Design Method, Device, Equipment and Storage Medium for Cloud Collaboration

Abstract

The present invention relates to the field of autonomous driving simulation design, and particularly to a design method, device, equipment and storage medium for cloud collaboration. The method includes the following steps: obtaining a driving video sample set and vehicle multi-modal driving parameters; extracting vehicle behavior details from the driving video sample set, and performing implicit behavior feature probability learning to generate a driving behavior probability transition rule; obtaining autonomous driving test requirements; performing three-dimensional simulation of the autonomous driving test requirements to generate a three-dimensional simulation driving scene model; performing elastic expansion processing on the vehicle multi-modal driving parameters to generate cloud driving data streams; generating cloud distributed architecture data streams based on the cloud driving data streams; performing distributed microservice function abstraction on the three-dimensional simulation driving scene model based on the driving behavior probability transition rule and the cloud distributed architecture data streams, and constructing a cloud distributed simulation engine. The present invention realizes an efficient and accurate autonomous driving simulation design method.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving simulation design, and particularly to a design method, device, equipment and storage medium for cloud collaboration. Background Art

[0002] With the rapid development of autonomous driving technology, future vehicles will integrate more emerging technologies such as artificial intelligence and the Internet of Things to achieve vehicle intelligence and autonomy. However, in practical applications, there are still many challenges in the design and debugging of autonomous driving systems. The core algorithms and decision-making mechanisms of autonomous driving engines are extremely complex and require a large amount of simulation testing and iterative optimization. Traditional stand-alone simulation systems are difficult to meet the needs of massive data processing and algorithm iteration. The efficiency of autonomous driving simulation data is low, and the accuracy of simulation is poor. Autonomous driving systems need to integrate multi-source heterogeneous data from in-vehicle sensors, maps, traffic information, etc. to achieve perception and decision-making in complex driving environments. How to effectively integrate and process these scattered data sources is another major difficulty in the design of autonomous driving systems. The verification of autonomous driving systems needs to run through the entire R & D life cycle, from algorithm design to final vehicle inspection. For the requirements of different stages, a flexible simulation test environment needs to be constructed to meet the requirements of rapid iteration and full-process verification. To solve the above problems, an intelligent simulation design method for autonomous driving engines is needed. Summary of the Invention

[0003] To solve the above technical problems, the present invention proposes a design method, device, equipment and storage medium for cloud collaboration to solve at least one of the above technical problems.

[0004] To achieve the above object, the present invention provides a design method for cloud collaboration, including the following steps:

[0005] Step S1: Obtain a driving video sample set and vehicle multi-modal driving parameters; extract vehicle behavior details from the driving video sample set, and perform implicit behavior feature probability learning to generate a driving behavior probability transition rule;

[0006] Step S2: Obtain autonomous driving test requirements; perform three-dimensional simulation of the autonomous driving test requirements to generate a three-dimensional simulation driving scene model;

[0007] Step S3: Perform elastic expansion processing on the vehicle multi-modal driving parameters to generate a cloud driving data stream; generate a cloud distributed architecture data stream based on the cloud driving data stream;

[0008] Step S4: Perform distributed microservice function abstraction on the three-dimensional simulation driving scene model based on the driving behavior probability transition rule and the cloud distributed architecture data stream, and perform distributed simulation calculation to construct a cloud distributed simulation engine;

[0009] Step S5: Perform cloud driving simulation based on the cloud distributed simulation engine to generate an autonomous driving simulation visualization view;

[0010] Step S6: Conduct dynamic autonomous driving detail learning on the autonomous driving simulation visualization view and optimize autonomous driving decisions, thereby constructing a distributed simulation optimization model to execute the autonomous driving simulation engine design task.

[0011] The present invention also provides a cloud collaboration design device, including:

[0012] A driving behavior analysis module, configured to obtain a driving video sample set and vehicle multimodal driving parameters; extract vehicle behavior details from the driving video sample set, and perform implicit behavior feature probability learning to generate a driving behavior probability transition rule;

[0013] A three-dimensional scene module, configured to obtain autonomous driving test requirements; perform three-dimensional simulation of the autonomous driving test requirements to generate a three-dimensional simulation driving scene model;

[0014] An elastic expansion module, configured to perform elastic expansion processing on the vehicle multimodal driving parameters to generate cloud driving data streams; generate cloud distributed architecture data streams based on the cloud driving data streams;

[0015] A distributed engine module, configured to perform distributed microservice function abstraction on the three-dimensional simulation driving scene model based on the driving behavior probability transition rule and the cloud distributed architecture data streams, and perform distributed simulation calculations to construct a cloud distributed simulation engine;

[0016] A simulation visualization module, configured to perform cloud driving simulation based on the cloud distributed simulation engine to generate an autonomous driving simulation visualization view;

[0017] A simulation optimization design module, configured to conduct dynamic autonomous driving detail learning on the autonomous driving simulation visualization view and optimize autonomous driving decisions, thereby constructing a distributed simulation optimization model to execute the autonomous driving simulation engine design task.

[0018] The present invention also provides a computer device, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps of the cloud collaboration design method described in any one of the above are implemented.

[0019] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the cloud collaboration design method described in any one of the above are implemented.

[0020] The design method, device, equipment and storage medium for cloud collaboration provided by the present invention have the following beneficial effects: By extracting vehicle behavior details and learning the probabilities of implicit behavior features from a driving video sample set, the driving behavior can be understood more deeply, and the probability transfer law of driving behavior can be generated, thereby improving the prediction ability and accuracy of driving behavior. Through three-dimensional simulation of the autonomous driving test requirements in a simulation scenario, the test requirements can be better understood, and a three-dimensional simulation driving scenario model can be generated to provide accurate scenario data for subsequent simulation tests. Elastic expansion processing makes the driving data more comprehensive and diverse, providing more diverse data support for subsequent data processing and analysis. Generating cloud driving data streams and cloud distributed architecture data streams realizes efficient management and sharing of data, improving data processing efficiency and scalability. The distributed microservice function abstraction and distributed simulation calculation improve the flexibility and performance of the system, optimize the simulation calculation process, and improve the calculation efficiency and response speed. Constructing a cloud distributed simulation engine realizes efficient simulation and calculation of driving scenarios, providing strong support for the design and test of autonomous driving systems. The autonomous driving simulation visualization view generated through cloud driving simulation intuitively displays the simulation results, helping to analyze driving scenarios and system performance. Dynamic autonomous driving detail learning and autonomous driving decision optimization improve the decision-making process of the autonomous driving system, enhancing the safety and reliability of the system. Constructing a distributed simulation optimization model optimizes the design of the simulation engine, improving the efficiency and accuracy of simulation, and better verifying the performance and stability of the autonomous driving system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic flow chart of the steps of a design method for cloud collaboration of the present invention;

[0022] Figure 2 It is a schematic detailed implementation step flow chart of step S1;

[0023] Figure 3 It is a schematic detailed implementation step flow chart of step S2;

[0024] Figure 4 It is a schematic detailed implementation step flow chart of step S3. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0026] The embodiments of the present application provide a design method, device, equipment and storage medium for cloud collaboration. The execution entities of the design method, device, equipment and storage medium for cloud collaboration include, but are not limited to, the following systems: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be regarded as general computing nodes of the present application. The data processing platform includes, but is not limited to, at least one of an audio and image management system, an information management system, and a cloud data management system.

[0027] In the embodiments of the present invention, refer to Figure 1 , which is a schematic flowchart of the steps of a design method for cloud collaboration of the present invention. In this example, the steps of the design method for cloud collaboration include:

[0028] Step S1: Obtain a driving video sample set and vehicle multi-modal driving parameters; extract vehicle behavior details from the driving video sample set, and perform implicit behavior feature probability learning to generate a driving behavior probability transition law;

[0029] In this embodiment, driving video data from various scenarios and environments is collected, covering different driving behaviors and states, ensuring the representativeness and diversity of video samples, and laying a foundation for subsequent behavior feature extraction and analysis. At the same time, a variety of driving-related data collected by in-vehicle sensors, cameras and other devices, such as speed, acceleration, steering angle, etc., is collected to ensure the temporal and spatial correspondence between these multi-modal driving parameters and the collected video samples. Using computer vision techniques, such as object detection and tracking, to extract detailed behavior features such as the movement trajectory and posture change of the vehicle from the driving video, and combining multi-modal driving parameters to further enrich and refine the description of these behavior features. Extract implicit behavior features from the extracted vehicle behavior details, for example, the driving style of the driver, the complexity of the road environment, etc. Select a suitable probability model, for example, Markov chain model, hidden Markov model, Bayesian network, etc., to learn the probability distribution of implicit behavior features. Use the collected driving video sample set and multi-modal driving parameters to train the selected probability model, learn the probability transition law of implicit behavior features, and generate a driving behavior probability transition matrix, which describes the transition probability between different driving behaviors. For example, in a specific situation, the probability that the driver switches from an accelerating state to a braking state.

[0030] Step S2: Obtain autonomous driving test requirements; perform three-dimensional simulation of the autonomous driving test requirements to generate a three-dimensional simulation driving scene model;

[0031] In this embodiment, in close cooperation with the R & D team of the autonomous driving system, various scenario requirements for autonomous driving tests are collected and sorted out, including various elements such as different road types, traffic conditions, weather conditions, obstacle types, etc. These requirement data provide the basic input for subsequent scenario modeling and simulation design. Using the previously constructed scenario knowledge graph, various scenario attributes (such as road type, traffic condition, weather condition, etc.) in the autonomous driving test requirements are mapped to the corresponding concepts and entities in the knowledge graph to obtain relevant semantic information and association relationships. Combining the scenario attribute information obtained from the knowledge graph, 3D modeling and physical engine simulation technologies are used to automatically construct a 3D simulation driving scenario that highly restores the actual environment. This scenario includes various elements such as road networks, traffic lights, vehicles, pedestrians, weather effects, etc., and simulates real physical interaction behaviors. Using the powerful distributed computing power of the cloud, the constructed 3D simulation scenario is deployed to the cloud to support multiple autonomous driving vehicles to perform real-time interaction simulations in this scenario. This simulation process can highly restore various dynamic factors in the actual traffic environment and provide a reliable simulation basis for the verification of autonomous driving algorithms.

[0032] Step S3: Perform elastic expansion processing on the vehicle multi-modal driving parameters to generate cloud driving data streams; generate cloud distributed architecture data streams based on the cloud driving data streams;

[0033] In this embodiment, the data is divided into multiple data shards, and each data shard contains a part of the data, which is convenient for distributed storage and processing. The data is compressed, for example, using compression algorithms such as gzip and zip, to reduce the data storage space and improve the data transmission efficiency. A suitable cloud data stream platform is selected, such as Apache Kafka, Apache Flink, Amazon Kinesis, etc., for constructing cloud driving data streams. According to the data characteristics and requirements, the parameters of the data stream platform are configured, such as data topics, data formats, data throughput, etc. The data after elastic expansion processing is published to the cloud data stream platform to form cloud driving data streams. A suitable cloud distributed architecture is selected, such as a microservices architecture, a message queue architecture, a data warehouse architecture, etc. According to the architecture selection, the corresponding architecture components are selected, such as message queues, databases, data processing engines, etc. The architecture components are deployed to the cloud environment, for example, using cloud service platforms such as AWS, Azure, and GCP for deployment.

[0034] Step S4: Based on the driving behavior probability transition law and the cloud distributed architecture data stream, perform distributed microservices function abstraction on the 3D simulation driving scenario model and perform distributed simulation calculations to construct a cloud distributed simulation engine;

[0035] In this embodiment, the three-dimensional simulation driving scenario model is decomposed into multiple functional modules. For example, road models, vehicle models, traffic signal models, pedestrian models, etc. Each functional module is abstracted into a microservice. For example, road model microservice, vehicle model microservice, traffic signal model microservice, pedestrian model microservice, etc. Service interfaces are defined for each microservice for communication and data interaction between other microservices. The driving behavior probability transition law is used to simulate the driving behavior of the driver. For example, steering, accelerating, braking, etc. The interaction between the vehicle and the environment is simulated. For example, the friction between the vehicle and the road, the collision between the vehicle and the pedestrian, etc. The data of vehicle sensors are simulated. For example, camera data, radar data, GPS data, etc. The data stream of the cloud distributed architecture is mapped into different microservices. For example, the vehicle sensor data is sent to the vehicle model microservice, and the road information is sent to the road model microservice, etc. Ensure the data stream synchronization between different microservices. For example, use technologies such as message queues and databases for data synchronization. Select a suitable distributed simulation framework. For example, SimPy, Mesa, Repast, etc., for building a cloud distributed simulation engine. According to the simulation requirements, configure the parameters of the distributed simulation framework. For example, simulation time step, simulation accuracy, simulation data storage, etc. Allocate the simulation tasks to different microservices. For example, allocate the simulation task of the vehicle model to the vehicle model microservice, and allocate the simulation task of the road model to the road model microservice, etc. Use the distributed simulation framework to achieve parallel computing between different microservices and improve the simulation efficiency. Ensure the data synchronization between different microservices. For example, use technologies such as message queues and databases for data synchronization. Integrate all microservices and the distributed simulation framework together to build a cloud distributed simulation engine. Deploy the cloud distributed simulation engine to the cloud environment. For example, use cloud service platforms such as AWS, Azure, GCP, etc. for deployment.

[0036] Step S5: Perform cloud driving simulation based on the cloud distributed simulation engine to generate an autonomous driving simulation visualization view;

[0037] In this embodiment, a simulation running environment is built in the cloud, and a distributed simulation engine is deployed. Here, virtual machine technology is adopted to run simulation computing tasks in public clouds or private clouds. Prepare basic map and traffic environment resources, including road network maps, traffic signal positions, vehicle driving rules, etc. These data need to be preprocessed into a computer-readable format. Define simulation scenarios and participating objects, such as scenario settings like simulation areas, time, types and quantities of participating vehicle models, etc. Configure the autonomous driving algorithm module, including two main parts: perception and decision-making, and simulate based on real algorithm models. Run the simulation task, and perform interactive simulation on participating objects through a perception-decision-control loop. Process and analyze the simulation results, such as vehicle driving trajectories, control instructions, following distance distributions, etc. Generate 3D visualization results, and package the simulation results into videos or roaming scenarios according to the time sequence for display. Compare and verify the simulation effects, and match and verify with the actual test results.

[0038] Step S6: Dynamically learn the details of autonomous driving from the autonomous driving simulation visualization view, and optimize autonomous driving decisions, thereby constructing a distributed simulation optimization model to execute the autonomous driving simulation engine design task.

[0039] In this embodiment, label the simulation visualization results to obtain action labels such as lane change and tracking. Extract vehicle-related behavior features from them, such as speed, acceleration and deceleration change trends, etc. Apply deep learning algorithms to train the labeled data set to learn the knowledge of driving dynamic details. Build a dynamic deep neural network model in the cloud, and connect the real-time simulation output to the input end of the model. When the model runs, collect vehicle perception inputs and output the optimal decision-making behavior within the frame. Implement the decision-making behavior as a control signal and input it into the simulation engine to perform directional simulation for verification. Compare the actual results with the model output, calculate the error and feedback it to the model to update the parameters. After the model parameters converge iteratively, the output is the distributed optimization model framework. The model is directly connected to the simulation platform to implement a parallel intelligent driving decision-making execution system. Continuously learn new samples for optimization to continuously improve the intelligent level of autonomous driving.

[0040] In this embodiment, refer to Figure 2 , which is a schematic diagram of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of the said step S1 include:

[0041] Step S11: Obtain a set of driving video samples and multi-modal driving parameters of the vehicle;

[0042] Step S12: Extract vehicle behavior details from the set of driving video samples to obtain vehicle behavior detail data;

[0043] Step S13: Perform convolutional spatial encoding on the vehicle behavior detail data to generate vehicle spatial encoding features;

[0044] Step S14: Perform time-domain continuous frame tracking on the vehicle space coding features to generate a vehicle behavior change trajectory;

[0045] Step S15: Mine deep behavior features from the vehicle behavior change trajectory to obtain driving behavior feature data;

[0046] Step S16: Perform implicit behavior feature probability learning on the driving behavior feature data to generate a driving behavior probability transition rule.

[0047] In this embodiment, after obtaining the authorization of the user and the cloud platform, a driving video sample set is collected from sources such as public data sets, on-vehicle cameras, and simulated driving environments. Vehicle multi-modal driving parameters are synchronously collected. For example: Video data: The driving video itself, containing information such as road environment, vehicle status, and driver behavior. Sensor data: Vehicle sensor data, such as speed, steering wheel angle, accelerator pedal, brake pedal, etc. GPS data: Vehicle location information, such as longitude, latitude, altitude, etc. Other data: Other relevant data, such as timestamp, weather condition, road condition information, etc. The driving video is preprocessed, such as removing noise, adjusting brightness, cropping the picture, etc. Object detection algorithms, such as YOLO, FasterR-CNN, etc., are used to identify objects such as vehicles, roads, pedestrians, traffic lights, etc. in the video frame. Behavior recognition algorithms, such as LSTM, CNN, etc., are used to identify vehicle behaviors, such as turning, accelerating, decelerating, braking, etc. The detailed information of vehicle behaviors is extracted, such as steering angle, acceleration time, deceleration time, braking distance, etc. Vehicle behavior detail data is generated, containing the vehicle behavior detail information of each video frame, as the input data for the subsequent steps. Convolutional neural networks, such as ResNet, VGG, etc., are used to perform convolutional spatial encoding on the vehicle behavior detail data. The convolutional neural network extracts the spatial features of the vehicle behavior detail data, such as vehicle location, shape, color, etc. Vehicle spatial encoding features are generated, containing the vehicle spatial feature information of each video frame, as the input data for the subsequent steps. Time-domain feature extraction algorithms, such as LSTM, RNN, etc., are used to extract the time-domain features of the vehicle spatial encoding features, such as vehicle movement trajectory, speed change, etc. According to the extracted time-domain features, a vehicle behavior change trajectory is generated, recording the vehicle behavior change information of each video frame, as the input data for the subsequent steps. Deep learning models, such as LSTM, RNN, Transformer, etc., are used to perform deep behavior feature mining on the vehicle behavior change trajectory. The deep learning model extracts the driving behavior features of the vehicle behavior change trajectory, such as driving style, driving habit, driving risk, etc. Driving behavior feature data is generated, containing the driving behavior feature information of each video frame, as the input data for the subsequent steps. The hidden Markov model (HMM) or other probability models are used to perform implicit behavior feature probability learning on the driving behavior feature data. The HMM model generates a driving behavior probability transition matrix, recording the probability transition relationships between different driving behaviors, such as the probability of transitioning from the acceleration state to the deceleration state, the probability of transitioning from the normal driving state to the dangerous driving state, etc. A driving behavior probability transition rule is generated, containing the probability transition relationships between driving behaviors, for predicting the future change trend of driving behaviors.

[0048] In this embodiment, refer to Figure 3, which is a schematic diagram of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0049] Step S21: Obtain the autonomous driving test requirements;

[0050] Step S22: Identify the scene attribute features of the autonomous driving test requirements to generate driving scene attribute data;

[0051] Step S23: Conduct cross-feature learning on the driving scene attribute data to construct a scene knowledge graph;

[0052] Step S24: Analyze the scenario requirements use cases of the autonomous driving test requirements to generate customized scenario use cases;

[0053] Step S25: Based on the scene knowledge graph, perform three-dimensional simulation of the customized scenario use cases to generate a three-dimensional simulation driving scene model.

[0054] In this embodiment, test requirements are collected from aspects such as the autonomous driving system development team and the road management department. The requirements include various requirements such as test scenarios, test objectives, and test metrics. Analyze the scene descriptions in the test requirements, identify attribute features such as road type, weather conditions, and traffic conditions, organize the identified scene attribute features into structured scene attribute data, and use methods such as machine learning to analyze the correlation and dependency relationships between the scene attribute data. Based on the associations between the attribute features, construct a knowledge graph describing different driving scenes. According to the test requirements, design scenario test cases that meet specific objectives. The use cases contain information such as scene attributes, test steps, and expected results. Use the attribute feature information in the scene knowledge graph to construct a three-dimensional simulation environment and map the customized scenario use cases to the three-dimensional simulation model to achieve visual simulation of the scenario.

[0055] In this embodiment, refer to Figure 4 , which is a schematic diagram of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0056] Step S31: Perform heterogeneous data fusion on the vehicle's multi-modal driving parameters to generate heterogeneous fusion driving parameters;

[0057] Step S32: Build a cloud platform-based cloud data transmission framework;

[0058] Step S33: Use the cloud data transmission framework to perform elastic expansion processing on the heterogeneous fusion driving parameters to generate cloud driving data streams;

[0059] Step S34: Conduct cloud storage structure aggregation on the cloud driving data streams to generate cloud distributed architecture data streams.

[0060] In this embodiment, driving-related data collected by various devices such as vehicle-mounted sensors, cameras, and GPS, such as vehicle speed, position, direction, etc., are collected. Through preprocessing such as data cleaning and format conversion, these heterogeneous data are effectively fused to generate unified heterogeneous fusion driving parameters. Functional modules such as data reception, transmission, and processing are built on the cloud computing platform, and a cloud data transmission channel based on technologies such as message queues and data buses is designed to achieve two-way data interaction between vehicle-mounted devices and the cloud. Elastic scalable cloud computing resources are used to handle the high-concurrency driving data generated by a large number of vehicles. The heterogeneous fusion driving parameters are converted into a cloud driving data stream that can be continuously and stably transmitted through the cloud data transmission framework, and the cloud driving data stream is stored in a distributed database or an object storage system. A distributed storage architecture is adopted to achieve high reliability, high scalability, and high-performance access to data.

[0061] In this embodiment, step S4 includes the following steps:

[0062] Step S41: Perform deep behavior coupling on the three-dimensional simulation driving scene model based on the driving behavior probability transition law to generate an initial driving simulation model;

[0063] Step S42: Perform dynamic driving behavior mapping simulation on the initial driving simulation model to construct a dynamic driving three-dimensional twin model;

[0064] Step S43: Perform distributed microservice function abstraction on the dynamic driving three-dimensional twin model to obtain multiple microservice modules;

[0065] Step S44: Perform distributed simulation calculations on multiple microservice modules according to the cloud distributed architecture data stream to construct a cloud distributed simulation engine.

[0066] In this embodiment, deep learning algorithms, such as reinforcement learning, deep neural networks, etc., are used to couple the driving behavior probability transition matrix with the three-dimensional simulation driving scenario model. An initial driving simulation model is generated, which can simulate the driving behavior of the driver in the three-dimensional simulation driving scenario model according to the driving behavior probability transition law. Define the driving behavior mapping rule to map different driving behaviors in the driving behavior probability transition matrix to specific operations in the three-dimensional simulation driving scenario model. For example, map the acceleration state to the acceleration operation of the vehicle, and map the braking state to the braking operation of the vehicle. Use dynamic mapping algorithms, such as time-step-based mapping, event-triggered mapping, etc., to dynamically map the driving behaviors in the driving behavior probability transition matrix to the three-dimensional simulation driving scenario model. Select a suitable dynamic driving three-dimensional twin model framework, such as a physics-based framework, a machine learning-based framework, etc. Set the parameters of the dynamic driving three-dimensional twin model according to the initial driving simulation model, the driving behavior mapping rule, and the dynamic mapping algorithm, such as the dynamic parameters of the vehicle, the environmental parameters, etc. Train the dynamic driving three-dimensional twin model using training data, such as historical driving data, simulation data, etc. Build a dynamic driving three-dimensional twin model that can simulate the driving behavior of the driver in the three-dimensional simulation driving scenario model and predict the driving results according to different driving conditions. Divide the functions of the dynamic driving three-dimensional twin model. For example, abstract functions such as vehicle dynamics simulation, environmental perception simulation, and driving decision-making simulation into independent modules. Define microservice interfaces for each functional module for other modules or external systems to call. Develop multiple microservice modules according to the function division and microservice interface definition, and each module is responsible for a specific function. Test each microservice module to ensure its normal function and meet the performance requirements. Select a suitable cloud platform, such as AWS, Azure, Google Cloud, etc., for deploying the distributed simulation engine. Design the data flow of the cloud distributed architecture, define the data interaction method between each microservice module, and the data transmission path on the cloud platform. Deploy multiple microservice modules to the cloud platform and connect them according to the data flow design. Build a cloud distributed simulation engine that can coordinate multiple microservice modules for distributed simulation calculations and perform data interactions according to the data flow.

[0067] In this embodiment, the specific steps of step S5 are as follows:

[0068] Step S51: Perform cloud driving simulation based on the cloud distributed simulation engine to generate autonomous driving simulation data;

[0069] Step S52: Perform interactive simulation response analysis on the autonomous driving simulation data to obtain autonomous driving interactive response data;

[0070] Step S53: Perform simulation visualization rendering on the autonomous driving interaction response data to generate an autonomous driving simulation visualization view.

[0071] In this embodiment, start the cloud distributed simulation engine and initialize it according to the simulation scenario settings. Run the simulation engine to simulate the driving process of the autonomous driving vehicle in the simulation scenario and record relevant data, such as the vehicle's position, speed, steering angle, sensor data, etc. Generate autonomous driving simulation data, including vehicle state data, environmental data, sensor data, etc. Build an autonomous driving interaction simulation model, such as using a neural network model, decision tree model, etc., to simulate the response of the autonomous driving system to different environments and sensor data. Input the autonomous driving simulation data into the interaction simulation model to simulate the response process of the autonomous driving system. Generate autonomous driving interaction response data, including the decisions, control commands, vehicle actions, etc. of the autonomous driving system. Use data analysis methods, such as statistical analysis, machine learning, etc., to analyze the autonomous driving interaction response data, such as analyzing the impact of different environments and sensor data on the decisions of the autonomous driving system.

[0072] According to evaluation metrics, such as safety performance, efficiency performance, comfort performance, etc., evaluate the performance of the autonomous driving system. Select a suitable simulation visualization rendering engine, such as Unity, Unreal Engine, etc., for rendering the autonomous driving simulation visualization view. Set the parameters of the rendering engine according to requirements, such as rendering quality, rendering effect, etc. Input the autonomous driving simulation data and interaction response data into the rendering engine. Use the rendering engine to render the autonomous driving simulation visualization view according to the input data, such as showing the vehicle driving trajectory, sensor data, decision-making process, etc. Output the autonomous driving simulation visualization view, such as presenting it in the form of video, image, animation, etc.

[0073] In this embodiment, the specific steps of Step S6 are as follows:

[0074] Step S61: Perform dynamic autonomous driving detail learning on the autonomous driving simulation visualization view to obtain autonomous driving detail features;

[0075] Step S62: Perform behavior feature evaluation on the autonomous driving detail features to generate an autonomous driving behavior performance evaluation value;

[0076] Step S63: Optimize the autonomous driving decisions of the cloud distributed simulation engine based on the autonomous driving behavior performance evaluation value, thereby constructing a distributed simulation optimization model to execute the autonomous driving simulation engine design task.

[0077] In this embodiment, image processing technologies such as edge detection, feature point extraction, and target recognition are used to extract the detailed features of autonomous driving from the visualization view of autonomous driving simulation. The types of features extracted include: vehicle driving trajectory, steering angle, vehicle speed change, sensor data change, environmental change, etc. Data analysis methods such as statistical analysis and machine learning are used to analyze the detailed features of autonomous driving extracted, such as analyzing the relationships between different detailed features and the impacts of different detailed features on the autonomous driving system. According to the analysis results, the detailed features of autonomous driving are clustered, for example, similar features are grouped together to facilitate subsequent evaluation and optimization. According to the performance requirements of the autonomous driving system, behavioral feature evaluation indicators are defined, such as safety performance, efficiency performance, comfort performance, etc. The calculation methods for each evaluation indicator are defined. For example, safety performance is measured using indicators such as the number of collisions and the severity of collisions, efficiency performance is measured using indicators such as driving time and driving distance, and comfort performance is measured using indicators such as acceleration change and steering angle change. An autonomous driving behavioral performance evaluation model is constructed, such as using a neural network model, a decision tree model, etc. According to the evaluation indicators and calculation methods, the detailed features of autonomous driving are evaluated. The evaluation model is used to evaluate each detailed feature of autonomous driving to generate an autonomous driving behavioral performance evaluation value. According to the performance requirements of the autonomous driving system, optimization goals are defined, such as improving safety performance, improving efficiency performance, improving comfort performance, etc. According to the optimization goals, appropriate optimization strategies are selected, such as optimization based on reinforcement learning, optimization based on genetic algorithms, etc. A suitable distributed simulation optimization model framework is selected, such as a framework based on surrogate models, a framework based on multi-objective optimization, etc. According to the optimization goals, optimization strategies, and autonomous driving behavioral performance evaluation values, the parameters of the distributed simulation optimization model are set. The distributed simulation optimization model is trained using training data, such as historical simulation data, simulation data, etc. The optimization tasks are assigned to multiple microservice modules in the cloud distributed simulation engine. Each microservice module optimizes the autonomous driving decision according to the optimization strategy, such as adjusting vehicle control parameters and optimizing path planning. According to the optimization results, the autonomous driving simulation is run again, and the optimization effect is evaluated using the autonomous driving behavioral performance evaluation model.

[0078] In this embodiment, the present invention also provides a cloud collaboration design device, including:

[0079] A driving behavior analysis module, configured to obtain a driving video sample set and vehicle multimodal driving parameters; extract vehicle behavior details from the driving video sample set, and perform implicit behavior feature probability learning to generate a driving behavior probability transition rule;

[0080] A three-dimensional scene module for obtaining autonomous driving test requirements; performing three-dimensional simulation of the autonomous driving test requirements to generate a three-dimensional simulation driving scene model;

[0081] An elastic expansion module for performing elastic expansion processing on vehicle multimodal driving parameters to generate cloud driving data streams; generating cloud distributed architecture data streams based on the cloud driving data streams;

[0082] A distributed engine module for performing distributed microservice function abstraction on the three-dimensional simulation driving scene model based on the driving behavior probability transition law and the cloud distributed architecture data streams, and performing distributed simulation calculations to construct a cloud distributed simulation engine;

[0083] A simulation visualization module for performing cloud driving simulation based on the cloud distributed simulation engine to generate an autonomous driving simulation visualization view;

[0084] A simulation optimization design module for performing dynamic autonomous driving detail learning on the autonomous driving simulation visualization view and performing autonomous driving decision optimization to construct a distributed simulation optimization model to execute the autonomous driving simulation engine design operation.

[0085] The present invention also provides a computer device, including a memory and a processor, where a computer program is stored in the memory, and when the processor executes the computer program, the method for cloud collaboration design described in any one of the above is implemented.

[0086] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method steps for cloud collaboration design described in any one of the above are implemented.

[0087] Those skilled in the art clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units are referred to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0088] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it is stored in a computer-readable storage medium. Based on such an understanding, the technical solution of this application essentially, or the part that contributes to the prior art, or all or part of this technical solution is embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media for storing program codes such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0089] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes that fall within the meaning and scope of the equivalent elements of the application documents within the present invention.

[0090] As described above, these are only the specific implementation manners of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A design method for cloud collaboration, characterized in that, Including the following steps: Step S1: Obtain a driving video sample set and vehicle multi-modal driving parameters; extract vehicle behavior details from the driving video sample set, and perform implicit behavior feature probability learning to generate a driving behavior probability transition rule; Step S2: Obtain autonomous driving test requirements; perform three-dimensional simulation of the autonomous driving test requirements to generate a three-dimensional simulation driving scene model; Step S3: Perform elastic expansion processing on the vehicle multi-modal driving parameters to generate cloud driving data streams; generate cloud distributed architecture data streams based on the cloud driving data streams; Step S4: Based on the driving behavior probability transition rule and the cloud distributed architecture data streams, perform distributed microservice function abstraction on the three-dimensional simulation driving scene model, and perform distributed simulation calculations to construct a cloud distributed simulation engine; Step S5: Perform cloud driving simulation based on the cloud distributed simulation engine to generate an autonomous driving simulation visualization view; Step S6: Perform dynamic autonomous driving detail learning on the autonomous driving simulation visualization view, and perform autonomous driving decision optimization to construct a distributed simulation optimization model to execute the autonomous driving simulation engine design task; Among them, the specific steps of Step S2 are: Step S21: Obtain autonomous driving test requirements; Step S22: Perform scene attribute feature recognition on the autonomous driving test requirements to generate driving scene attribute data; Step S23: Perform cross-feature learning on the driving scene attribute data to construct a scene knowledge graph; Step S24: Perform scene requirement use case analysis on the autonomous driving test requirements to generate customized scene use cases; Step S25: Perform three-dimensional simulation of the customized scene use cases based on the scene knowledge graph to generate a three-dimensional simulation driving scene model; Among them, the specific steps of Step S4 are: Step S41: Perform deep behavior coupling on the three-dimensional simulation driving scene model based on the driving behavior probability transition rule to generate an initial driving simulation model; Step S42: Perform dynamic driving behavior mapping simulation on the initial driving simulation model to construct a dynamic driving three-dimensional twin model; Step S43: Perform distributed microservice function abstraction on the dynamic driving three-dimensional twin model to obtain multiple microservice modules; Step S44: Perform distributed simulation calculations on the multiple microservice modules according to the cloud distributed architecture data streams to construct a cloud distributed simulation engine.

2. The design method for cloud collaboration according to claim 1, characterized in that The specific steps of Step S1 are: Step S11: Obtain a driving video sample set and vehicle multi-modal driving parameters; Step S12: Extract vehicle behavior details from the driving video sample set to obtain vehicle behavior detail data; Step S13: Perform convolutional space encoding on the vehicle behavior detail data to generate vehicle space encoding features; Step S14: Perform time-domain continuous frame tracking on the vehicle space encoding features to generate a vehicle behavior change trajectory; Step S15: Perform deep behavior feature mining on the vehicle behavior change trajectory to obtain driving behavior feature data; Step S16: Perform implicit behavior feature probability learning on the driving behavior feature data to generate a driving behavior probability transition rule.

3. The design method for cloud collaboration according to claim 1, characterized in that, The specific steps of Step S3 are: Step S31: Perform heterogeneous data fusion on vehicle multimodal driving parameters to generate heterogeneous fusion driving parameters; Step S32: Build a cloud data transmission framework based on the cloud platform; Step S33: Use the cloud data transmission framework to perform elastic expansion processing on the heterogeneous fusion driving parameters to generate cloud driving data streams; Step S34: Aggregate the cloud driving data streams into a cloud storage structure to generate cloud distributed architecture data streams.

4. The design method for cloud collaboration according to claim 1, wherein The specific steps of Step S5 are as follows: Step S51: Perform cloud driving simulation based on the cloud distributed simulation engine to generate autonomous driving simulation data; Step S52: Analyze the interactive simulation response of the autonomous driving simulation data to obtain autonomous driving interactive response data; Step S53: Perform simulation visualization rendering on the autonomous driving interactive response data to generate an autonomous driving simulation visualization view.

5. The design method for cloud collaboration according to claim 1, wherein The specific steps of Step S6 are as follows: Step S61: Perform dynamic autonomous driving detail learning on the autonomous driving simulation visualization view to obtain autonomous driving detail features; Step S62: Evaluate the behavioral features of the autonomous driving detail features to generate an autonomous driving behavior performance evaluation value; Step S63: Optimize the autonomous driving decision-making of the cloud distributed simulation engine based on the autonomous driving behavior performance evaluation value to build a distributed simulation optimization model for performing autonomous driving simulation engine design tasks.

6. A design device for cloud collaboration, characterized in that, A design method for performing cloud collaboration as described in claim 1, including: A driving behavior analysis module for obtaining a driving video sample set and vehicle multimodal driving parameters; extracting vehicle behavior details from the driving video sample set and performing implicit behavioral feature probability learning to generate driving behavior probability transition rules; A three-dimensional scene module for obtaining autonomous driving test requirements; performing three-dimensional simulation of the autonomous driving test requirements to generate a three-dimensional simulation driving scene model; An elastic expansion module for performing elastic expansion processing on vehicle multimodal driving parameters to generate cloud driving data streams; generating cloud distributed architecture data streams based on the cloud driving data streams; A distributed engine module for abstracting the distributed microservice functions of the three-dimensional simulation driving scene model based on the driving behavior probability transition rules and the cloud distributed architecture data streams and performing distributed simulation calculations to build a cloud distributed simulation engine; A simulation visualization module for performing cloud driving simulation based on the cloud distributed simulation engine to generate an autonomous driving simulation visualization view; A simulation optimization design module for performing dynamic autonomous driving detail learning on the autonomous driving simulation visualization view and performing autonomous driving decision optimization to build a distributed simulation optimization model for performing autonomous driving simulation engine design tasks.

7. A computer device, comprising a memory and a processor, wherein a computer program is stored in the memory, characterized in that When the processor executes the computer program, it implements the steps of the cloud collaboration design method described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the cloud collaboration design method described in any one of claims 1 to 5.

Images