Vehicle automatic driving simulation system and simulation method

Abstract

The invention discloses a vehicle automatic driving simulation system and simulation method, and the method comprises the steps: transmitting a built simulation configuration to a simulator of each cluster node at a resource configurator side; the cluster comprises a plurality of nodes, the nodes are divided into main nodes and sub-nodes, and the sub-nodes comprise computing nodes and rendering nodes; each cluster node simulator side receives the simulation configuration issued by the resource configurator and obtains AB packet resources and scene resource parameters according to the resource id in the simulation configuration, and a simulator of the cluster node asynchronously loads a scene according to the AB packet resources and sets and initializes the scene according to the scene resource parameters; and according to cluster configuration in the simulation configuration, realizing data communication and frame synchronization among the simulation resources of the main node, the computing node and the rendering node through cluster simulation scheduling until the loading of the simulation resources is finished, and running simulation. Through distributed deployment and dynamic resource management, scene rendering efficiency and simulation quality in a multi-host-vehicle and multi-sensor scene are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving simulation, and particularly to a vehicle autonomous driving simulation system and a simulation method. Background Art

[0002] At present, the unmanned driving technology is in a stage of rapid development. Among them, the unmanned driving technology of mining construction machinery is an important branch. Due to the special geographical conditions of the mining area, extreme environmental factors such as altitude and weather are likely to exist. Using manual driving has a large number of potential safety hazards and extremely high labor costs. The driving speed of mining trucks is slow, the transportation routes in the mining area are relatively fixed, and the vehicle operation speed is slow. All these make the mining area one of the most suitable environments for using unmanned vehicles. However, as an emerging industry, the technology of unmanned mining in mines is not yet fully mature and requires time to develop. The safety, stability, and reliability of its algorithms need to be verified. The unmanned mining operation tests in mining areas are generally divided into four forms: real vehicle running tests in mining areas, real vehicle running tests in test fields, hardware-in-the-loop tests, and software-in-the-loop tests.

[0003] Real vehicle running tests in mining areas: Due to the special terrain and slow running speed of mining trucks in mining areas, the efficiency of such tests is extremely low and accidents are likely to occur, causing potential safety hazards.

[0004] Running tests in test fields: The construction cost of test fields is too high, and it is difficult to simulate the extreme environment in mining areas.

[0005] Hardware-in-the-loop tests: Dependent on physical hardware, it is difficult to conduct simulation work at any time; it is difficult to conduct verification work at any time when carrying out R & D work such as control planning for mining trucks.

[0006] Therefore, it is very necessary to conduct software-in-the-loop simulation before the above-mentioned tests. By simulating the scene as much as possible to restore the mining area environment, the functionality, safety, and stability of the unmanned driving system can be tested at the simulation level, thus avoiding the problems existing in real vehicle tests in mining areas, test field tests, and hardware-in-the-loop tests. With the progress of 3D rendering engine technology, more and more enterprises begin to use such software for scene simulation and establish a software-in-the-loop simulation environment for algorithms.

[0007] Autonomous driving simulation in mine scenes requires a large amount of map resources to be loaded, many vehicle models, and many heavy sensors. A single host cannot independently complete the rendering of the entire scene. Therefore, it is necessary to introduce a distributed cluster rendering framework. With the improvement of algorithm complexity and the increasing requirements for computer rendering performance, the market urgently needs a new simulation framework for autonomous driving algorithms of construction machinery. Summary of the Invention

[0008] Distributed cluster rendering refers to leveraging the computing resources of multiple computers to improve the rendering efficiency and processing power of scene simulation, which is particularly important for the testing of autonomous driving algorithms that require large-scale or highly complex scene simulations. Through distributed cluster rendering, the simulator can simulate more complex traffic environments and weather conditions, providing more realistic test scenarios. In distributed cluster rendering, each computer can undertake a part of the rendering task or simulate a specific part of the scene, significantly enhancing the running speed of the simulation and the ability to process large-scale data. Additionally, distributed cluster rendering also improves the scalability of the simulation, allowing users to add more computing nodes as needed to expand the scale of the simulation.

[0009] The objective of the present invention is to provide a vehicle autonomous driving simulation system and simulation method, which introduce a modular construction method and a distributed simulation architecture to improve resource utilization and scene rendering efficiency.

[0010] To achieve the above objective, the technical solution adopted by the present invention is as follows: In the first aspect, the present invention provides a vehicle autonomous driving simulation system, which deploys a distributed simulation environment in the manner of cluster simulation. The simulation system includes: A resource configurator, which is used for the management of the cluster and nodes, as well as the configuration and management of scene resources; wherein the cluster includes multiple nodes, and the nodes are divided into a master node and slave nodes, and the slave nodes include computing nodes and rendering nodes; Multiple simulators, with one simulator deployed on each node. All the simulators form a simulation cluster through local area network connection. The simulators on each node are used to receive the simulation configuration issued by the resource configurator and load and run the actual environment of the simulation resources according to the simulation configuration.

[0011] In the second aspect, the present invention provides a vehicle autonomous driving simulation method, which is implemented by using the above vehicle autonomous driving simulation system. The method includes: On the side of the resource configurator, issue the created simulation configuration to the simulators on each cluster node; wherein the cluster includes multiple nodes, and the nodes are divided into a master node and slave nodes, and the slave nodes include computing nodes and rendering nodes; On the side of the simulators on each cluster node, receive the simulation configuration issued by the resource configurator, obtain the AB package resources and scene resource parameters according to the resource id in the simulation configuration. The simulators on the cluster nodes asynchronously load the scene according to the AB package resources, set and initialize the scene according to the scene resource parameters; according to the cluster configuration in the simulation configuration, realize data communication and frame synchronization among the simulation resources of the master node, computing nodes, and rendering nodes through cluster simulation scheduling until the simulation resources are loaded and the simulation runs.

[0012] In some embodiments, the configuration method of the cluster includes: Select multiple nodes; Set one main node from the multiple nodes, and the remaining nodes are all sub - nodes; Assign the sub - nodes as computing nodes and rendering nodes to form a cluster.

[0013] In some embodiments, the method for creating the simulation configuration includes: creating a new sensor configuration according to the sensor parameters on the main vehicle configuration page, and successively selecting the map, main vehicle, main vehicle sensor configuration, weather environment, NPC, cluster, and bridge type on the new simulation page.

[0014] In some embodiments, the simulator receives the simulation configuration sent by the resource configurator through the HTTP communication method.

[0015] In some embodiments, obtaining the AB - package resources according to the resource id in the simulation configuration includes: The simulator retrieves the corresponding AB - package resource id in the parameter database according to the resource id in the simulation configuration; The simulator checks whether there is such an AB - package resource in the local cache directory; if so, directly load the AB - package resource from the local cache directory into the scene; if not, download the corresponding AB - package resource from the resource server.

[0016] In some embodiments, obtaining the scene resource parameters and setting and initializing the scene according to the scene resource parameters includes: Retrieve the relevant scene resource parameters in the parameter database according to the resource id in the simulation configuration, and then set the parameters of the sensor, main vehicle, weather environment, and NPC behavior according to the scene resource parameters and initialize the scene.

[0017] In some embodiments, the cluster simulation scheduling includes: After each node receives the simulation configuration, the main node will first enter the scene preparation stage to wait for the slave nodes to access the simulation. After the slave nodes access the simulation, the scene initialization and loading stage starts, and each node will initialize the scene and load the corresponding resources; among them, the main node is responsible for resource allocation, the computing nodes are responsible for computing tasks, and the rendering nodes are responsible for rendering tasks; Each computing node and rendering node feedbacks the sub - node load to the main node through the Websocket communication method; After the main node receives the loads of each sub - node, it dynamically allocates resources according to the loads of each sub - node; After the main node receives the calculation end flags of all computing nodes and the rendering end flags of all rendering nodes, it re - allocates the calculation and rendering tasks according to the loads of each sub - node, so as to ensure frame synchronization among the nodes.

[0018] In some embodiments, dynamically allocating resources according to the loads of each sub - node includes: First, decompose the sensor tasks into computing tasks and rendering tasks; then allocate computing resources according to the loads of all computing nodes, and allocate rendering resources according to the loads of all rendering nodes. Among them, the computing resource allocation includes: allocating the computing tasks of heavy computing sensors to computing nodes with lower computing loads, and allocating the computing tasks of light computing sensors to computing nodes with higher computing loads. Among them, the rendering resource allocation includes: allocating the rendering tasks of heavy rendering sensors to rendering nodes with lower rendering loads, and allocating the rendering tasks of light rendering sensors to rendering nodes with higher rendering loads.

[0019] In some embodiments, each cluster has only one master node, which can run multiple master vehicle models, centrally manage the computing and rendering tasks of all master vehicle sensors, and dynamically allocate the sensor computing and rendering tasks according to the loads of the rendering nodes and computing nodes.

[0020] Beneficial effects brought by the technical solution of the present invention: (1) Improve rendering efficiency: Distributed cluster rendering can process a large number of rendering tasks in parallel, significantly improving the efficiency of scene rendering, especially in complex scenarios with multiple master vehicles and multiple sensors.

[0021] (2) Load balancing: Through dynamic resource allocation, cluster rendering can achieve load balancing, allocate heavy sensors and light rendering sensors to nodes with different loads, optimize resource utilization, and avoid overload.

[0022] (3) Improve simulation quality: Cluster rendering can handle more complex simulation tasks through distributed computing, provide higher-quality simulation results, and provide more accurate test data for the research and development of autonomous driving technology.

[0023] (4) Scalability: As the simulation requirements increase, the cluster rendering architecture can be easily expanded by configuring more nodes in the simulation configurator to improve computing and rendering capabilities and meet the needs of larger-scale simulations.

[0024] (5) Fine-grained management: The master node can dynamically adjust resource allocation according to the behavior and state of each master vehicle and the associated sensor data, ensuring that each master vehicle and its sensors can obtain reasonable computing resources. This fine-grained resource management not only improves the efficiency of the simulation but also enhances the accuracy and reliability of the simulation results, providing strong support for the research of autonomous driving technology in multi-master vehicle scenarios.

[0025] (6) Adaptive adjustment: The master node adjusts the resource allocation strategy in a timely manner by monitoring the operation status of each master vehicle and the changes in sensor data in real time, so as to adapt to the simulation requirements of different master vehicles in different scenarios. This adaptive adjustment ability enables the simulation system to more flexibly handle complex situations in the multi-master vehicle scenario and provides a more reliable test platform for the research and development of autonomous driving technology. Brief Description of the Drawings

[0026] Figure 1 It is a schematic diagram of the cluster simulation creation process in an embodiment of the present invention; Figure 2 It is a schematic diagram of the resource loading process of each node simulator in an embodiment of the present invention; Figure 3 It is a schematic diagram of the scheduling process of each node in the cluster simulation in an embodiment of the present invention; Figure 4 It is a schematic diagram of the resource allocation process in the cluster simulation in an embodiment of the present invention. Detailed Embodiment

[0027] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in combination with the embodiments and the drawings. Herein, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but do not limit the present invention.

[0028] Herein, it should also be noted that in order to avoid obscuring the present invention due to unnecessary details, only the structures and / or processing steps closely related to the solution of the present invention are shown in the drawings, and other details less related to the present invention are omitted.

[0029] It should be emphasized that the term "including / comprising" when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

[0030] Herein, it should also be noted that if not specifically stated, the term "connection" in this document can not only refer to direct connection, but also represent indirect connection with an intermediate.

[0031] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0032] It should be emphasized here that the step labels mentioned hereinafter are not intended to limit the order of the steps. Instead, it should be understood that the steps can be executed in the order mentioned in the embodiments, or different from the order in the embodiments, or several steps can be executed simultaneously.

[0033] An embodiment of the present invention provides a vehicle autonomous driving simulation system, which deploys a distributed simulation environment in a cluster simulation manner. The simulation system includes: A resource configurator for managing the cluster and nodes, as well as configuring and managing scene resources, including the main vehicle, map, sensors, NPCs, and the management of test scenarios and test tasks. The cluster includes multiple nodes, which are divided into a main node and sub-nodes. The sub-nodes include computing nodes and rendering nodes. Multiple simulators, with one simulator deployed on each node. All simulators form a simulation cluster through a local area network connection. The simulators on each node are used to receive the simulation configurations issued by the resource configurator and load and run the actual environment of the simulation resources according to the simulation configurations.

[0034] This system is constructed using the design concept of separating the front end from the back end. In this architecture, the front end undertakes the tasks of configuring and managing scene resources, while the back end focuses on the management of data and resources, the creation of scene resources, and the operation of the simulation client. In this embodiment, the simulation system includes an autonomous driving simulator developed based on a virtual scene rendering engine, a front-end web resource configurator developed based on the React scaffolding, as well as a supporting resource server, parameter database, and web server.

[0035] In some embodiments, AB package resources of various scenarios are stored in the resource server. The parameter database uses a MongoDB database for structured management of data. The data structure is re-divided into three parts: data related to simulation resources, data related to simulation data, and data related to user management. This makes the structured management of data more convenient. Further, the data related to simulation resources includes NPCs, the main vehicle, sensors, maps, and scene descriptions; the data related to simulation data includes simulation configurations and task configurations; the data related to user management includes user information and permission management.

[0036] Based on the above simulation system, a second embodiment of the present invention provides a vehicle autonomous driving simulation method, including: On the side of the resource configurator, issue the created simulation configurations to the simulators on each cluster node. The cluster includes multiple nodes, which are divided into a main node and sub-nodes. The sub-nodes include computing nodes and rendering nodes. On the side of each cluster node simulator, receive the simulation configurations issued by the resource configurator, obtain the AB package resources and scene resource parameters according to the resource IDs in the simulation configurations. The simulator on the cluster node asynchronously loads the scene according to the AB package resources, sets and initializes the scene according to the scene resource parameters. According to the cluster configuration in the simulation configuration, through cluster simulation scheduling, achieve data communication and frame synchronization between the main node and the simulation resources of the computing nodes and rendering nodes until the simulation resources are loaded and the simulation runs.

[0037] In some embodiments, as Figure 1 shown, the method for configuring the cluster includes: Select multiple nodes; Set one master node from the multiple nodes, and the remaining nodes are all slave nodes; Allocate the slave nodes as computing nodes and rendering nodes to form a cluster.

[0038] In some embodiments, as Figure 1 shown, the method for creating the simulation configuration includes: creating a new sensor configuration according to the sensor parameters on the master vehicle configuration page, and sequentially selecting the map, master vehicle, master vehicle sensor configuration, weather environment, NPC, cluster, and bridge type on the new simulation page.

[0039] The user needs to select the map, master vehicle, and sensor configuration under the resource configurator, and needs to complete the settings of the weather environment and the bridge IP. Finally, the simulation configuration is sent to each cluster node. Specifically, as Figure 1 shown, the user needs to create a simulation cluster according to the cluster creation process, then create a new sensor configuration according to the sensor parameters on the master vehicle configuration page, and sequentially select the map, master vehicle, master vehicle sensor configuration, weather environment, NPC, simulation cluster, and bridge type (including ROS1 and ROS2) on the new simulation page to create a simulation configuration, and send it to each cluster node through the cluster configuration.

[0040] The simulator on each cluster node receives the simulation configuration and loads the simulation resources, and sets the relevant scene resources. This part of the process mainly refers to downloading the AB package resources and loading the scene resources. The specific process of loading this part of the resources is as Figure 2 shown. First, the simulator receives the simulation configuration sent by the resource configurator through the HTTP communication method. Then, according to the resource id in the simulation configuration, the simulator retrieves the corresponding AB package resource id in the parameter database. Then, the simulator checks whether the AB package resource exists in the local cache directory. If the scene resource already exists in the local cache, there is no need to download it again, and the AB package resource is directly loaded from the local cache directory into the scene. If the resource does not exist in the local cache, the simulator will download the corresponding AB package resource from the resource server. After the download is completed, the simulator will asynchronously load the resource into the scene, and retrieve the relevant scene resource parameters according to the resource id in the parameter database, so as to set the parameters of the sensor, master vehicle, weather environment (including cloud density, rain density, fog density, snow density, ground humidity, and time), and NPC behavior and then initialize the scene. Finally, according to the cluster configuration settings in the simulation configuration, the simulator will realize the data communication and frame synchronization between the master node and the simulation resources of the computing node and the rendering node through the cluster simulation scheduling process.

[0041] In some embodiments, the emulator receives the simulation configuration sent by the resource configurator through HTTP communication.

[0042] In some embodiments, obtaining the AB package resources according to the resource ID in the simulation configuration includes: The emulator retrieves the corresponding AB package resource ID in the parameter database according to the resource ID in the simulation configuration; The emulator checks whether the AB package resource exists in the local cache directory; if it exists, directly load the AB package resource from the local cache directory into the scene; if not, download the corresponding AB package resource from the resource server.

[0043] In some embodiments, obtaining the scene resource parameters and setting and initializing the scene according to the scene resource parameters includes: Retrieve the relevant scene resource parameters in the parameter database according to the resource ID in the simulation configuration, so as to set the parameters of the sensor, the main vehicle, the weather environment, and the NPC behavior according to the scene resource parameters and then initialize the scene.

[0044] To improve the scene rendering efficiency in the case of multiple main vehicles and multiple sensors, the system deploys a distributed simulation environment in the way of cluster simulation, where a simulator is deployed on each node, and all simulators form a simulation cluster through local area network connection. As Figure 3 shown, the cluster simulation scheduling includes: After each node receives the simulation configuration, the master node will first enter the scene preparation stage to wait for the slave nodes to access the simulation. After the slave nodes access the simulation, the scene initialization and loading stage will start, and each node will initialize the scene and load the corresponding resources; among them, the master node is responsible for resource allocation, the computing nodes are responsible for computing tasks, and the rendering nodes are responsible for rendering tasks; Each computing node and rendering node feedbacks the child node load to the master node through the Websocket communication method; After the master node receives the loads of each child node, it dynamically allocates resources according to the loads of each child node; After the master node receives the calculation end flags of all computing nodes and the rendering end flags of all rendering nodes, after all tasks are completed, it reallocates the calculation and rendering tasks according to the loads of each child node, so as to ensure frame synchronization among nodes. Ensure that each node is performing the calculation and rendering tasks at the current moment of rendering, and the clock cannot be synchronized due to some nodes calculating and rendering too fast or too slow.

[0045] In some embodiments, as Figure 4 shown, dynamically allocating resources according to the loads of each child node includes: First, decompose the sensor tasks into computing tasks and rendering tasks; then allocate computing resources according to the loads of all computing nodes, and allocate rendering resources according to the loads of all rendering nodes. Among them, the computing resource allocation includes: allocating the computing tasks of heavy computing sensors (such as lidar) to computing nodes with lower computing loads, and allocating the computing tasks of light computing sensors (such as IMU, GPS) to computing nodes with higher computing loads. Among them, the rendering resource allocation includes: allocating the rendering tasks of heavy rendering sensors (such as lidar, camera) to rendering nodes with lower rendering loads, and allocating the rendering tasks of light rendering sensors (such as IMU, GPS) to rendering nodes with higher rendering loads.

[0046] Each cluster has only one master node, which can run multiple master vehicle models and can centrally manage the computing and rendering tasks of all master vehicle sensors, and dynamically allocate the sensor computing and rendering tasks according to the loads of the rendering nodes and computing nodes. Such a resource allocation strategy helps to achieve load balancing management among nodes and accelerate scene rendering.

[0047] In this cluster simulation architecture, the master node, as the center of the entire simulation system, undertakes the important task of coordinating and managing the entire simulation process. It not only has to be responsible for resource allocation, but also monitor the progress and quality of the simulation to ensure the smooth progress of the simulation. The master node, through efficient algorithms and strategies, intelligently adjusts resource allocation according to the load information fed back in real time to adapt to various changes that may occur during the simulation process. This dynamic adjustment ability enables the simulation system to more flexibly respond to complex and changeable simulation requirements.

[0048] In addition, the rendering nodes and computing nodes in the cluster simulation environment described in this patent each undertake different tasks. The rendering nodes focus on the visualization of the scene and sensors to ensure the realism and details of the scene, while the computing nodes handle complex computing tasks such as physical simulation and sensor data processing. Through the real-time communication protocol of Websocket, the information exchange between nodes becomes rapid and reliable, thus ensuring the real-time and accuracy of the simulation.

[0049] In the case of multiple master vehicles, the dynamic resource allocation ability of the master node is particularly important. It can dynamically adjust resource allocation according to the behavior and state of each master vehicle and the related sensor data to ensure that each master vehicle and its sensors can obtain reasonable computing resources. This refined resource management not only improves the efficiency of the simulation, but also enhances the accuracy and reliability of the simulation results.

[0050] In summary, through distributed deployment and dynamic resource management, the cluster simulation method has greatly improved the scene rendering efficiency and simulation quality in scenarios with multiple master vehicles and multiple sensors. The application of this simulation architecture provides a powerful and flexible platform for the research and development and testing of the autonomous driving technology of construction machinery vehicles, enabling complex simulation tasks to be completed efficiently and accurately. With the continuous progress and optimization of technology, the cluster simulation method will play an increasingly important role in the field of autonomous driving of mining vehicles.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent substitutions can still be made to the specific embodiments of the present invention. Any modification or equivalent substitution that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A vehicle autonomous driving simulation system, characterized in that, The distributed simulation environment is deployed by means of cluster simulation. The simulation system includes: A resource configurator for managing clusters and nodes, as well as configuring and managing scenario resources. The cluster includes multiple nodes, which are divided into a master node and slave nodes. The slave nodes include computing nodes and rendering nodes. Multiple simulators, with one simulator deployed on each node. All simulators form a simulation cluster through a local area network connection. The simulators on each node are used to receive the simulation configurations issued by the resource configurator and load and run the actual environment of the simulation resources according to the simulation configurations.

2. The vehicle autonomous driving simulation system according to claim 1, wherein, It also includes: A parameter database, using a MongoDB database for structured management of data. The data structure is re-divided into three parts: data related to simulation resources, data related to simulation data, and data related to user management.

3. The vehicle automatic driving simulation system according to claim 2, wherein The data related to simulation resources includes NPCs, host vehicles, sensors, maps, and scene descriptions. The data related to simulation data includes simulation configurations and task configurations. The data related to user management includes user information and permission management.

4. A vehicle autonomous driving simulation method, characterized in that, Implemented by using the vehicle autonomous driving simulation system according to any one of claims 1 to 3. The method includes: On the side of the resource configurator, the created simulation configurations are issued to the simulators on each cluster node. The cluster includes multiple nodes, which are divided into a master node and slave nodes. The slave nodes include computing nodes and rendering nodes. On the side of each cluster node simulator, receive the simulation configurations issued by the resource configurator, obtain the AB package resources and scene resource parameters according to the resource IDs in the simulation configurations. The simulator on the cluster node asynchronously loads the scene according to the AB package resources, sets and initializes the scene according to the scene resource parameters. According to the cluster configuration in the simulation configuration, data communication and frame synchronization are achieved among the simulation resources of the master node, computing nodes, and rendering nodes through cluster simulation scheduling until the simulation resource loading is completed and the simulation runs.

5. The vehicle autonomous driving simulation method according to claim 4, wherein The configuration method of the cluster includes: Select multiple nodes; Set one master node from the multiple nodes, and the remaining nodes are all slave nodes; Allocate the slave nodes as computing nodes and rendering nodes to form a cluster.

6. The vehicle autonomous driving simulation method according to claim 4, wherein The creation method of the simulation configuration includes: creating a new sensor configuration according to the sensor parameters on the host vehicle configuration page, and sequentially selecting the map, host vehicle, host vehicle sensor configuration, weather environment, NPC, cluster, and bridge type on the new simulation page.

7. The vehicle autonomous driving simulation method according to claim 4, wherein The simulator receives the simulation configurations issued by the resource configurator through the HTTP communication method.

8. The vehicle automatic driving simulation method according to claim 4, wherein Obtaining the AB package resources according to the resource IDs in the simulation configurations includes: The simulator retrieves the corresponding AB package resource ID in the parameter database according to the resource ID in the simulation configuration; The simulator checks whether there is such an AB package resource in the local cache directory. If so, directly load the AB package resource from the local cache directory into the scene. If not, download the corresponding AB package resource from the resource server.

9. The vehicle autonomous driving simulation method according to claim 4, characterized in that Obtaining the scene resource parameters, setting and initializing the scene according to the scene resource parameters, includes: Retrieve relevant scenario resource parameters in the parameter database according to the resource ID in the simulation configuration, and then set the parameters of the sensors, the host vehicle, the weather environment, and the NPC behavior according to the scenario resource parameters, and then initialize the scenario.

10. The vehicle autonomous driving simulation method according to claim 4, characterized in that, The cluster simulation scheduling includes: After each node receives the simulation configuration, the master node will first enter the scenario preparation stage and wait for the slave nodes to access the simulation. After the slave nodes access the simulation, the scenario initialization and loading stage will start, and each node will initialize the scenario and load the corresponding resources; among them, the master node is responsible for resource allocation, the computing nodes are responsible for computing tasks, and the rendering nodes are responsible for rendering tasks; Each computing node and rendering node feedback the child node load to the master node through the Websocket communication method; After the master node receives the loads of each child node, it performs dynamic resource allocation according to the loads of each child node; After the master node receives the calculation end flags of all computing nodes and the rendering end flags of all rendering nodes, it reallocates the calculation and rendering tasks according to the loads of each child node, so as to ensure frame synchronization among the nodes.

11. The vehicle automatic driving simulation method according to claim 10, wherein, Performing dynamic resource allocation according to the loads of each child node includes: First, decompose the sensor tasks into computing tasks and rendering tasks; then allocate computing resources according to the loads of all computing nodes, and allocate rendering resources according to the loads of all rendering nodes; Among them, the computing resource allocation includes: allocating the computing tasks of heavy computing sensors to the computing nodes with lower computing loads, and allocating the computing tasks of light computing sensors to the computing nodes with higher computing loads; Among them, the rendering resource allocation includes: allocating the rendering tasks of heavy rendering sensors to the rendering nodes with lower rendering loads, and allocating the rendering tasks of light rendering sensors to the rendering nodes with higher rendering loads.

12. The vehicle autonomous driving simulation method according to claim 11, characterized in that, Each cluster has only one master node, which can run multiple host vehicle models, and can centrally manage the computing and rendering tasks of all host vehicle sensors, and perform dynamic allocation of sensor computing and rendering tasks according to the loads of the rendering nodes and computing nodes.

Images