Space-ground integrated multi-domain-fusion high-level autonomous driving system for vehicle

Abstract

A space-ground integrated multi-domain-fusion high-level autonomous driving system for a vehicle, which achieves more comprehensive and blind-zone-free sensing and positioning of an autonomous vehicle by means of multi-domain information fusion of a satellite platform, a cloud platform, a vehicle end platform, and the like, thereby promoting a comprehensive upgrade of comprehensive efficiency of an intelligent transportation system. On the basis of cloud error broadcasting technology and high and low orbit satellite fusion enhancement technology, the problem of vehicle communication loss caused by single ground communication being limited by base station coverage is solved; on the basis of a cloud reinforcement learning architecture, the communication network stability of a space-ground integrated system is improved by means of adaptive switching of a communication network; a road semantic map and a decision control experience pool are established on the basis of high-precision vehicle trajectory data collected by a platform, and the limitation of the sensing capability of a single vehicle is compensated by means of cloud federal learning and scene experience sharing; a vehicle end dynamically updates a cloud road semantic map by means of local map matching; and on the basis of high-precision positioning, information support is provided for implementation of chassis multi-domain fusion technology.

Description

Advanced autonomous driving system integrating space, air, and ground domains Technical Field

[0001] This invention belongs to the field of vehicle engineering and transportation engineering, and relates to a high-level autonomous driving system for automobiles that integrates space, ground, and multi-domain technologies. Background Technology

[0002] With the rapid development of vehicle-to-everything (V2X) communication technology, information interconnection has been achieved between vehicles, roads, the cloud, and satellites. The key issue is how to effectively utilize this integrated vehicle-road-satellite-cloud communication network to build a seamless information interaction system, improve the safety and comfort of high-level autonomous driving, and promote a comprehensive upgrade of the overall efficiency of intelligent transportation systems.

[0003] Traditional single-mode ground communication suffers from reliability challenges such as weak signals and limited network coverage, failing to meet the high real-time and safety requirements of autonomous driving and limiting its widespread application. For example, cellular vehicle-to-everything (CV2X)-based autonomous driving systems experience data transmission delays or interruptions in remote areas, urban fringes, and areas obstructed by buildings due to insufficient cellular network coverage. This makes them unable to handle complex dynamic driving scenarios, reducing system adaptability and reliability. Furthermore, CV2X-based autonomous driving systems lack multi-domain information integration capabilities, hindering the full realization of multi-domain information collaboration from perception and decision-making to control. This results in information silos due to insufficient information sharing, preventing rapid and comprehensive responses to changes in driving scenarios. In summary, existing systems have limitations in scenario adaptability and multi-domain information integration, restricting their further development.

[0004] The integrated space-ground autonomous driving system combines terrestrial and satellite communication networks to achieve highly reliable and seamless coverage of the communication network. Through satellite signals, it provides accurate time synchronization and reliable communication transmission between multiple domains, and provides high-precision positioning as the basic data support for the integration of multiple domains such as intelligent driving, chassis, cockpit, and power. It also provides information support such as traffic, weather, and road conditions, providing a new solution for autonomous driving in complex traffic scenarios. Summary of the Invention

[0005] This invention provides a high-level autonomous driving system that integrates multiple domains, including satellite, cloud, and vehicle platforms. By fusing information from different data sources, it enables autonomous vehicles to achieve more comprehensive and blind-spot-free perception and positioning, improves the safety and reliability of planning and control, and promotes a comprehensive upgrade of the overall efficiency of intelligent transportation systems. Based on cloud-based error broadcasting and high / low orbit satellite fusion enhancement technologies, the system solves the problem of vehicle communication disconnection caused by the limited coverage of base stations in single-terrestrial communication, ensuring the high-precision positioning and high-security communication requirements of autonomous driving. Based on a cloud-based reinforcement learning architecture, the system improves the stability of the integrated space-ground communication network through adaptive switching. Road semantic maps and decision-making control experience pools are built based on high-precision vehicle trajectory data collected by the platform. Through cloud-based federated learning and scenario experience sharing, the system compensates for the limitations of single-vehicle perception capabilities, solving redundant safety and long-tail problems that single-vehicle intelligence cannot address. The vehicle-side system dynamically updates the cloud-based road semantic map by matching the semantic map provided by the cloud with the local driving map built by the vehicle, improving the cross-scenario integration capability of the entire autonomous driving system. Based on high-precision positioning, the system provides vehicles with traffic, weather, and road condition information for driving scenarios, supporting the implementation of chassis multi-domain fusion technology.

[0006] This invention presents a high-level autonomous driving system integrating space, ground, and multi-domain technologies, comprising three main components and eight domains. The three components are the vehicle-side, cloud-side, and satellite-side. The eight domains are the intelligent cockpit domain, intelligent driving domain, powertrain domain, intelligent chassis domain, knowledge management domain, cloud control domain, satellite service domain, and information communication domain. Specifically, the intelligent cockpit, intelligent driving, powertrain, and intelligent chassis domains are deployed on the vehicle-side; the cloud control and knowledge management domains are deployed on the cloud; the satellite service domain is deployed on high- and low-Earth orbit satellites; and the information communication domain serves the communication between the vehicle-side, cloud-side, and satellite-side.

[0007] For the satellite end, it consists of a satellite service domain and related support platforms. Through satellite signals, it provides accurate time synchronization and reliable communication transmission between multiple domains, offers high-precision positioning as fundamental data support for multi-domain fusion, and provides additional data support such as traffic information, weather changes, and road condition updates. The satellite service domain provides high-precision positioning information via satellite communication. The related support platforms include, but are not limited to, a traffic management platform that monitors traffic information in real time and provides a complete traffic network; a logistics platform that provides real-time logistics network information; a map platform that provides navigation services and regularly updates offline map information; a positioning platform that provides high-precision positioning services and location information sharing functions; and a meteorological platform that provides real-time weather monitoring and analysis.

[0008] The cloud is primarily used for information transmission and further processing, enabling centralized information management and knowledge extraction through various cloud control applications. The cloud consists of a cloud control domain and a knowledge management domain, achieving efficient information exchange between the vehicle, the cloud, and the satellite terminals through Vehicle-to-Network (V2N) communication combined with satellite communication.

[0009] The cloud control domain, based on three cloud control levels—edge cloud, regional cloud, and central cloud—enables various cloud control applications through three types of cloud control empowerment: connected vehicle empowerment, traffic management and control, and traffic data empowerment. These cloud control applications include communication network monitoring applications, communication convergence enhancement applications, network switching strategies, blind-spot-free high-precision bird's-eye view maps, semantic bird's-eye view matrices, and driving safety field construction.

[0010] The communication network monitoring application is primarily used to continuously monitor the status of the communication network, including signal strength, latency, bandwidth, and packet loss. When a network failure or performance degradation occurs, the communication network monitoring application can promptly detect it and issue an alarm, allowing for subsequent network switching to resolve the issue.

[0011] The aforementioned communication fusion enhancement application refers to the process of sending phase deviations, which characterize satellite clock errors, satellite orbital errors, and phase delay errors, to high-orbit satellites via a Satellite Ground Reference Station (SGRS). The high-orbit satellites then calculate satellite ephemeris errors using the phase deviation data and send them to the corresponding low-orbit satellite set. Finally, the low-orbit satellites return navigation enhancement information to the Satellite Ground Station (SGS), thereby offsetting the orbital errors from the satellites during the satellite positioning process.

[0012] The network handover strategy is trained using deep reinforcement learning to switch between satellite and cellular networks. The deep reinforcement learning is described by a tuple (S, P, A, R), where:

[0013] S represents the input state variables for deep reinforcement learning:

[0014] S = [s cell ,s sat ,d cell ,d sat ,b cell ,b sat ,p lat ,p long ,c,h]

[0015] Among them, s cell Indicates the current cellular network signal strength, s satd represents the current satellite network signal strength. cell Indicates the current cellular network latency, d sat Indicates the current satellite network latency, b cell Indicates the current cellular network bandwidth, b sat p represents the current satellite network bandwidth. lat p represents the latitude of the current vehicle. long The current longitude of the vehicle is represented by c, the current network connection status is represented by c=0 (currently connected to a cellular network) and c=1 (currently connected to a satellite network), and h represents the communication quality history over a recent period, expressed through the communication network delay.

[0016] P represents the state transition equation for deep reinforcement learning, which is fitted using a neural network in this invention.

[0017] A represents the output of the deep reinforcement learning communication network switching model. A=0 means maintaining the current communication network connection, A=1 means switching to a cellular network, and A=2 means switching to a satellite network.

[0018] R represents the reward function (environmental feedback) after executing the communication network switching strategy:

[0019] R t =w1R stab +w2R cost +w3R quality

[0020] Among them, R t This represents the environmental feedback at time t, where w1, w2, and w3 represent the weighted weights of each sub-reward function, and R... stab R represents the reward function related to connection stability. cost R represents the reward function related to switching costs. quality This represents the reward function related to communication service quality. Table 1 details the values ​​of each sub-reward function, where Δd represents the latency difference after implementing switching strategy c, and θ... d θ represents the latency threshold, Δb represents the bandwidth difference after implementing switching strategy c, and θ represents the bandwidth difference after implementing switching strategy c. b This indicates the bandwidth threshold.

[0021] Table 1

[0022] The blind-spot-free high-precision bird's-eye view map is constructed by receiving high-precision positioning data from the satellite service domain deployed on the satellite via a Satellite Ground Station (SGS) responsible for receiving and transmitting satellite signals. Based on vehicle trajectory data collected by the platform under this high-precision positioning, a road semantic map is created. The blind-spot-free high-precision bird's-eye view map contains four channels of information: the drivable road area, lane boundaries, dynamic obstacles (vehicles, pedestrians), and traffic infrastructure (traffic lights, traffic signs).

[0023] The semantic bird's-eye view matrix divides the blind-spot-free high-precision bird's-eye view into matrices with different semantic meanings according to different channels. The size of the blind-spot-free semantic bird's-eye view matrix is ​​[0,1]. W×H×C , where W represents the width of the blind-spot-free semantic bird's-eye view matrix, H represents the height of the blind-spot-free semantic bird's-eye view matrix, and C represents the number of channels of the blind-spot-free semantic bird's-eye view matrix.

[0024] The driving safety field, based on a high-precision bird's-eye view with no blind spots, is constructed to guide vehicle driving behavior. The driving safety field is modeled using Gaussian equations, and by describing the dynamic transfer process of the risk center in the driving safety field during the movement of intelligent connected vehicles, it assesses the driving risks of intelligent connected vehicles in their environment.

[0025] The knowledge management domain acquires neural network parameters of intelligent connected vehicles within the vehicle-to-network (V2N) group through vehicle-to-network (V2N) communication, and aggregates these parameters using feedback from the vehicle's intelligent chassis for local neural network Actor-Critic parameter aggregation. The knowledge management domain further trains a scenario causal inference model in cross-scenario training sequences, extracts transferable scenario factors for each scenario, and establishes a scenario experience database. When a new vehicle connects to the system, the knowledge management domain pushes a beyond-line-of-sight road semantic map of the intelligent connected vehicle's driving environment, predicts scenario factors of the vehicle's environment based on the scenario causal inference model, and pushes parameter experience to the vehicle based on the neural network parameters corresponding to the scenario factors in the scenario experience database, enabling rapid vehicle access.

[0026] For the vehicle side, there are multiple vehicle groups, each containing N intelligent connected vehicles. Each intelligent connected vehicle includes an intelligent cockpit domain, an intelligent driving domain, a powertrain domain, and an intelligent chassis domain.

[0027] The intelligent cockpit domain, primarily designed to enhance driving safety, improve user experience, and provide personalized services, includes an information system, an interaction system, and a connectivity system. The information system displays real-time vehicle information and entertainment content on a high-resolution screen, providing the driver with a comprehensive understanding of the vehicle's status. The interaction system enables comprehensive information exchange between the vehicle and the driver through an intuitive user interface and voice recognition. The connectivity system utilizes vehicle-to-everything (V2X) technology to achieve seamless connection between the vehicle and external networks.

[0028] The intelligent driving domain includes a perception processing module, a trajectory prediction module, and a reinforcement learning module. The perception processing module acquires a blind-spot-free semantic bird's-eye view provided by the cloud control domain via Vehicle-to-Network (V2N) communication, and then crops it based on high-precision positioning provided by the satellite service domain. The cropped blind-spot-free semantic bird's-eye view matrix has a size of [0,1]. w×h×c Where w represents the width of the cropped blind-spot-free semantic bird's-eye view matrix, h represents the height of the cropped blind-spot-free semantic bird's-eye view matrix, and c represents the number of channels of the cropped blind-spot-free semantic bird's-eye view matrix. The cropped blind-spot-free semantic bird's-eye view is stacked with sensor information and used as part of the input to the reinforcement learning module. Simultaneously, the perception processing module uses vehicle-side sensors to construct and upload a real-time dynamic map, achieving real-time local updates of the offline map through dynamic matching based on high-precision positioning. The trajectory prediction module stacks the blind-spot-free semantic bird's-eye view provided by the cloud control domain over multiple time steps and combines it with scene factors provided by the knowledge management domain to achieve multi-time-step trajectory prediction of dynamic obstacles around the intelligent connected vehicle, serving as another part of the input to the reinforcement learning module. The reinforcement learning module stacks the input provided by the perception processing module over multiple time steps and integrates it with the multi-time-step trajectory prediction results of surrounding dynamic obstacles provided by the trajectory prediction module into a complete reinforcement learning input. Furthermore, the reinforcement learning module constructs a fusion reward function based on the driving safety field provided by the cloud control domain to guide the training process of the vehicle-side reinforcement learning neural network. Finally, the reinforcement learning module outputs control commands to the powertrain domain.

[0029] The powertrain domain primarily outputs power control based on control commands from the intelligent driving domain. Through the coordinated action of the battery unit, motor unit, and electronic control unit, it enhances the overall performance and efficiency of the powertrain domain, ensuring the electric vehicle's power performance, energy management, and safety during driving. The battery unit, with its high-energy-density battery pack, provides a continuous and robust power supply, achieving long driving range and high efficiency. The motor unit, with its high-efficiency motor and advanced rotor design, provides rapid acceleration performance, resulting in a responsive driving experience. The electronic control unit, through advanced control algorithms, achieves intelligent coordination between the motor and battery, optimizing power output and energy management.

[0030] The intelligent chassis domain primarily optimizes vehicle driving comfort through an intelligent chassis system. Based on high-precision positioning, it provides the vehicle with traffic, weather, and road condition information for the driving scenario, supporting the implementation of multi-domain chassis fusion technology. The intelligent chassis system includes, but is not limited to, a steering subsystem, a direct yaw control (DYC) subsystem, a suspension subsystem, and a drive subsystem, which are coupled in terms of state and cost. The steering subsystem and the DYC subsystem calculate the lateral control quantities of the intelligent connected vehicle, the suspension subsystem calculates the vertical control quantities, and the drive subsystem calculates the longitudinal control quantities. Through information transmission, the intelligent chassis domain connects to three major domains: the intelligent cockpit domain, the intelligent driving domain, and the powertrain domain. First, through interaction within the intelligent cockpit domain, the intelligent chassis domain dynamically adjusts, achieving interaction between the intelligent cockpit domain and the intelligent chassis domain. Second, this invention incorporates chassis feedback as part of the state input to the intelligent driving domain, enabling interaction between the intelligent driving domain and the intelligent chassis domain. Finally, the steering subsystem and DYC subsystem of the intelligent chassis domain calculate the yaw moment of the intelligent connected vehicle based on the output of the powertrain domain and the pre-aiming error between the actual path and the desired path. The drive subsystem obtains the desired target speed and calculates the four-wheel drive or braking torque of the intelligent connected vehicle based on the output of the powertrain domain and the current speed of the intelligent connected vehicle, thus realizing the interaction between the powertrain domain and the intelligent chassis domain.

[0031] The information and communication domain serves the communication between the vehicle, the cloud, and the satellite. On the satellite platform, it provides inter-satellite communication, information monitoring, high-precision positioning, and time synchronization services. On the cloud platform, it provides information storage, data analysis, experience summarization, and communication enhancement services. On the vehicle platform, it provides scene recognition, information sharing, experience reception, and network switching services.

[0032] The implementation process of the high-level autonomous driving system integrating space, air, and ground domains according to this invention includes the following steps:

[0033] Step 1: Establish a satellite platform to provide accurate time synchronization and reliable communication transmission between multiple domains via satellite signals. This provides high-precision positioning as the foundational data support for multi-domain fusion and additional data support such as traffic information, weather changes, and road condition updates. Deploy satellite service domains on the satellite platform; these domains provide high-precision positioning information through satellite communication.

[0034] Step 2: Build a cloud platform and deploy the cloud control domain and knowledge management domain on it. Centralized information management and knowledge extraction are achieved through different cloud control applications. The cloud platform consists of the cloud control domain and the knowledge management domain, and efficient information exchange between the vehicle, the cloud, and satellite is achieved through Vehicle-to-Network (V2N) communication combined with satellite communication.

[0035] Step 3: Build the vehicle-side platform and deploy the intelligent cockpit domain, intelligent driving domain, powertrain domain, and intelligent chassis domain on the vehicle-side platform. The vehicle-side platform includes multiple vehicle-side groups, each containing N intelligent connected vehicles, and each intelligent connected vehicle includes the intelligent cockpit domain, intelligent driving domain, powertrain domain, and intelligent chassis domain.

[0036] Step 4: Establish an information and communication domain to serve the vehicle, cloud, and satellite terminals. The information and communication domain on the satellite platform provides inter-satellite communication, information monitoring, high-precision positioning, and time synchronization services. On the cloud platform, it provides information storage, data analysis, experience summarization, and communication enhancement services. On the vehicle platform, it provides scene recognition, information sharing, experience reception, and network switching services.

[0037] Step 5: The intelligent connected vehicle group on the vehicle platform integrates complete reinforcement learning state inputs based on the perception processing module and trajectory prediction module in the intelligent driving domain.

[0038] Step 6: The intelligent connected vehicle group on the vehicle platform further constructs a fusion reward function based on the driving safety field provided by the cloud control domain to guide the training process of the intelligent connected vehicle reinforcement learning model.

[0039] Step 7: During reinforcement learning training, the intelligent connected vehicle on the vehicle platform uploads its neural network Actor-Critic parameters to the knowledge management domain of the cloud platform. In different scenarios where the intelligent connected vehicle operates, the knowledge management domain selects neural network Actor-Critic parameters for parameter aggregation based on the intelligent chassis domain, and aggregates the local neural network Actor-Critic parameters based on federated learning. Furthermore, the cross-scenario training sequences of the intelligent connected vehicle's neural network are stored in the knowledge management domain. Based on these cross-scenario sequences, a scenario causal reasoning model is trained, scene factors corresponding to different scenarios are extracted, and a scenario experience database is established.

[0040] Step 8: The knowledge management domain of the cloud platform sends the aggregated local neural network Actor-Critic parameters to the corresponding intelligent connected vehicles on the vehicle platform, repeating this process until the network converges. When a new intelligent connected vehicle connects to the integrated space-ground multi-domain high-level autonomous driving system on the vehicle platform, the knowledge management domain of the cloud platform predicts the scene factors of the scene based on the observation sequence of the intelligent connected vehicle, and retrieves the corresponding scene experience from the scene experience library to push the experience based on federated learning, enabling the intelligent connected vehicle to quickly connect to the system.

[0041] Preferably, in step 1, the relevant support platform includes, but is not limited to, a traffic management platform that monitors traffic information in real time and provides a complete traffic network, a logistics platform that provides real-time logistics network information, a map platform that provides navigation services and regularly updates offline map information, a positioning platform that provides high-precision positioning services and location information sharing functions, and a meteorological platform that provides real-time weather monitoring and analysis.

[0042] Preferably, in step 2, the cloud control domain, based on three cloud control levels—edge cloud, regional cloud, and central cloud—achieves different cloud control applications through three types of cloud control empowerment: connected vehicle empowerment, traffic management and control, and traffic data empowerment. These cloud control applications include communication network monitoring applications, communication convergence enhancement applications, providing network switching strategies, providing blind-spot-free high-precision bird's-eye view maps, providing semantic bird's-eye view matrices, and constructing driving safety scenarios.

[0043] Preferably, the communication network monitoring application is mainly used to continuously monitor the status of the communication network, including signal strength, latency, bandwidth, and packet loss. When a network failure or performance degradation occurs, the communication network monitoring application can promptly detect it and issue an alarm, so that the problem can be resolved through subsequent network switching applications.

[0044] Preferably, the communication fusion enhancement application refers to sending phase deviation, which characterizes satellite clock error, satellite orbit error, and phase delay error, to high-orbit satellites via a Satellite Ground Reference Station (SGRS). The high-orbit satellites further calculate satellite ephemeris errors using the phase deviation data and send them to the corresponding low-orbit satellite set. Finally, the low-orbit satellites return navigation enhancement information to the Satellite Ground Station (SGS), thereby offsetting the orbit errors from the satellites during satellite positioning through the navigation enhancement information.

[0045] Preferably, the network switching strategy is trained using deep reinforcement learning to train the switching strategies for satellite and cellular networks. The deep reinforcement learning is described by a tuple (S, P, A, R), where:

[0046] S represents the input state variables for deep reinforcement learning:

[0047] S = [s cell ,s sat ,d cell ,d sat ,b cell ,b sat ,p lat ,p long ,c,h]

[0048] Among them, s cell Indicates the current cellular network signal strength, s sat d represents the current satellite network signal strength. cell Indicates the current cellular network latency, d sat Indicates the current satellite network latency, b cell Indicates the current cellular network bandwidth, b sat p represents the current satellite network bandwidth. lat p represents the latitude of the current vehicle. long The current longitude of the vehicle is represented by c, the current network connection status is represented by c=0 (currently connected to a cellular network) and c=1 (currently connected to a satellite network), and h represents the communication quality history over a recent period, expressed through the communication network delay.

[0049] P represents the state transition equation for deep reinforcement learning, which is fitted using a neural network in this invention.

[0050] A represents the output of the deep reinforcement learning communication network switching model. A=0 means maintaining the current communication network connection, A=1 means switching to a cellular network, and A=2 means switching to a satellite network.

[0051] R represents the reward function (environmental feedback) after executing the communication network switching strategy:

[0052] R t =w1R stab +w2R cost +w3R quality

[0053] Among them, R t This represents the environmental feedback at time t, where w1, w2, and w3 represent the weighted weights of each sub-reward function, and R... stab R represents the reward function related to connection stability. cost R represents the reward function related to switching costs. quality This represents the reward function related to communication service quality. Table 1 details the values ​​of each sub-reward function, where Δd represents the latency difference after implementing switching strategy c, and θ... dθ represents the latency threshold, Δb represents the bandwidth difference after implementing switching strategy c, and θ represents the bandwidth difference after implementing switching strategy c. b This indicates the bandwidth threshold.

[0054] Table 1

[0055] Preferably, the blind-spot-free high-precision bird's-eye view is constructed by receiving high-precision positioning data from the satellite service domain deployed on the satellite via a Satellite Ground Station (SGS) responsible for receiving and transmitting satellite signals. A road semantic map is then built based on vehicle trajectory data collected by the platform under this high-precision positioning. The blind-spot-free high-precision bird's-eye view includes four channels of information: the drivable road area, lane boundaries, dynamic obstacles (vehicles, pedestrians), and traffic infrastructure (traffic lights, traffic signs).

[0056] Preferably, the semantic bird's-eye view matrix divides the blind-spot-free high-precision bird's-eye view into matrices with different semantic meanings according to different channels. The size of the blind-spot-free semantic bird's-eye view matrix is ​​[0,1]. W×H×C , where W represents the width of the blind-spot-free semantic bird's-eye view matrix, H represents the height of the blind-spot-free semantic bird's-eye view matrix, and C represents the number of channels of the blind-spot-free semantic bird's-eye view matrix.

[0057] Preferably, the driving safety field is constructed based on a high-precision bird's-eye view with no blind spots, and a dynamic driving safety field is built to guide vehicle driving behavior. The driving safety field is modeled using Gaussian equations, and by describing the dynamic transfer process of the driving safety field risk center during the movement of the intelligent connected vehicle, the driving risk of the intelligent connected vehicle in its environment is assessed.

[0058] Preferably, the knowledge management domain acquires the neural network parameters of intelligent connected vehicles in the vehicle group through Vehicle-to-Network (V2N) communication, and performs local neural network Actor-Critic parameter aggregation by filtering neural network parameters through intelligent chassis feedback on the vehicle. The knowledge management domain further trains a scenario causal inference model in cross-scenario training sequences, extracts transferable scenario factors for each scenario, and establishes a scenario experience library. When a new vehicle connects to the system, the knowledge management domain pushes a beyond-line-of-sight road semantic map of the intelligent connected vehicle's driving environment, predicts scenario factors of the vehicle's environment based on the scenario causal inference model, and pushes parameter experience to the vehicle based on the neural network parameters corresponding to the scenario factors in the scenario experience library, enabling rapid vehicle access.

[0059] Preferably, in step 3, the intelligent cockpit domain is primarily used to improve driving safety, enhance user experience, and provide personalized services, and includes an information system, an interaction system, and a connectivity system. The information system displays vehicle information and entertainment content in real time through a high-resolution screen, enabling the driver to have a comprehensive understanding of the vehicle's status. The interaction system achieves comprehensive information interaction between the vehicle and the driver through an intuitive user interface and voice recognition. The connectivity system achieves seamless connection between the vehicle and external networks through vehicle-to-everything (V2X) technology.

[0060] Preferably, in step 3, the intelligent driving domain includes a perception processing module, a trajectory prediction module, and a reinforcement learning module. The perception processing module acquires a blind-spot-free semantic bird's-eye view provided by the cloud control domain via Vehicle-to-Network (V2N) communication and crops it based on high-precision positioning provided by the satellite service domain. The trajectory prediction module stacks the blind-spot-free semantic bird's-eye view provided by the cloud control domain across multiple time steps and combines it with scene factors provided by the knowledge management domain to achieve multi-time-step trajectory prediction of dynamic obstacles around the intelligent connected vehicle, serving as another input to the reinforcement learning module. The reinforcement learning module stacks the multi-time-step input provided by the perception processing module and integrates it with the multi-time-step trajectory prediction results of surrounding dynamic obstacles provided by the trajectory prediction module into a complete reinforcement learning input. Further, the reinforcement learning module constructs a fusion reward function based on the driving safety field provided by the cloud control domain to guide the training process of the vehicle-side reinforcement learning neural network. Finally, the reinforcement learning module outputs control commands to the powertrain domain.

[0061] Preferably, in step 3, the powertrain domain primarily outputs power control based on control commands from the intelligent driving domain. Through the synergistic effect of the battery unit, motor unit, and electronic control unit, the overall performance and efficiency of the powertrain domain are improved, ensuring the electric vehicle's power performance, energy management, and safety during driving. The battery unit, with its high-energy-density battery pack, provides a continuous and robust power supply, achieving long driving range and high efficiency. The motor unit, with its high-efficiency motor and advanced rotor design, provides rapid acceleration performance, achieving a responsive driving experience. The electronic control unit, through advanced control algorithms, achieves intelligent coordination between the motor and battery, optimizing power output and energy management.

[0062] Preferably, in step 3, the intelligent chassis domain primarily optimizes vehicle driving comfort through the intelligent chassis system and provides the vehicle with traffic, weather, and road condition information based on high-precision positioning, supporting the realization of chassis multi-domain fusion technology. The intelligent chassis system includes, but is not limited to, a steering subsystem, a direct yaw control (DYC) subsystem, a suspension subsystem, and a drive subsystem, which are coupled in both state and cost. The steering subsystem and the DYC subsystem calculate the lateral control quantities of the intelligent connected vehicle, the suspension subsystem calculates the vertical control quantities, and the drive subsystem calculates the longitudinal control quantities. Through information transmission, the intelligent chassis domain connects to the intelligent cockpit domain, the intelligent driving domain, and the powertrain domain. First, through interaction in the intelligent cockpit domain, the intelligent chassis domain dynamically adjusts, achieving interaction between the intelligent cockpit domain and the intelligent chassis domain. Second, this invention uses chassis feedback as part of the state input of the intelligent driving domain, realizing interaction between the intelligent driving domain and the intelligent chassis domain. Finally, the steering subsystem and DYC subsystem of the intelligent chassis domain calculate the yaw moment of the intelligent connected vehicle based on the output of the powertrain domain and the pre-aiming error between the actual path and the desired path. The drive subsystem obtains the desired target speed and calculates the four-wheel drive / braking torque of the intelligent connected vehicle based on the output of the powertrain domain and the current speed of the intelligent connected vehicle, thus realizing the interaction between the powertrain domain and the intelligent chassis domain.

[0063] Preferably, in step 5, the integrated and complete reinforcement learning state input includes two parts. The first part involves the perception processing module acquiring a blind-spot-free semantic bird's-eye view provided by the cloud control domain via Vehicle-to-Network (V2N) communication, and cropping it based on high-precision positioning provided by the satellite service domain. The cropped blind-spot-free semantic bird's-eye view matrix has a size of [0,1]. w×h×c Where w represents the width of the cropped blind-spot-free semantic bird's-eye view matrix, h represents the height of the cropped blind-spot-free semantic bird's-eye view matrix, and c represents the number of channels of the cropped blind-spot-free semantic bird's-eye view matrix. The cropped blind-spot-free semantic bird's-eye view is stacked with sensor information and used as part of the input to the reinforcement learning module. In the second part, the trajectory prediction module stacks the blind-spot-free semantic bird's-eye view provided by the cloud control domain over multiple time steps, and combines it with scene factors provided by the knowledge management domain to realize multi-time-step trajectory prediction of dynamic obstacles around the intelligent connected vehicle, which serves as another part of the input to the reinforcement learning module. The reinforcement learning module stacks the input provided by the perception processing module over multiple time steps and integrates it with the multi-time-step trajectory prediction results of the surrounding dynamic obstacles provided by the trajectory prediction module into a complete reinforcement learning input.

[0064] Preferably, in step 6, the fusion reward function considers both driving safety and driving expectation. The driving safety-related reward function is constructed based on the driving risks of the intelligent connected vehicle predicted by the driving safety location provided by the cloud control domain. The driving expectation reward function is constructed based on the lateral error and heading angle deviation between the actual path and the expected path of the intelligent connected vehicle.

[0065] The beneficial effects of this invention are:

[0066] (1) This invention provides a high-level autonomous driving system that integrates multiple domains, including satellite platform, cloud platform and vehicle platform, and integrates different data sources to achieve more comprehensive and blind-spot-free perception and positioning of autonomous vehicles, improve the safety and reliability of planning and control, and promote the comprehensive upgrade of the overall efficiency of intelligent transportation system.

[0067] (2) Based on cloud error broadcasting and high and low orbit satellite fusion enhancement technology, the problem of vehicle communication disconnection caused by the limitation of single ground communication on base station coverage is solved, ensuring the high-precision positioning and high-safety communication requirements of autonomous driving; based on cloud reinforcement learning architecture, the communication network stability of the space-ground integrated system is improved through adaptive switching of the communication network.

[0068] (3) Based on the high-precision vehicle trajectory data collected by the platform, a road semantic map and decision control experience pool are established. Through cloud-based federated learning scenario experience sharing, the limitations of single-vehicle perception capabilities are made up for, and redundant safety and long-tail problems that single-vehicle intelligence cannot solve are addressed.

[0069] (4) The vehicle-side system dynamically updates the cloud-based road semantic map by matching the semantic map provided in the cloud with the local driving map constructed by the vehicle itself. High-precision positioning can provide the vehicle with information such as traffic, weather, and road conditions in the driving scenario, improve the cross-scenario integration capability of the entire autonomous driving system, support the realization of chassis multi-domain fusion technology, and has important value for the implementation of high-level autonomous driving technology. Attached Figure Description

[0070] Figure 1 is a schematic diagram of the high-level autonomous driving system integrating space, air, and multi-domain fusion proposed in this invention;

[0071] Figure 2. Communication schematic diagram of the high-level autonomous driving system with integrated space-ground and multi-domain fusion proposed in this invention;

[0072] Figure 3. Schematic diagram of the neural network structure of the network switching strategy proposed in this invention;

[0073] Figure 4 is a schematic diagram of the blind-spot-free semantic bird's-eye view used in this invention. Detailed Implementation

[0074] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings, but the content of the present invention is not limited thereto.

[0075] This invention provides a high-level autonomous driving system integrating space, air, and ground domains, as shown in Figure 1. It enables high-level autonomous driving in complex driving scenarios based on this system. A communication diagram of the high-level autonomous driving system integrating space, air, and ground domains is shown in Figure 2. The system specifically includes the following steps:

[0076] (1) Establish a satellite platform to provide accurate time synchronization and reliable communication transmission between multiple domains via satellite signals, provide high-precision positioning as the basic data support for multi-domain fusion, and provide additional data support such as traffic information, weather changes, and road condition updates. Deploy satellite service domains on the satellite platform, which provide high-precision positioning information through satellite communication. Related supporting platforms include, but are not limited to, traffic management platforms that monitor traffic information in real time and provide a complete traffic network, logistics platforms that provide real-time logistics network information, map platforms that provide navigation services and regularly update offline map information, positioning platforms that provide high-precision positioning services and location information sharing functions, and meteorological platforms that provide real-time weather monitoring and analysis.

[0077] (2) Build a cloud platform and deploy the cloud control domain and knowledge management domain on the cloud platform. Achieve centralized information management and knowledge extraction through different cloud control applications. The cloud consists of the cloud control domain and the knowledge management domain, and achieves efficient information exchange between the vehicle, the cloud, and the satellite through Vehicle-to-Network (V2N) communication combined with satellite communication. The cloud control domain is based on three cloud control levels: edge cloud, regional cloud, and central cloud. It realizes different functions of cloud control applications through three types of cloud control empowerment: connected vehicle empowerment, traffic management and control, and traffic data empowerment. Cloud control applications include communication network monitoring applications, communication convergence enhancement applications, network switching strategies, blind-spot-free high-precision bird's-eye view, semantic bird's-eye view matrix, and driving safety field construction.

[0078] Communication network monitoring applications are primarily used to continuously monitor the status of communication networks, including signal strength, latency, bandwidth, and packet loss. When network failures or performance degradation occur, these applications can promptly detect the problem and issue alerts, allowing for subsequent network switching to resolve the issue.

[0079] Communication convergence enhancement applications refer to the process where a satellite ground reference station (SGRS) sends phase deviation data, representing satellite clock errors, orbital errors, and phase delay errors, to high-orbit satellites. The high-orbit satellites then calculate satellite ephemeris errors using the phase deviation data and send this data to the corresponding low-orbit satellite set. Finally, the low-orbit satellites return navigation enhancement information to the satellite ground station (SGS), using the navigation enhancement information to offset the orbital errors from the satellites during the satellite positioning process.

[0080] The network handover strategy, and the neural network structure used for the network handover strategy, are shown in Figure 3. The handover strategy for the satellite network and the cellular network is trained through deep reinforcement learning. The deep reinforcement learning is described by the tuple (S, P, A, R), where:

[0081] S represents the input state variables for deep reinforcement learning:

[0082] S = [s cell ,s sat ,d cell ,d sat ,b cell ,b sat ,p lat ,p long ,c,h]

[0083] Among them, s cell Indicates the current cellular network signal strength, s sat d represents the current satellite network signal strength. cell Indicates the current cellular network latency, d sat Indicates the current satellite network latency, b cell Indicates the current cellular network bandwidth, b sat p represents the current satellite network bandwidth. lat p represents the latitude of the current vehicle. long The current longitude of the vehicle is represented by c, the current network connection status is represented by c=0 (currently connected to a cellular network) and c=1 (currently connected to a satellite network), and h represents the communication quality history over a recent period, expressed through the communication network delay.

[0084] P represents the state transition equation for deep reinforcement learning, which is fitted using a neural network in this invention.

[0085] A represents the output of the deep reinforcement learning communication network switching model. A=0 means maintaining the current communication network connection, A=1 means switching to a cellular network, and A=2 means switching to a satellite network.

[0086] R represents the reward function (environmental feedback) after executing the communication network switching strategy:

[0087] R t =w1R stab +w2R cost +w3R quality

[0088] Among them, R t This represents the environmental feedback at time t, where w1, w2, and w3 represent the weighted weights of each sub-reward function, and R... stab R represents the reward function related to connection stability. cost R represents the reward function related to switching costs. quality This represents the reward function related to communication service quality. Table 1 details the values ​​of each sub-reward function, where Δd represents the latency difference after implementing switching strategy c, and θ... d θ represents the latency threshold, Δb represents the bandwidth difference after implementing switching strategy c, and θ represents the bandwidth difference after implementing switching strategy c. b This indicates the bandwidth threshold.

[0089] Table 1

[0090] As shown in Figure 4, the blind-spot-free high-precision bird's-eye view is generated by receiving high-precision positioning data from the satellite service domain deployed on the satellite via a Satellite Ground Station (SGS) responsible for receiving and transmitting satellite signals. Based on the vehicle trajectory data collected by the platform under this high-precision positioning, a road semantic map is constructed. This blind-spot-free high-precision bird's-eye view contains four channels of information: the drivable road area, lane boundaries, dynamic obstacles (vehicles, pedestrians), and traffic infrastructure (traffic lights, traffic signs).

[0091] The semantic bird's-eye view matrix divides the high-precision bird's-eye view image without blind spots into matrices with different semantic meanings according to different channels. The size of the semantic bird's-eye view matrix without blind spots is [0,1]. W×H×C , where W represents the width of the blind-spot-free semantic bird's-eye view matrix, H represents the height of the blind-spot-free semantic bird's-eye view matrix, and C represents the number of channels of the blind-spot-free semantic bird's-eye view matrix.

[0092] The driving safety field, based on a high-precision bird's-eye view with no blind spots, constructs a dynamic driving safety field to guide vehicle driving behavior. The driving safety field is modeled using Gaussian equations and assesses the driving risks of intelligent connected vehicles in their environment by describing the dynamic transfer process of the driving safety field risk center during the movement of the vehicle.

[0093] The knowledge management domain acquires neural network parameters of intelligent connected vehicles within the vehicle-to-network (V2N) cloud communication group, and aggregates these parameters using feedback from the vehicle's intelligent chassis for local neural network Actor-Critic parameter aggregation. The knowledge management domain further trains a scenario-based causal inference model in cross-scenario training sequences, extracts transferable scenario factors for each scenario, and establishes a scenario experience database. When a new vehicle connects to the system, the knowledge management domain pushes a beyond-line-of-sight road semantic map of the intelligent connected vehicle's driving environment, predicts scenario factors of the vehicle's environment based on the scenario-based causal inference model, and pushes parameter experience to the vehicle based on the neural network parameters corresponding to the scenario factors in the scenario experience database, enabling rapid vehicle integration.

[0094] (3) Build a vehicle-side platform and deploy the intelligent cockpit domain, intelligent driving domain, powertrain domain, and intelligent chassis domain on the vehicle-side platform. The vehicle-side platform includes multiple vehicle-side groups, each vehicle-side group includes N intelligent connected vehicles, and each intelligent connected vehicle includes the intelligent cockpit domain, intelligent driving domain, powertrain domain, and intelligent chassis domain.

[0095] The intelligent cockpit domain, primarily designed to enhance driving safety, improve user experience, and provide personalized services, includes an information system, an interaction system, and a connectivity system. The information system displays real-time vehicle information and entertainment content on a high-resolution screen, providing the driver with a comprehensive understanding of the vehicle's status. The interaction system enables comprehensive information exchange between the vehicle and the driver through an intuitive user interface and voice recognition. The connectivity system utilizes vehicle-to-everything (V2X) technology to achieve seamless connection between the vehicle and external networks.

[0096] The intelligent driving domain includes a perception processing module, a trajectory prediction module, and a reinforcement learning module. The perception processing module acquires a blind-spot-free semantic bird's-eye view from the cloud control domain via Vehicle-to-Network (V2N) communication and crops it based on high-precision positioning provided by the satellite service domain. Simultaneously, the perception processing module constructs and uploads a real-time dynamic map using vehicle-side sensors, achieving real-time local updates of the offline map based on dynamic matching dependent on high-precision positioning. The trajectory prediction module stacks the blind-spot-free semantic bird's-eye view from the cloud control domain across multiple time steps and combines it with scene factors provided by the knowledge management domain to predict the trajectories of dynamic obstacles around the intelligent connected vehicle across multiple time steps, serving as another input to the reinforcement learning module. The reinforcement learning module stacks the input from the perception processing module across multiple time steps and integrates it with the trajectories of surrounding dynamic obstacles predicted by the trajectory prediction module into a complete reinforcement learning input. Furthermore, the reinforcement learning module constructs a fusion reward function based on the driving safety field provided by the cloud control domain to guide the training process of the vehicle-side reinforcement learning neural network. Finally, the reinforcement learning module outputs control commands to the powertrain domain.

[0097] The powertrain domain primarily outputs power control based on control commands from the intelligent driving domain. Through the coordinated action of the battery unit, motor unit, and electronic control unit, it enhances the overall performance and efficiency of the powertrain domain, ensuring the electric vehicle's power performance, energy management, and safety during driving. The battery unit, with its high-energy-density battery pack, provides a continuous and robust power supply, achieving long driving range and high efficiency. The motor unit, with its high-efficiency motor and advanced rotor design, provides rapid acceleration performance, resulting in a responsive driving experience. The electronic control unit, through advanced control algorithms, achieves intelligent coordination between the motor and battery, optimizing power output and energy management.

[0098] The intelligent chassis domain primarily optimizes vehicle driving comfort through an intelligent chassis system. Based on high-precision positioning, it provides the vehicle with information on traffic, weather, and road conditions for various driving scenarios, supporting the implementation of multi-domain chassis fusion technology. The intelligent chassis system includes, but is not limited to, a steering subsystem, a direct yaw control (DYC) subsystem, a suspension subsystem, and a drive subsystem, which are coupled in terms of state and cost. The steering subsystem and the DYC subsystem calculate the lateral control quantities of the intelligent connected vehicle, the suspension subsystem calculates the vertical control quantities, and the drive subsystem calculates the longitudinal control quantities. Through information transmission, the intelligent chassis domain connects to three major domains: the intelligent cockpit domain, the intelligent driving domain, and the powertrain domain. First, through interaction within the intelligent cockpit domain, the intelligent chassis domain dynamically adjusts, achieving interaction between the intelligent cockpit domain and the intelligent chassis domain. Second, this invention incorporates chassis feedback as part of the state input to the intelligent driving domain, enabling interaction between the intelligent driving domain and the intelligent chassis domain. Finally, the steering subsystem and DYC subsystem of the intelligent chassis domain calculate the yaw moment of the intelligent connected vehicle based on the output of the powertrain domain and the pre-aiming error between the actual path and the desired path. The drive subsystem obtains the desired target speed and calculates the four-wheel drive / braking torque of the intelligent connected vehicle based on the output of the powertrain domain and the current speed of the intelligent connected vehicle, thus realizing the interaction between the powertrain domain and the intelligent chassis domain.

[0099] (4) The intelligent connected vehicle group on the vehicle platform integrates the complete reinforcement learning state input based on the perception processing module and trajectory prediction module in the intelligent driving domain.

[0100] (5) The intelligent connected vehicle group on the vehicle-side platform further constructs a fusion reward function based on the driving safety field provided by the cloud control domain to guide the training process of the intelligent connected vehicle reinforcement learning model. The fusion reward function considers both driving safety and driving expectation. The driving safety-related reward function is constructed based on the driving risks of intelligent connected vehicles predicted by the driving safety field provided by the cloud control domain. The driving expectation reward function is constructed based on the lateral error and heading angle deviation between the actual path and the expected path of the intelligent connected vehicle.

[0101] (6) During reinforcement learning training, the intelligent connected vehicle on the vehicle-side platform uploads its own neural network Actor-Critic parameters to the knowledge management domain of the cloud platform. In different scenarios where the intelligent connected vehicle is located, the knowledge management domain selects neural network Actor-Critic parameters for parameter aggregation based on the intelligent chassis domain, and aggregates the local neural network Actor-Critic parameters based on federated learning. Furthermore, the cross-scenario training sequences of the intelligent connected vehicle's neural network are stored in the knowledge management domain. Based on these cross-scenario sequences, a scenario causal reasoning model is trained, and scenario factors corresponding to different scenarios are extracted to establish a scenario experience database.

[0102] (7) The knowledge management domain of the cloud platform sends the aggregated local neural network Actor-Critic parameters to the corresponding vehicle-side intelligent connected vehicles, repeating this process until the network converges. When a new intelligent connected vehicle connects to the integrated space-ground multi-domain high-level autonomous driving system on the vehicle-side platform, the knowledge management domain of the cloud platform predicts the scene factors of the scene based on the observation sequence of the intelligent connected vehicle, and retrieves the corresponding scene experience from the scene experience library to push the experience based on federated learning, thereby enabling the intelligent connected vehicle to quickly connect to the system.

[0103] In summary, this invention proposes a multi-domain integrated space-ground autonomous driving system for automobiles. By fusing information from satellite platforms, cloud platforms, and vehicle-side platforms, and integrating different data sources, it achieves more comprehensive and blind-spot-free perception and positioning for autonomous vehicles, improving the safety and reliability of planning and control, and promoting a comprehensive upgrade of the overall efficiency of intelligent transportation systems. Based on cloud error broadcasting and high- and low-orbit satellite fusion enhancement technologies, it solves the problem of vehicle communication disconnection caused by the limited coverage of base stations in single-terrestrial communication, ensuring the high-precision positioning and high-security communication requirements of autonomous driving. Based on a cloud-based reinforcement learning architecture, it improves the stability of the communication network of the integrated space-ground system through adaptive switching of the communication network. Based on the high-precision trajectory data of vehicles collected by the platform, it establishes a road semantic map and a decision-making and control experience pool. Through cloud-based federated learning scenario experience sharing, it compensates for the limitations of single-vehicle perception capabilities and solves the redundant safety and long-tail problems that single-vehicle intelligence cannot solve. The vehicle-side system dynamically updates the cloud-based road semantic map by matching the semantic map provided by the cloud with the local driving map constructed by the vehicle. High-precision positioning can provide vehicles with information such as traffic, weather, and road conditions in driving scenarios, improve the cross-scenario integration capability of the entire autonomous driving system, support the realization of chassis multi-domain fusion technology, and is of great value for the implementation of high-level autonomous driving technology.

[0104] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent methods or modifications that do not depart from the technology of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-domain integrated space-ground fusion automotive high-level automatic driving system, characterized in that, It includes three platforms and eight domains; the three platforms are the vehicle platform, the cloud platform, and the satellite platform; the eight domains are the intelligent cockpit domain, the intelligent driving domain, the powertrain domain, the intelligent chassis domain, the knowledge management domain, the cloud control domain, the satellite service domain, and the information and communication domain. The vehicle-side platform comprises multiple vehicle-side groups, each containing N intelligent connected vehicles. Each intelligent connected vehicle deploys an intelligent cockpit domain, an intelligent driving domain, a powertrain domain, and an intelligent chassis domain. The cloud platform deploys a cloud control domain and a knowledge management domain. The satellite platform deploys a satellite service domain; the information and communication domain enables information interconnection between the vehicle-side platform, the cloud platform, and the satellite platform. The satellite platform, based on the deployed satellite service domain, provides accurate time synchronization and reliable communication transmission between multiple domains through satellite signals, provides high-precision positioning as the basic data support for multi-domain fusion, and provides traffic information, weather changes, and road condition update data support. The cloud platform, based on the deployed cloud control domain and knowledge management domain, enables information transmission and processing, and achieves centralized information management and knowledge extraction through different cloud control applications. The vehicle-side platform, based on the deployed intelligent cockpit domain, intelligent driving domain, powertrain domain, and intelligent chassis domain, can dynamically update the cloud-based road semantic map, achieve status interaction with the driver, realize trajectory prediction, realize motor control and energy management, and realize vehicle driving comfort. 2.The integrated multi-domain and multi-level fusion automotive advanced driving system according to claim 1, wherein, The satellite platform includes a satellite service domain and related support platforms; the satellite service domain provides high-precision positioning information through satellite communication; the related support platforms include a traffic management platform that monitors traffic information in real time and provides a complete traffic network, a logistics platform that provides real-time logistics network information, a map platform that provides navigation services and regularly updates offline map information, a positioning platform that provides high-precision positioning services and location information sharing functions, and a meteorological platform that provides real-time weather monitoring and analysis. 3.The integrated multi-domain and multi-level automotive advanced driving system according to claim 1, wherein, The cloud platform includes a cloud control domain and a knowledge management domain, and achieves efficient information exchange between the vehicle, the cloud, and the satellite through Vehicle-to-Network (V2N) communication combined with satellite communication. The cloud control domain is based on three cloud control levels: edge cloud, regional cloud, and central cloud. It enables different cloud control applications through three types of cloud control empowerment: connected vehicle empowerment, traffic management and control, and traffic data empowerment. The knowledge management domain acquires neural network parameters of intelligent connected vehicles in the vehicle group through Vehicle-to-Network (V2N) communication, and performs local neural network Actor-Critic parameter aggregation by filtering neural network parameters through intelligent chassis feedback on the vehicle. The knowledge management domain trains a scenario causal reasoning model in cross-scenario training sequences, extracts transferable scenario factors under each scenario, and establishes a scenario experience library. When a new vehicle connects to the system, the knowledge management domain pushes a beyond-line-of-sight road semantic map of the driving environment in which the intelligent connected vehicle is located, and predicts the scenario factors of the environment in which the vehicle is located according to the scenario causal reasoning model. Based on the neural network parameters corresponding to the scenario factors in the scenario experience library, the knowledge management domain pushes parameter experience to the vehicle to achieve rapid vehicle access. 4.The integrated multi-domain and multi-level automotive advanced driving system according to claim 3, wherein, The cloud control applications include communication network monitoring applications, communication convergence enhancement applications, network switching strategies, blind-spot-free high-precision bird's-eye view maps, semantic bird's-eye view matrices, and the construction of driving safety fields. The communication network monitoring application is used to continuously monitor the status of the communication network, including signal strength, latency, bandwidth and packet loss. When the network fails or its performance degrades, the communication network monitoring application can detect it in time and issue an alarm so that the problem can be solved through subsequent network switching applications. The aforementioned communication fusion enhancement application sends phase deviation, which characterizes satellite clock error, satellite orbit error, and phase delay error, to high-orbit satellites via a satellite ground reference station. The high-orbit satellites calculate satellite ephemeris errors using the phase deviation data and send them to the corresponding low-orbit satellite set. Finally, the low-orbit satellites return navigation enhancement information to the satellite ground base station, using the navigation enhancement information to offset the orbit errors from the satellites during satellite positioning. The network switching strategy is trained by deep reinforcement learning to switch between satellite networks and cellular networks. The blind-spot-free high-precision bird's-eye view is obtained by receiving high-precision positioning provided by the satellite service domain deployed on the satellite platform through the satellite ground base station responsible for receiving and transmitting satellite signals. Based on the vehicle trajectory data collected by the platform under high-precision positioning, a road semantic map is established. The blind-spot-free high-precision bird's-eye view contains 4 channels of information, namely the road drivable area, lane boundaries, dynamic obstacles, and traffic infrastructure. The semantic bird's eye view matrix divides the high-precision bird's eye view without blind area into different semantic matrices according to different channels, and the size of the semantic bird's eye view matrix without blind area is [0, 1] W×H×C Wherein W represents the width of the semantic bird's eye view matrix without blind area, H represents the height of the semantic bird's eye view matrix without blind area, and C represents the number of channels of the semantic bird's eye view matrix without blind area. The driving safety field is constructed based on a high-precision bird's-eye view with no blind spots to guide vehicle driving behavior. The driving safety field is modeled using Gaussian equations and assesses the driving risks of intelligent connected vehicles in their environment by describing the dynamic transfer process of the driving safety field risk center during the movement of intelligent connected vehicles. 5.The integrated multi-domain and multi-level automotive advanced driving system according to claim 4, wherein, The deep reinforcement learning in the network switching strategy is described by the tuple (S, P, A, R), where: S represents the input state variables for deep reinforcement learning: S = [s cell ,s sat ,d cell ,d sat ,b cell ,b sat ,p lat ,p long ,c,h] wherein s cell represents the current cellular network signal strength, s sat represents the current satellite network signal strength, d cell represents the current cellular network delay, d sat represents the current satellite network delay, b cell represents the current cellular network bandwidth, b sat represents the current satellite network bandwidth, p lat represents the current vehicle latitude, p long represents the current vehicle longitude, c represents the current network connection state, c = 0 represents the current connection cellular network, c = 1 represents the current connection satellite network, h represents the communication quality history in the recent period of time, represented by the communication network delay; P represents the state transition equation of deep reinforcement learning, which is fitted using a neural network; A represents the output of the deep reinforcement learning communication network switching model. A=0 means maintaining the current communication network connection, A=1 means switching to a cellular network, and A=2 means switching to a satellite network. R represents the reward function output after executing the communication network switching strategy: R t = w1R stab + w2R cost + w3R quality wherein R t represents the environment feedback at time t, w1, w2, and w3 represent the weighting weights of each sub-reward function, respectively, R stab represents the connection stability related reward function, R cost represents the switching cost related reward function, R quality represents the communication service quality related reward function. 6.The integrated multi-domain and multi-level automotive advanced driving system according to claim 1, wherein, The intelligent cockpit domain in the vehicle platform is used to improve driving safety, enhance user experience, and provide personalized services, and includes information systems, interaction systems, and connectivity systems. The information system displays vehicle information and entertainment content in real time through a high-resolution screen, enabling the driver to have a comprehensive understanding of the vehicle's status; the interactive system enables comprehensive information interaction between the vehicle and the driver through an intuitive user interface and voice recognition; and the network system achieves seamless connection between the vehicle and external networks through vehicle networking technology. 7.The integrated multi-domain and multi-level automotive advanced driving system according to claim 1, wherein, The intelligent driving domain in the vehicle platform includes a perception processing module, a trajectory prediction module, and a reinforcement learning module. The perception processing module acquires a non-blind-area semantic bird's-eye view provided by a cloud control domain through Vehicle-to-Network (V2N) communication, and performs clipping according to high-precision positioning provided by a satellite service domain, and a size of the clipped non-blind-area semantic bird's-eye view matrix is [0, 1] w×h×c where w represents the width of the clipped non-blind-area semantic bird's-eye view matrix, h represents the height of the clipped non-blind-area semantic bird's-eye view matrix, and c represents the channel number of the clipped non-blind-area semantic bird's-eye view matrix; the clipped non-blind-area semantic bird's-eye view is stacked with sensor information as part of an input of a reinforcement learning module, and the perception processing module constructs a real-time dynamic map using vehicle-end sensors and uploads the real-time dynamic map; through dynamic matching based on high-precision positioning, real-time local updating of an offline map is realized. The trajectory prediction module stacks the blind-spot-free semantic bird's-eye view provided by the cloud control domain in multiple time steps, and combines the scene factors provided by the knowledge management domain to realize the multi-time step trajectory prediction of dynamic obstacles around the intelligent connected vehicle, which serves as another part of the input to the reinforcement learning module. The reinforcement learning module stacks the multi-time-step inputs provided by the perception processing module and integrates them with the multi-time-step trajectory prediction results of the surrounding dynamic obstacles provided by the trajectory prediction module into a complete reinforcement learning input. Based on the driving safety field provided by the cloud control domain, the reinforcement learning module constructs a fusion reward function to guide the training process of the vehicle-side reinforcement learning neural network. Finally, the reinforcement learning module outputs control commands to the powertrain domain. 8.The integrated multi-domain and multi-level fusion automotive advanced driving system according to claim 1, wherein, The powertrain domain in the vehicle platform outputs power control according to the control commands of the intelligent driving domain. Through the synergistic effect of the battery unit, motor unit and electronic control unit, the overall performance and efficiency of the powertrain domain are improved, ensuring the power performance, energy management and safety of electric vehicles during driving. The battery unit provides a continuous and powerful power supply through a high-energy-density battery pack, enabling the vehicle to achieve long range and high efficiency. The motor unit, through its high-efficiency motor and advanced rotor design, provides rapid acceleration performance and delivers a responsive driving experience. The electronic control unit uses advanced control algorithms to achieve intelligent coordination between the motor and the battery, optimizing power output and energy management. 9.The integrated multi-domain and multi-level automotive advanced driving system of claim 1, wherein, The intelligent chassis domain in the vehicle platform is connected to three major domains: intelligent cockpit domain, intelligent driving domain, and powertrain domain. First, through the interaction in the intelligent cockpit domain, the intelligent chassis domain is dynamically adjusted, realizing the interaction between the intelligent cockpit domain and the intelligent chassis domain. Secondly, chassis feedback is used as part of the state input of the intelligent driving domain to realize the interaction between the intelligent driving domain and the intelligent chassis domain. Finally, the steering subsystem and DYC subsystem of the intelligent chassis domain calculate the yaw moment of the intelligent connected vehicle based on the output of the powertrain domain and the pre-aiming error between the actual path and the desired path. The drive subsystem obtains the desired target speed and calculates the four-wheel drive or braking torque of the intelligent connected vehicle based on the output of the powertrain domain and the current speed of the intelligent connected vehicle, thus realizing the interaction between the powertrain domain and the intelligent chassis domain. 10.The integrated multi-domain and multi-level automotive advanced driving system of claim 1, wherein, The information communication domain realizes inter-satellite communication, information monitoring, high-precision positioning and time synchronization service on a satellite platform; the information communication domain realizes information storage, data analysis, experience summary and communication enhancement service on a cloud platform; and the information communication domain realizes scene recognition, information sharing, experience receiving and network switching service on a vehicle platform.

Images