Autonomous driving interactive simulation method and system based on interactive trajectory prediction model

Abstract

The present invention provides an interactive simulation method and system for autonomous driving based on an interactive trajectory prediction model, comprising: obtaining real traffic data, establishing an interactive trajectory prediction model based on the traffic data; outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result; performing autonomous driving simulation on the optimized trajectory prediction result through a preset open source platform of a road motion planning combinable benchmark, and outputting a visualized simulation result. The present invention solves the problem that the existing autonomous driving simulation has low authenticity and is difficult to ensure the safety of autonomous driving.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to an automatic driving interactive simulation method and system based on an interactive trajectory prediction model. Background Art

[0002] Autonomous driving has become a development trend in the automotive industry, and research on autonomous driving perception, positioning, planning, decision-making and control is progressing rapidly. However, safety accidents occur frequently in the field of autonomous driving, and safety has become the essential problem to be solved in the field of autonomous driving. The industry and society need autonomous driving technology with higher reliability to lay a solid foundation for development. Among them, testing scenarios, enriching and improving testing technology are extremely important steps to improve the safety performance of autonomous driving.

[0003] In order to ensure the safety of autonomous vehicles, the vehicles must be tested and verified for millions of kilometers, so autonomous driving relies more and more on simulation platforms. Generally, the system of autonomous vehicles can be divided into three modules: the first is the perception module, which is equivalent to the human eye and collects the surrounding environment status in real time through sensors; the second is the planning and decision-making module, which is equivalent to the human brain, plans the driving path and converts the planned path into executable throttle, brake, steering and other instructions; the third is the control module, which is equivalent to the human hand and foot, used to control the vehicle to perform throttle, brake, steering and other operations.

[0004] There are many autonomous driving simulation platforms in the existing technology, but the existing platforms usually focus on simulating the driving scene and ignore the authenticity of the vehicle's driving behavior. Specifically, the existing technology is usually limited to building real vehicle models and rendering road scenes, lacking modeling for vehicle trajectory prediction and planning, and thus unable to provide real vehicle interaction scenarios. Summary of the invention

[0005] The present invention provides an autonomous driving interactive simulation method and system based on an interactive trajectory prediction model, which are used to solve the problems of low authenticity of existing autonomous driving simulation and difficulty in ensuring the safety of autonomous driving.

[0006] The present invention provides an autonomous driving interactive simulation method based on an interactive trajectory prediction model, comprising:

[0007] Acquire real traffic data, and establish an interactive trajectory prediction model based on the traffic data;

[0008] Outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result;

[0009] The optimized trajectory prediction results are simulated for autonomous driving through an open source platform with a preset road motion planning combinable benchmark, and a visual simulation result is output.

[0010] According to an automatic driving interactive simulation method based on an interactive trajectory prediction model provided by the present invention, the step of acquiring real traffic data and establishing an interactive trajectory prediction model based on the traffic data specifically includes:

[0011] Determine influencers and responders in the interactive conflict based on the traffic data, and distinguish and mark them;

[0012] Based on the distinction between the labeled influencers and responders, the interaction relationship is predicted through the preset relationship prediction network;

[0013] The interaction relationship prediction is abstracted into two trajectory predictions, and an interaction trajectory prediction model is output.

[0014] According to an automatic driving interaction simulation method based on an interaction trajectory prediction model provided by the present invention, influencers and responders in the interaction conflict are determined based on the traffic data, and distinguished and labeled, specifically including:

[0015] The traffic data includes map information and vehicle information, and the conflict yielding party in the map information and vehicle information is determined as a responder, and the conflicting party that does not need yielding is determined as an influencer;

[0016] Responders and influencers are distinguished and labeled through the preset time-space trajectory interleaving method.

[0017] According to an automatic driving interaction simulation method based on an interactive trajectory prediction model provided by the present invention, the interactive relationship prediction is abstracted into two trajectory predictions, and the interactive trajectory prediction model is output, which specifically includes:

[0018] The influencer trajectory is predicted through a preset marginal prediction model to generate influencer trajectory prediction results;

[0019] Based on the influencer trajectory prediction result, predict the responder's trajectory through a preset conditional prediction model to generate a responder trajectory prediction result;

[0020] Based on the influencer trajectory prediction results and the responder trajectory prediction results, combined screening is performed on multiple agents, and the output interactive trajectory prediction model is trained by using the real trajectory results in the traffic data as the supervised loss function of the prediction output.

[0021] According to an interactive trajectory prediction model provided by the present invention, an interactive simulation method for autonomous driving based on the interactive trajectory prediction model outputs an initial trajectory prediction result based on the interactive trajectory prediction model, optimizes the initial trajectory prediction result through a preset advanced process control algorithm, and outputs an optimized trajectory prediction result, specifically comprising:

[0022] Outputting an initial trajectory prediction result through the interactive trajectory prediction model;

[0023] Based on the initial trajectory prediction result, the vehicle driving trajectory is subjected to state constraints and control constraints through an advanced process control algorithm, and an optimized trajectory prediction result is output.

[0024] According to an interactive trajectory prediction model-based autonomous driving interactive simulation method provided by the present invention, an open source platform of a preset road motion planning combinable benchmark is used to perform autonomous driving simulation on the optimized trajectory prediction result, and a visual simulation result is output, which specifically includes:

[0025] Obtain real-time traffic data, extract road map information and vehicle movement information, and convert them into specific format files;

[0026] The specific format file is parsed through a preset trajectory and map element management model and abstracted into a specific object;

[0027] Input the specific object as raw data into the interactive trajectory prediction model, output the trajectory prediction result, and optimize it through the advanced process control algorithm to obtain the optimized trajectory prediction result;

[0028] The optimized trajectory prediction result is updated as the vehicle behavior in the current period, and the updated trajectory data is input into the trajectory and map element management model as a new initial value for iterative update, and traffic flow simulation is performed within a set time period to generate a traffic simulation scene;

[0029] The trajectory and map element management model visualizes the traffic simulation scene frame by frame in chronological order and sends it to a preset visualization platform for display.

[0030] The present invention also provides an autonomous driving interactive simulation system based on an interactive trajectory prediction model, the system comprising:

[0031] A modeling module, used to obtain real traffic data and establish an interactive trajectory prediction model based on the traffic data;

[0032] A prediction module, configured to output an initial trajectory prediction result based on the interactive trajectory prediction model, optimize the initial trajectory prediction result through a preset advanced process control algorithm, and output an optimized trajectory prediction result;

[0033] The simulation module is used to perform autonomous driving simulation on the optimized trajectory prediction result through an open source platform of a preset road motion planning combinable benchmark, and output a visualized simulation result.

[0034] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, an autonomous driving interactive simulation method based on an interactive trajectory prediction model as described above is implemented.

[0035] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for interactive simulation of autonomous driving based on an interactive trajectory prediction model as described in any one of the above is implemented.

[0036] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned autonomous driving interactive simulation methods based on the interactive trajectory prediction model.

[0037] The present invention provides an interactive simulation method and system for autonomous driving based on an interactive trajectory prediction model. By learning the information in large-scale traffic scene data, the trajectory prediction process utilizes big data prior knowledge, thereby achieving more realistic traffic simulation results than the existing technology. Therefore, the present invention can provide a large amount of simulation data that is close to reality for the research and development of autonomous driving technology, thereby saving data collection costs and improving efficiency. In addition, the present invention can also provide simulation test scenarios for algorithm research and development, avoiding the dangers of developing autonomous driving vehicles and testing them on real roads, and improving the safety of autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 It is one of the flow diagrams of an autonomous driving interactive simulation method based on an interactive trajectory prediction model provided by the present invention;

[0040] Figure 2 This is the second flow chart of an interactive simulation method for autonomous driving based on an interactive trajectory prediction model provided by the present invention;

[0041] Figure 3This is a third flow chart of an interactive simulation method for autonomous driving based on an interactive trajectory prediction model provided by the present invention;

[0042] Figure 4 This is a fourth flow chart of an autonomous driving interactive simulation method based on an interactive trajectory prediction model provided by the present invention;

[0043] Figure 5 This is a fifth flow chart of an interactive simulation method for autonomous driving based on an interactive trajectory prediction model provided by the present invention;

[0044] Figure 6 This is the sixth flow chart of an interactive simulation method for autonomous driving based on an interactive trajectory prediction model provided by the present invention;

[0045] Figure 7 It is a schematic diagram of module connection of an autonomous driving interactive simulation system based on an interactive trajectory prediction model provided by the present invention;

[0046] Figure 8 It is a structural schematic diagram of the electronic device provided by the present invention.

[0047] Reference numerals:

[0048] 110: modeling module; 120: prediction module; 130: simulation module;

[0049] 810: processor; 820: communication interface; 830: memory; 840: communication bus. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Combine the following Figure 1-Figure 6 The present invention describes an autonomous driving interactive simulation method based on an interactive trajectory prediction model, comprising:

[0052] S100, acquiring real traffic data, and establishing an interactive trajectory prediction model based on the traffic data;

[0053] S200, outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result;

[0054] S300, performing autonomous driving simulation on the optimized trajectory prediction result through an open source platform of a preset road motion planning combinable benchmark, and outputting a visualized simulation result.

[0055] The present invention uses a deep neural network to model the interactive trajectory prediction model, and conducts large-scale training on the collected vehicle and pedestrian trajectory data to achieve data-driven interactive prediction trajectory generation with good real-world adaptability and feasibility. Based on the interactive prediction model, the present invention supports the systematic deployment of simulation environments and related algorithms, develops an autonomous driving simulation platform that supports multi-agent traffic scene simulation, and visualizes the simulation results through a visualization module.

[0056] Acquire real traffic data and establish an interactive trajectory prediction model based on the traffic data, specifically including:

[0057] S101, determining influencers and responders in the interactive conflict based on the traffic data, and distinguishing and marking them;

[0058] S102, based on distinguishing the marked influencers and responders, predicting the interactive relationship through a preset relationship prediction network;

[0059] S103: abstract the interactive relationship prediction into two trajectory predictions, and output an interactive trajectory prediction model.

[0060] In the present invention, multiple intersections use vehicle-road collaboration to collect large-scale real traffic data and process the data. The processed data information includes map information and vehicle information. Specifically, map information includes information such as traffic lights, lane lines, virtual and real lines, and vehicle information includes physical information such as vehicle position, speed, and direction. Based on this data, the present invention models the interactive trajectory prediction model, combines the existing cutting-edge neural network model, and studies the generation of interactive prediction trajectories based on data-driven, good real-scene adaptability and feasibility.

[0061] Influencers and responders in the interaction conflict are determined based on the traffic data, and distinguished and labeled, specifically including:

[0062] S1011, the traffic data includes map information and vehicle information, and the conflict yielding party in the map information and vehicle information is determined as a responder, and the conflicting party that does not need yielding is determined as an influencer;

[0063] S1012. Distinguish and mark responders and influencers using a preset time-space trajectory interleaving method.

[0064] The interactive trajectory prediction model of the present invention is constructed based on the M2I network model, and the subjects such as vehicles, roads and pedestrians in the interactive conflict are divided into influencers and responders. The responder is defined as the party that needs to yield in the conflict, and the influencer is defined as the party that does not need to yield. In the data labeling process, the model uses a method based on the interlacing of spatiotemporal trajectories to distinguish and label the vehicles in the existing data set as influencers and responders. Different data types can be distinguished by distinguishing and labeling.

[0065] The interaction relationship prediction is abstracted into two trajectory predictions, and an interaction trajectory prediction model is output, which specifically includes:

[0066] S1031. Predicting the influencer trajectory by using a preset marginal prediction model to generate an influencer trajectory prediction result;

[0067] S1032, based on the influencer trajectory prediction result, predict the responder's trajectory through a preset conditional prediction model to generate a responder trajectory prediction result;

[0068] S1033. Based on the influencer trajectory prediction results and the responder trajectory prediction results, a combined screening is performed on multiple agents, and the actual trajectory results in the traffic data are used as the supervised loss function of the prediction output to train the output interactive trajectory prediction model.

[0069] In the model prediction process, the present invention first uses the relationship prediction network to predict the interaction relationship; then the trajectory prediction problem in the interaction is abstracted into two trajectory predictions. First, the marginal prediction model is used to predict the influencer's trajectory, and on this basis, the conditional prediction model is used to predict the responder's trajectory.

[0070] The interactive prediction model further extends the relationship prediction to the relationship prediction of multiple agents. For multiple agents, pairwise prediction is performed, and the final prediction results are combined and screened based on the joint possibility of the prediction. The above model is trained based on real data collected in advance, and the real trajectory results are used as the supervised loss function of the prediction output. The interactive prediction results after training can realize interactive behaviors such as avoidance.

[0071] Outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result, specifically including:

[0072] S201, outputting an initial trajectory prediction result through the interactive trajectory prediction model;

[0073] S202: Based on the initial trajectory prediction result, the vehicle driving trajectory is subjected to state constraints and control constraints through an advanced process control algorithm, and an optimized trajectory prediction result is output.

[0074] The present invention takes into account the phenomenon that the prediction module itself is prone to error accumulation. Based on the interactive prediction model, the MPC control algorithm (advanced process control algorithm) is introduced to further control and optimize the interactive prediction results. The MPC algorithm is an advanced process control method. Based on the vehicle dynamics model, the present invention models the state space equation of vehicle interactive trajectory tracking, and uses the MPC algorithm to perform state constraints and control constraints on the vehicle's driving trajectory to solve the optimal control result. The output result is the control sequence of the vehicle in a limited number of time steps in the future. Test results show that the MPC control algorithm can effectively constrain the vehicle trajectory and realize interactive behaviors such as tracking and obstacle avoidance.

[0075] The optimized trajectory prediction result is simulated for autonomous driving through an open source platform of a preset road motion planning combinable benchmark, and a visual simulation result is output, specifically including:

[0076] S301, obtaining real-time traffic data, extracting road map information and vehicle movement information, and converting them into a specific format file;

[0077] S302, parsing the specific format file through a preset trajectory and map element management model, and abstracting it into a specific object;

[0078] S303, inputting the specific object as raw data into the interactive trajectory prediction model, outputting the trajectory prediction result, and optimizing it through the advanced process control algorithm to obtain an optimized trajectory prediction result;

[0079] S304, updating the optimized trajectory prediction result as the vehicle behavior in the current period, inputting the updated trajectory data as a new initial value into the trajectory and map element management model for iterative update, performing traffic flow simulation within a set time period, and generating a traffic simulation scene;

[0080] S305: The trajectory and map element management model visualizes the traffic simulation scene frame by frame in chronological order and sends it to a preset visualization platform for display.

[0081] Based on the interactive prediction model, the present invention further develops an autonomous driving simulation platform that supports multi-agent traffic scenario simulation on the basis of the open source platform CommonRoad.

[0082] CommonRoad is an open source platform for combinable benchmarks of road motion planning. The benchmark consists of road scenarios containing planning problems, vehicle dynamics models, vehicle parameters, vehicle ID and other information. The simulation platform is lightweight and scenario generalized. Based on this platform, the present invention further integrates the above-mentioned interactive prediction model as an interactive prediction module in the platform. On this basis, the interactive trajectory prediction in the process of autonomous driving is explored, and the visualization of traffic flow is realized while building a road scene control system.

[0083] The present invention first converts the data format, parses and extracts road map information (lane lines, traffic lights, lane width) and vehicle movement information (vehicle ID, vehicle type, speed, direction, starting position coordinates) from the original data, and converts the above information into the XML file format supported by the simulation platform.

[0084] The simulation platform first uses the trajectory and map element management model to parse the scene information in the above XML format, abstracts the road information and vehicle information into specific objects respectively, sends the specific objects as initial data to the interactive prediction network and optimizes them through the MPC control algorithm, outputs the vehicle's trajectory prediction output in the future time period, uses the output result to update the vehicle behavior in the current time period, and solves the vehicle motion behavior through the vehicle dynamics model planning. The updated trajectory data is used as the new initial value and re-sent to the trajectory and map element management model for iterative update to achieve complete traffic flow simulation within a given time period.

[0085] The traffic simulation scene is visualized frame by frame in chronological order, the trajectory and map element management model are connected with the visualization module, the vehicle is visualized as a rectangle with a given ID, size, direction, and position, and the structured map elements are visualized at the same time. The lane is visualized as two sets of parallel line segments of a given width.

[0086] A simulation platform based on the interactive model has been built, which enables visualization of traffic flow while building a road scene control system, providing platform and method support for path planning and driving safety of autonomous driving vehicles in interactive scenarios.

[0087] The present invention uses a mature deep learning method to learn the information in large-scale traffic scene data, so that the trajectory prediction process uses big data prior knowledge, thereby achieving more realistic traffic simulation results than existing technologies. Therefore, this technology can provide a large amount of close-to-real simulation data for the research and development of autonomous driving technology, saving data collection costs while improving efficiency. In addition, this technology can also provide simulation test scenarios for algorithm research and development, avoiding the dangers of developing autonomous driving cars and testing them on real roads.

[0088] And because this technology is lightweight, it does not require the support of special hardware equipment, is easy to install, and can be directly applied to existing computing devices for simulation, with strong portability.

[0089] refer to Figure 7 The present invention also discloses an autonomous driving interactive simulation system based on an interactive trajectory prediction model, the system comprising:

[0090] A modeling module 110, for acquiring real traffic data and establishing an interactive trajectory prediction model based on the traffic data;

[0091] A prediction module 120, configured to output an initial trajectory prediction result based on the interactive trajectory prediction model, optimize the initial trajectory prediction result through a preset advanced process control algorithm, and output an optimized trajectory prediction result;

[0092] The simulation module 130 is used to perform autonomous driving simulation on the optimized trajectory prediction result through an open source platform of a preset road motion planning combinable benchmark, and output a visualized simulation result.

[0093] A modeling module, which determines influencers and responders in the interactive conflict based on the traffic data, and distinguishes and labels them;

[0094] Based on the distinction between the labeled influencers and responders, the interaction relationship is predicted through the preset relationship prediction network;

[0095] The interaction relationship prediction is abstracted into two trajectory predictions, and an interaction trajectory prediction model is output.

[0096] The traffic data includes map information and vehicle information, and the conflict yielding party in the map information and vehicle information is determined as the responder, and the conflict without yielding is determined as the influencer;

[0097] Responders and influencers are distinguished and labeled through the preset time-space trajectory interleaving method.

[0098] The influencer trajectory is predicted through a preset marginal prediction model to generate influencer trajectory prediction results;

[0099] Based on the influencer trajectory prediction result, predict the responder's trajectory through a preset conditional prediction model to generate a responder trajectory prediction result;

[0100] Based on the influencer trajectory prediction results and the responder trajectory prediction results, combined screening is performed on multiple agents, and the output interactive trajectory prediction model is trained by using the real trajectory results in the traffic data as the supervised loss function of the prediction output.

[0101] A prediction module, which outputs an initial trajectory prediction result through the interactive trajectory prediction model;

[0102] Based on the initial trajectory prediction result, the vehicle driving trajectory is subjected to state constraints and control constraints through an advanced process control algorithm, and an optimized trajectory prediction result is output.

[0103] The simulation module obtains real-time traffic data, extracts road map information and vehicle movement information, and converts them into files in a specific format;

[0104] The specific format file is parsed through a preset trajectory and map element management model and abstracted into a specific object;

[0105] Input the specific object as raw data into the interactive trajectory prediction model, output the trajectory prediction result, and optimize it through the advanced process control algorithm to obtain the optimized trajectory prediction result;

[0106] The optimized trajectory prediction result is updated as the vehicle behavior in the current period, and the updated trajectory data is input into the trajectory and map element management model as a new initial value for iterative update, and traffic flow simulation is performed within a set time period to generate a traffic simulation scene;

[0107] The trajectory and map element management model visualizes the traffic simulation scene frame by frame in chronological order and sends it to a preset visualization platform for display.

[0108] The present invention discloses an interactive simulation system for autonomous driving based on an interactive trajectory prediction model. By learning the information in large-scale traffic scene data, the trajectory prediction process utilizes big data prior knowledge, thereby achieving more realistic traffic simulation results than the existing technology. Therefore, the present invention can provide a large amount of simulation data close to reality for the research and development of autonomous driving technology, saving data collection costs and improving efficiency. In addition, the present invention can also provide simulation test scenarios for algorithm research and development, avoiding the dangers of developing autonomous driving vehicles and testing them on real roads, and improving the safety of autonomous driving.

[0109] Figure 8 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 8As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820 and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute an automatic driving interaction simulation method based on an interaction trajectory prediction model, the method comprising: obtaining real traffic data, and establishing an interaction trajectory prediction model based on the traffic data;

[0110] Outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result;

[0111] The optimized trajectory prediction results are simulated for autonomous driving through an open source platform with a preset road motion planning combinable benchmark, and a visual simulation result is output.

[0112] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0113] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute an automatic driving interactive simulation method based on an interactive trajectory prediction model provided by the above methods, the method comprising: obtaining real traffic data, and establishing an interactive trajectory prediction model based on the traffic data;

[0114] Outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result;

[0115] The optimized trajectory prediction results are simulated for autonomous driving through an open source platform with a preset road motion planning combinable benchmark, and a visual simulation result is output.

[0116] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement an autonomous driving interactive simulation method based on an interactive trajectory prediction model provided by the above methods, the method comprising: obtaining real traffic data, and establishing an interactive trajectory prediction model based on the traffic data;

[0117] Outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result;

[0118] The optimized trajectory prediction results are simulated for autonomous driving through an open source platform with a preset road motion planning combinable benchmark, and a visual simulation result is output.

[0119] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0120] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An autonomous driving interactive simulation method based on an interactive trajectory prediction model, characterized in that: include: Acquire real traffic data, and establish an interactive trajectory prediction model based on the traffic data; Outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result; Performing autonomous driving simulation on the optimized trajectory prediction result through a preset open source platform of road motion planning combinable benchmarks, and outputting a visual simulation result; The step of acquiring real traffic data and establishing an interactive trajectory prediction model based on the traffic data specifically includes: Determine influencers and responders in the interactive conflict based on the traffic data, and distinguish and mark them; Based on the distinction between the labeled influencers and responders, the interaction relationship is predicted through the preset relationship prediction network; Abstracting the interaction relationship prediction into two trajectory predictions, and outputting an interaction trajectory prediction model; The interactive relationship prediction is abstracted into two trajectory predictions, and an interactive trajectory prediction model is output, which specifically includes: The influencer trajectory is predicted through a preset marginal prediction model to generate influencer trajectory prediction results; Based on the influencer trajectory prediction result, predict the responder's trajectory through a preset conditional prediction model to generate a responder trajectory prediction result; Based on the influencer trajectory prediction results and the responder trajectory prediction results, combined screening is performed on multiple agents, and the output interactive trajectory prediction model is trained by using the real trajectory results in the traffic data as the supervised loss function of the prediction output.

2. The autonomous driving interactive simulation method based on the interactive trajectory prediction model according to claim 1, characterized in that: Influencers and responders in the interaction conflict are determined based on the traffic data, and distinguished and labeled, specifically including: The traffic data includes map information and vehicle information, and the conflicting courteous party in the map information and the vehicle information is determined as a responder, and the conflicting party that does not need to be courteous is determined as an influencer; Responders and influencers are distinguished and labeled through the preset time-space trajectory interleaving method.

3. The autonomous driving interactive simulation method based on the interactive trajectory prediction model according to claim 1, characterized in that: Outputting an initial trajectory prediction result based on the interactive trajectory prediction model, optimizing the initial trajectory prediction result through a preset advanced process control algorithm, and outputting an optimized trajectory prediction result, specifically including: Outputting an initial trajectory prediction result through the interactive trajectory prediction model; Based on the initial trajectory prediction result, the vehicle driving trajectory is subjected to state constraints and control constraints through an advanced process control algorithm, and an optimized trajectory prediction result is output.

4. The autonomous driving interactive simulation method based on the interactive trajectory prediction model according to claim 1, characterized in that: The optimized trajectory prediction result is simulated for autonomous driving through an open source platform of a preset road motion planning combinable benchmark, and a visual simulation result is output, specifically including: Obtain real-time traffic data, extract road map information and vehicle movement information, and convert them into specific format files; The specific format file is parsed through a preset trajectory and map element management model and abstracted into a specific object; Input the specific object as raw data into the interactive trajectory prediction model, output the trajectory prediction result, and optimize it through the advanced process control algorithm to obtain the optimized trajectory prediction result; The optimized trajectory prediction result is updated as the vehicle behavior in the current period, and the updated trajectory data is input into the trajectory and map element management model as a new initial value for iterative update, and traffic flow simulation is performed within a set time period to generate a traffic simulation scene; The trajectory and map element management model visualizes the traffic simulation scene frame by frame in chronological order and sends it to a preset visualization platform for display.

5. An autonomous driving interactive simulation system based on an interactive trajectory prediction model, characterized in that: The system comprises: A modeling module, used to obtain real traffic data and establish an interactive trajectory prediction model based on the traffic data; A prediction module, configured to output an initial trajectory prediction result based on the interactive trajectory prediction model, optimize the initial trajectory prediction result through a preset advanced process control algorithm, and output an optimized trajectory prediction result; A simulation module, used to perform autonomous driving simulation on the optimized trajectory prediction result through a preset open source platform of a road motion planning combinable benchmark, and output a visual simulation result; The step of acquiring real traffic data and establishing an interactive trajectory prediction model based on the traffic data specifically includes: Determine influencers and responders in the interactive conflict based on the traffic data, and distinguish and mark them; Based on the distinction between the labeled influencers and responders, the interaction relationship is predicted through the preset relationship prediction network; Abstracting the interaction relationship prediction into two trajectory predictions, and outputting an interaction trajectory prediction model; The interactive relationship prediction is abstracted into two trajectory predictions, and an interactive trajectory prediction model is output, which specifically includes: The influencer trajectory is predicted through a preset marginal prediction model to generate influencer trajectory prediction results; Based on the influencer trajectory prediction result, predict the responder's trajectory through a preset conditional prediction model to generate a responder trajectory prediction result; Based on the influencer trajectory prediction results and the responder trajectory prediction results, combined screening is performed on multiple agents, and the output interactive trajectory prediction model is trained by using the real trajectory results in the traffic data as the supervised loss function of the prediction output.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the autonomous driving interactive simulation method based on the interactive trajectory prediction model as described in any one of claims 1 to 4 is implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the autonomous driving interactive simulation method based on the interactive trajectory prediction model as described in any one of claims 1 to 4 is implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the autonomous driving interactive simulation method based on the interactive trajectory prediction model as described in any one of claims 1 to 4 is implemented.

Images