula:

[0065] Loss=-∑ k p k ·log(q k )

[0066] In the formula, p k represents the true score of the kth trajectory; q k Represents the trajectory score predicted by the vehicle kinematic model. By minimizing the value of the cross entropy loss, this application can effectively guide the model to learn how to correctly evaluate and select the optimal driving path.

[0067] The present invention proposes an autonomous driving trajectory planning algorithm based on a sampling scoring strategy, which aims to provide an efficient and safe driving trajectory for autonomous driving vehicles. It can not only adapt to the changing road environment and social traffic rules, but also significantly improve the safety and user experience of the autonomous driving system. The present invention selects multiple possible driving trajectories that can meet specific conditions. These trajectories not only fully consider the physical limitations of the vehicle (such as the maximum front wheel turning angle and the maximum vehicle speed), but also combine safety considerations to ensure optimal control stability and safety at different speeds.

[0068] In order to effectively evaluate the quality of each candidate trajectory, the present invention adopts a cross-attention mechanism network. This network structure allows the model to simultaneously focus on the relationship between multiple input sources, and has a certain degree of interpretability and good known performance. In addition, the model comprehensively considers environmental information, vehicle status information, and historical trajectory information to effectively guide the model to learn how to correctly evaluate and select the optimal driving path, thereby ensuring the efficiency and reliability of the entire system.

[0069] Embodiment 2

[0070] An embodiment of the present invention provides an autonomous driving trajectory planning system based on a sampling scoring strategy, referring to Figure 4 As shown, including:

[0071] Model building module 10: used to build a trajectory cluster data set based on the vehicle model by sampling the front wheel turning angle and the vehicle speed;

[0072] Model training module 20: used to take environmental information, vehicle state information, vehicle historical trajectory information, time series feature information and trajectory cluster data as vehicle model input;

[0073] Result prediction module 30: used for evaluating each trajectory generated by the vehicle model based on a cross-attention mechanism network, so as to predict a trajectory suitable for the current vehicle and environment conditions;

[0074] The loss determination module 40 is used to use the cross entropy as the loss function of the vehicle model to measure the gap between the predicted result and the actual performance.

[0075] In a specific embodiment of the present invention, the model building module 10 is used to: The vehicle kinematic model is determined according to the following formula:

[0076]

[0077] Where x(t) and y(t) represent the lateral and longitudinal positions of the vehicle at time t, respectively; ψ(t) represents the heading angle of the vehicle at time t; v represents the speed of the vehicle; β represents the sideslip angle; L represents the wheelbase of the vehicle; θ represents the front wheel turning angle; and Δt represents the time step.

[0078] In a specific embodiment of the present invention, the model training module 20 is used for: the environmental information is based on image data collected by cameras around the vehicle body, including: front left, front right, side left, side right, rear left and rear right, and each camera provides a three-channel RGB image and its corresponding parameter matrix.

[0079] In a specific embodiment of the present invention, the model training module 20 is used to: The vehicle state information includes the front wheel turning angle, speed and acceleration of the current vehicle.

[0080] In a specific embodiment of the present invention, the model training module 20 is used for: the vehicle historical trajectory information is the coordinates of each trajectory point on the optimal path selected from the trajectory cluster data set.

[0081] In a specific embodiment of the present invention, the model training module 20 is used for: the temporal feature information is an information carrier connecting the previous and next frames of the vehicle trajectory.

[0082] In a specific embodiment of the present invention, the result prediction module 30 is used to: the cross attention mechanism is determined according to the following formula:

[0083]

[0084] In the formula, softmax() represents the activation function, Q represents the query output, and K T represents the transpose of the key output, while V represents the value output, d k Represents the scaling parameter.

[0085] In a specific embodiment of the present invention, the result prediction module 30 is also used to: assign a score to each trajectory generated by the vehicle kinematic model based on a cross-attention mechanism network, the value of the score is determined according to the safety, comfort and degree of matching with the target path of the trajectory, and store the result of the current frame of the vehicle trajectory to update and maintain the temporal feature information.

[0086] In a specific embodiment of the present invention, the loss determination module 40 is used to: the loss function is determined according to the following formula:

[0087] Loss=-∑ k p k ·log(q k )

[0088] In the formula, p k represents the true score of the k-th trajectory; q k Represents the score of the trajectory predicted by the vehicle kinematic model.

[0089] The present invention proposes an autonomous driving trajectory planning algorithm based on a sampling scoring strategy, which aims to provide an efficient and safe driving trajectory for autonomous driving vehicles. It can not only adapt to the changing road environment and social traffic rules, but also significantly improve the safety and user experience of the autonomous driving system. The present invention selects multiple possible driving trajectories that can meet specific conditions. These trajectories not only fully consider the physical limitations of the vehicle (such as the maximum front wheel turning angle and the maximum vehicle speed), but also combine safety considerations to ensure optimal control stability and safety at different speeds.

[0090] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0091] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referenced to each other. The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the functions in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing terminal device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1The steps of the functions specified in one or more boxes. Although the preferred embodiments of the embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they know the basic creative concept. Therefore, the attached claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention. Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal device including a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0092] The method and device provided by the present invention are introduced in detail above. Specific examples are used in this article to illustrate the principle and implementation mode of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation mode and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

[0093] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", "one specific embodiment" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An autonomous driving trajectory planning method based on sampling scoring strategy, characterized in that: include: Step S1: Based on the vehicle model, a trajectory cluster dataset is constructed by sampling the front wheel turning angle and vehicle speed; Step S2: using environmental information, vehicle state information, vehicle historical trajectory information, time series feature information, and trajectory cluster data as vehicle model input; Step S3: Evaluate each trajectory generated by the vehicle model based on the cross-attention mechanism network to predict a trajectory suitable for the current vehicle and environment conditions; Step S4: Use cross entropy as the loss function of the vehicle model to measure the gap between the predicted results and the actual performance.

2. The automatic driving trajectory planning method based on sampling scoring strategy according to claim 1 is characterized in that: Step S1 includes: the vehicle kinematic model is determined according to the following formula: Where x(t) and y(t) represent the lateral and longitudinal positions of the vehicle at time t, respectively; ψ(t) represents the heading angle of the vehicle at time t; v represents the speed of the vehicle; β represents the sideslip angle; L represents the wheelbase of the vehicle; θ represents the front wheel turning angle; and Δt represents the time step.

3. The automatic driving trajectory planning method based on sampling scoring strategy according to claim 1 is characterized in that: Step S2 includes: the environmental information is based on image data collected by cameras around the vehicle body, including: front left, front right, side left, side right, rear left and rear right, each camera provides a three-channel RGB image and its corresponding parameter matrix.

4. The automatic driving trajectory planning method based on sampling scoring strategy according to claim 1 is characterized in that: Step S2 includes: the vehicle status information includes the front wheel turning angle, speed and acceleration of the current vehicle.

5. The automatic driving trajectory planning method based on sampling scoring strategy according to claim 1 is characterized in that: Step S2 includes: the vehicle historical trajectory information includes the coordinates of each trajectory point on the optimal path selected from the trajectory cluster data set.

6. The automatic driving trajectory planning method based on sampling scoring strategy according to claim 1 is characterized in that: In step S2, the time series feature information includes an information carrier connecting the previous and next frames of the vehicle trajectory.

7. The automatic driving trajectory planning method based on sampling scoring strategy according to claim 1 is characterized in that: Step S3 includes: the cross attention mechanism is determined according to the following formula: In the formula, softmax() represents the activation function, Q represents the query output, and K T represents the transpose of the key output, while V represents the value output, d k Represents the scaling parameter.

8. The automatic driving trajectory planning method based on sampling scoring strategy according to claim 1 is characterized in that: Step S3 also includes: assigning a score to each trajectory generated by the vehicle kinematic model based on a cross-attention mechanism network, wherein the score is determined according to the safety, comfort, and degree of match with the target path of the trajectory, and storing the result of the current frame of the vehicle trajectory to update and maintain the temporal feature information.

9. The automatic driving trajectory planning method based on sampling scoring strategy according to claim 1 is characterized in that: Step S4 includes: the loss function is determined according to the following formula: Loss=-∑ k p k ·log(q k ) In the formula, p k represents the true score of the k-th trajectory; q k Represents the score of the trajectory predicted by the vehicle kinematic model.

10. An autonomous driving trajectory planning system based on sampling scoring strategy, characterized in that: include: Model building module: used to build a trajectory cluster dataset based on the vehicle model by sampling the front wheel turning angle and vehicle speed; Model training module: used to take environmental information, vehicle status information, vehicle historical trajectory information, time series feature information and trajectory cluster data as vehicle model input; Result prediction module: used to evaluate each trajectory generated by the vehicle model based on a cross-attention mechanism network, so as to predict a trajectory suitable for the current vehicle and environment conditions; Loss determination module: used to use cross entropy as the loss function of the vehicle model to measure the gap between the predicted results and the actual performance.

Images