A cloud and vehicle end cooperative method, cloud processing method, system and vehicle

Abstract

The application discloses a cloud and vehicle end cooperative method, a cloud processing method, a system and a vehicle. The cloud and vehicle end cooperative method is applied to the vehicle end, and comprises the following steps: sending sensor data of the vehicle end to the cloud; receiving at least two different types of control quantities from the cloud, wherein the control quantities are generated by the cloud based on the sensor data processing and are used for directly controlling vehicle actuators; analyzing the at least two different types of control quantities to generate an analysis result; and generating a control signal based on the analysis result to simultaneously control at least two different vehicle actuators. The method generates control instructions through cloud processing of complex calculations, and the vehicle end performs analysis and execution, thereby overcoming the limitation of vehicle-mounted computing resources and improving the processing capacity and performance of the automatic driving system.

Description

Technical Field

[0001] This application relates to the field of vehicle networking technology, specifically to a cloud-vehicle collaboration method, a cloud processing method, a system, and a vehicle. Background Technology

[0002] In the field of autonomous driving, achieving advanced levels of autonomous driving requires powerful artificial intelligence models for real-time environmental perception, decision-making, planning, and control. Related technologies typically rely on locally integrated computing devices within the vehicle to perform these complex processing tasks. However, due to inherent constraints such as limited physical space, available power, and heat dissipation within the vehicle, the scale of integrated computing and storage resources has a natural upper limit. Therefore, within the current technological framework, the vehicle's local computing power struggles to continuously adapt to and process increasingly complex and large-scale artificial intelligence models, becoming a major factor limiting further improvements in the performance of autonomous driving systems. Summary of the Invention

[0003] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a cloud-to-vehicle collaboration method, a cloud processing method, a system, and a vehicle, which can improve the processing capabilities for complex autonomous driving tasks.

[0004] According to a first aspect of this application, a cloud-vehicle collaboration method is provided, applied to a vehicle, comprising: sending sensor data from the vehicle to the cloud; receiving at least two different types of control quantities from the cloud, wherein the control quantities are generated by the cloud based on the sensor data and are used to directly control vehicle actuators; parsing the at least two different types of control quantities to generate a parsing result; and generating a control signal based on the parsing result to simultaneously control at least two different vehicle actuators.

[0005] As one possible implementation, the vehicle actuators include a steering system, a power system, and a braking system; the generation of control signals based on the analysis results to simultaneously control at least two different vehicle actuators includes: generating control signals for coordinated control of the vehicle steering system, power system, and braking system according to at least two types of the control quantities, to perform at least two of the following operations: adjusting the vehicle's perception focus direction and recognition priority; updating the vehicle's driving route and speed planning; adjusting the vehicle's steering angle, throttle opening, and braking force ratio; triggering the vehicle's emergency braking or evasive steering operation; and adjusting the vehicle's lane keeping, following distance, or traffic signal response behavior.

[0006] As one possible implementation, after generating the control signal based on the parsing result, the cloud-vehicle collaboration method further includes: after executing the control signal, sending vehicle status data back to the cloud so that the cloud can optimize and iterate the model used to generate the control quantity based on the vehicle status data.

[0007] According to a second aspect of this application, a cloud processing method is provided, comprising: receiving sensor data from a vehicle; processing the sensor data to generate at least two different types of control quantities; the control quantities being used to directly control vehicle actuators; and sending the at least two control quantities to the vehicle so that the vehicle can simultaneously control at least two different vehicle actuators after parsing.

[0008] As one possible implementation, the control quantity is generated by adjusting the internal parameters of the generation model in the cloud based on scene data; wherein, the scene data includes one or more of real-time weather information, road type information, real-time traffic flow information, and traffic event information.

[0009] As one possible implementation, sending at least two control quantities to the vehicle includes: the cloud selecting a target link from multiple preset communication links based on the vehicle's real-time location information and network status information; the cloud switching the current communication link to the target link and sending the control quantities through the target link.

[0010] According to a third aspect of this application, a cloud-vehicle collaborative system is provided, comprising: a cloud processing module configured to: receive sensor data from a vehicle; process the received sensor data to generate at least two different types of control quantities; the control quantities being used to directly control vehicle actuators; and sending the at least two control quantities to the vehicle; and a vehicle actuator module communicatively connected to the cloud processing module, configured to: send sensor data from the vehicle to the cloud; receive and parse the at least two control quantities from the cloud processing module to generate a parsing result; and generate a control signal based on the parsing result to simultaneously control the at least two different vehicle actuators.

[0011] As one possible implementation, the control quantity is a set of instructions or parameters that can be directly mapped and converted into underlying drive signals by the vehicle-side execution module; the underlying drive signals are used to drive at least two vehicle actuators in the vehicle's steering system, drive system, and braking system.

[0012] As one possible implementation, the cloud processing module adopts a multi-node redundant computing architecture, which is configured to switch computing tasks to other available nodes to continue execution when a single computing node fails.

[0013] According to a fourth aspect of this application, a vehicle is provided, comprising: a power system, a steering system, and a braking system; and a cloud-vehicle collaborative system as described in the third aspect or any implementation thereof, wherein the vehicle-side execution module of the cloud-vehicle collaborative system is communicatively connected to the power system, the steering system, and the braking system, and is used to control at least two vehicle actuators in the power system, the steering system, and the braking system based on control signals.

[0014] The cloud-vehicle collaboration method, cloud processing method, system, and vehicle provided in this application involve the vehicle sending sensor data to the cloud, and the cloud generating at least two different types of control quantities based on the data to directly control vehicle actuators. The cloud then sends these data back to the vehicle for parsing and generation of control signals to simultaneously control multiple vehicle actuators. This offloads the computationally intensive, complete autonomous driving processing task from resource-constrained onboard equipment to the elastically scalable cloud for execution. Compared to existing solutions that only generate single-dimensional optimization instructions, this application achieves multi-dimensional real-time collaborative control of vehicle driving strategies and behaviors, overcoming the physical bottlenecks of onboard hardware in terms of computing and storage resources, and improving the processing capacity of complex autonomous driving tasks and the overall performance of the system. Attached Figure Description

[0015] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 This is a schematic diagram of the structure of a cloud-vehicle collaborative system provided in an exemplary embodiment of this application.

[0017] Figure 2 This is a flowchart illustrating a cloud-to-vehicle collaboration method provided in an exemplary embodiment of this application.

[0018] Figure 3 This is a flowchart illustrating a cloud processing method provided in an exemplary embodiment of this application.

[0019] Figure 4 This is a schematic diagram of the structure of a vehicle provided in an exemplary embodiment of this application.

[0020] Figure 5 This is a schematic diagram of the structure of a cloud-vehicle collaborative device provided in an exemplary embodiment of this application.

[0021] Figure 6 This is a schematic diagram of the structure of a cloud processing device provided in an exemplary embodiment of this application.

[0022] Figure 7 This is a structural diagram of an electronic device provided in an exemplary embodiment of this application. Detailed Implementation

[0023] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0024] Application Overview In the field of autonomous driving technology, achieving advanced environmental perception, path planning, and vehicle control typically relies on onboard computing devices to process most AI (Artificial Intelligence) algorithms. For example, an onboard computing unit is deployed inside the vehicle, containing a high-performance processor, a dedicated AI chip, and large-capacity storage. Specifically, various sensors on the vehicle collect real-time data on the surrounding environment and the vehicle's own state, inputting this data into the onboard computing unit. The onboard computing unit runs a series of complex AI models, processes the input data, and ultimately generates commands to control the vehicle's steering, power, and braking mechanisms. This approach is widely used primarily because it ensures autonomous vehicle operation even without a network connection or with an unstable network, thus enabling the basic functions of an autonomous driving system.

[0025] However, the solutions perform poorly when applied to complex autonomous driving scenarios. This is because these scenarios require processing large-scale and high-precision AI models. To optimize local processing power and real-time performance, these solutions are inherently designed to compromise vehicle performance in other areas. For example, when performing multi-object tracking and intent prediction at complex urban intersections, or during dynamic path replanning at high speeds over long distances, the limited computing power and storage space of onboard computing devices prevent the operation of advanced AI models with large parameter scales. This leads to reduced perception distance, conservative planning decisions, or insufficient control precision, thus affecting the safety and traffic efficiency of autonomous driving. Simultaneously, the continuous operation of high-performance chips generates a large amount of heat, posing a challenge to the vehicle's cooling system, and its high power consumption directly shortens the driving range of electric vehicles.

[0026] The root causes of the aforementioned unsatisfactory performance lie in the following aspects: From a hardware perspective, the limited interior space and stringent heat dissipation requirements of vehicles necessitate a primary focus on driving needs, which fundamentally restricts the power available for onboard computing devices, thus limiting their size and computing power. From a data and model perspective, the sheer volume of road information, traffic rules, high-precision maps, and continuously iterating AI model parameters required for autonomous driving is enormous. This data far exceeds the capacity of onboard storage devices, preventing the onboard system from acquiring the latest and most comprehensive environmental knowledge. These factors collectively limit the ability of existing solutions to meet the demands of high-level autonomous driving.

[0027] To overcome the aforementioned problems, this invention proposes a different technical approach. The core concept is the introduction of a novel architecture for collaborative processing between the cloud and the vehicle. This architecture offloads computationally intensive AI model processing tasks from resource-constrained in-vehicle equipment, entrusting these tasks to remote cloud servers with elastic computing and massive storage resources. This improves the processing flow and division of responsibilities for autonomous driving tasks. The cloud generates multi-dimensional control variables that can directly control the vehicle's actuators. Instruction parsing and execution, which require extremely high real-time performance, remain on the vehicle. This division of labor overcomes the computing power and storage bottlenecks of in-vehicle hardware without sacrificing system real-time performance and reliability. Simultaneously, it reduces energy consumption and hardware costs on the in-vehicle side. This solves the problem that existing solutions cannot efficiently run complex AI models due to the physical limitations of the vehicle, thus overcoming the constraints on autonomous driving performance.

[0028] Exemplary System To concretize the inventive concept, Figure 1 This is a schematic diagram of the structure of a cloud-vehicle collaborative system provided in an exemplary embodiment of this application. First, as shown... Figure 1 The diagram illustrates a cloud-vehicle collaborative system 1, comprising a cloud processing module 11 and a vehicle execution module 12. The cloud processing module 11 is configured to: receive sensor data from the vehicle; process the received sensor data to generate at least two different types of control quantities; these control quantities are used to directly control vehicle actuators; and send the at least two control quantities to the vehicle. The vehicle execution module 12 is configured to: send sensor data from the vehicle to the cloud; receive and parse the at least two control quantities from the cloud processing module 11 to generate a parsing result; and generate a control signal based on the parsing result to simultaneously control at least two different vehicle actuators.

[0029] The vehicle-side execution module 12 is typically deployed inside the vehicle. It is electrically connected to various sensors on the vehicle, including cameras, lidar, millimeter-wave radar, and inertial measurement units. The module is also electrically connected to the vehicle's actuators, such as the steering system, powertrain, and braking system. The cloud processing module 11 is deployed in a remote data center. It typically consists of a server cluster with flexible computing and storage resources. The cloud processing module 11 connects to the vehicle-side execution module 12 via a communication network, which can include cellular networks, satellite networks, and dedicated vehicle-to-everything (V2X) networks. In this system environment, the data flow is as follows: vehicle sensors collect raw data, which is then sent from the vehicle-side execution module 12 to the cloud processing module 11. The cloud processing module 11 processes the data, generating control signals. These control signals are then sent to the vehicle-side execution module 12, which analyzes the control signals and generates control signals. These control signals ultimately drive the corresponding vehicle actuators to move.

[0030] In this application, sensor data refers to data collected by various sensing devices carried by the vehicle. This data is used to characterize the vehicle's own state and the state of its surrounding environment. The data can be raw data or data that has undergone preliminary processing. For example, sensor data may include image data, point cloud data, millimeter-wave radar signals, vehicle speed, acceleration, yaw rate, and positioning information. Control quantity refers to information generated by the cloud and sent to the vehicle. This control quantity can be parsed by the vehicle and used to directly drive the vehicle's actuators to produce actions. The control quantity can be instruction information or a set of parameters; for example, it can be a set of structured control instruction codes. It can also be a set of parameters containing specific values, such as a target steering angle, throttle opening percentage, and brake pressure value. Direct control means that the operations performed by the vehicle based on the control quantity do not require a complex secondary decision-making process involving multiple possibilities, but can be directly and deterministically mapped to the driving signals of the underlying hardware. Vehicle actuators refer to components or systems on the vehicle that receive electronic control signals and generate mechanical actions, thereby changing the vehicle's driving state. For example, vehicle actuators may include steering motors, electronic throttle pedal controllers, and electronic brake controllers.

[0031] Exemplary Method 1 Figure 2 This is a flowchart illustrating a cloud-vehicle collaboration method provided in an exemplary embodiment of this application. The method mainly describes the execution logic on the vehicle side.

[0032] See Figure 2 S210: Sends sensor data from the vehicle to the cloud.

[0033] S210 is used to initiate the cloud-based collaborative processing flow, uploading the raw perception data collected by the vehicle to the cloud. This data can serve as the data foundation for deep computing in the cloud.

[0034] In some embodiments, the vehicle-side execution module aggregates data streams from various onboard sensors. This data can undergo necessary preprocessing before transmission, including timestamp synchronization, data packaging, and compression encoding. Preprocessing reduces transmission bandwidth consumption and latency. Furthermore, the vehicle can continuously or on-demand send data to the cloud. For example, when the vehicle detects a complex scene, it can trigger high-frequency data uploads to ensure the cloud has sufficient information for accurate calculations. Adopting an on-demand transmission method allows for dynamic adjustment of the communication load based on actual needs, ensuring computational accuracy while optimizing network resource utilization.

[0035] In some embodiments, sending data to the cloud can be achieved in various ways. For example, it can include periodic uploads at a fixed frequency, event-triggered uploads, or uploads requested by cloud instructions.

[0036] See Figure 2 S220: Receives at least two different types of control inputs from the cloud.

[0037] S220 is the step where the vehicle acquires the results of intelligent computing from the cloud. The control variables are generated by the cloud based on sensor data processing and are used to directly control the vehicle's actuators. Receiving at least two different types of control variables means that the vehicle does not obtain a single-dimensional suggestion or intermediate result from the cloud. Instead, the vehicle acquires multiple types of directly executable control commands or parameters covering different control dimensions of the vehicle. For example, one control variable might correspond to the target steering angle of the steering system, while another might correspond to the target torque of the powertrain. These two control variable types are different and are used to control different actuators.

[0038] As one possible implementation, the control quantity is a set of instructions or parameters that can be directly mapped and converted into underlying drive signals by the vehicle-side execution module. These underlying drive signals are used to drive at least two vehicle actuators in the vehicle's steering, drive, and braking systems. This distinguishes the technical attributes of the control quantity from traditional intermediate calculation results. By designing the control quantity as a structured, semantically clear set of instructions or parameters, the vehicle-side eliminates the need for complex secondary calculations or decisions. It only requires low-complexity decoding and mapping operations to generate the underlying drive signals. This significantly reduces the computational and power consumption requirements of the vehicle-side processing unit, simplifies the vehicle-side hardware design, and shortens the end-to-end latency from instruction reception to execution, thus improving system response speed. The vehicle-side execution module internally incorporates corresponding parsing and mapping logic, which can be either hardware circuitry or software modules. This logic can unambiguously convert the received instruction or parameter set into the PWM waveform, analog signal, or CAN message required to drive the actuators through table lookup, calculation, or simple logical judgment. This control design minimizes processing latency and computational overhead on the vehicle before execution, maximizing the conversion of cloud intelligence into real-time vehicle actions and enhancing the advantages of the collaborative architecture.

[0039] As one possible implementation, the types of control variables received can correspond to different stages of the autonomous driving function. For example, control variables may include those corresponding to environmental perception analysis results, which can be used to adjust the vehicle's perception and recognition focus direction. Control variables may include those corresponding to path planning results, which can be used to indicate the coordinates of the next path point and the desired vehicle speed. Control variables may include vehicle attitude control parameters, such as optimal steering angle, throttle-brake ratio coefficients for hill driving, and tire grip adaptation parameters for slippery surfaces. Control variables may also include hazard warning decision results, such as emergency avoidance commands. Control variables may include traffic rule adaptation results, such as following distance adjustment parameters. Receiving control variables that include at least two of the above types allows the computational results of powerful cloud-based AI models at multiple stages to be distributed in the form of executable commands, achieving comprehensive and coordinated adjustments to the vehicle's driving strategies and behaviors.

[0040] As another possible implementation, the types of control signals received are not limited to those listed above. Any control signal can be used to directly control different vehicle actuators. For example, control signals can also include lighting control commands or suspension adjustment parameters. Lighting control commands are used to control high beams and turn signals, etc.

[0041] See also Figure 2 S230: Analyze at least two different types of control variables and generate analysis results.

[0042] S230 is the process by which the vehicle-side converts control signals sent from the cloud into usable internal control information. The core of the parsing process lies in identifying the type of control signal. Parsing extracts key values ​​or instruction logic from the control signal that can be directly used to generate drive signals. Because the control signal itself is designed to be directly mapped to the actuator, the parsing process is typically a deterministic decoding or conversion process. That is, the parsing process is not a decision-making process that requires significant computational resources.

[0043] In some embodiments, the parsing work can be performed by a lightweight AI processing unit on the vehicle side. This lightweight AI processing unit is a low-power chip or embedded system specifically designed for instruction parsing. It can receive AI calculation results from the cloud, such as environmental perception analysis results, optimal path planning solutions, vehicle attitude control parameters, hazard warning decision results, and traffic rule adaptation results, and adjust the vehicle's driving strategy and behavior in real time based on these results. Specifically, it adjusts the vehicle's perception threshold and recognition priority based on environmental perception results, dynamically updates the driving route and speed planning based on path planning results, precisely adjusts the steering angle, throttle opening, and braking force based on attitude control parameters, performs hazard avoidance operations based on hazard warning results, and completes actions such as lane switching, following distance adjustment, and traffic light response based on traffic rule adaptation results. In other words, by decoding and converting the structured instruction code or control parameter set issued from the cloud through the lightweight AI processing unit, the computational power requirements and power consumption on the vehicle side can be greatly reduced, the dependence on network bandwidth and latency can be reduced, and the response speed and stability of the autonomous driving system can be improved. In addition, high-performance chips in pure automotive solutions consume extremely high energy during operation. If a vehicle only deploys a lightweight AI processing unit, its power consumption is far lower than that of traditional automotive high-performance chips. This significantly reduces the vehicle's computing power consumption and directly improves the vehicle's range.

[0044] Understandably, the parsing process can be implemented in various ways. For example, it can be achieved through instruction mapping based on lookup tables, state machine transitions based on simple rules, or by running a lightweight neural network to understand and decompose the control variables.

[0045] See also Figure 2 S240: Generate control signals based on the analysis results to simultaneously control at least two different vehicle actuators.

[0046] The S240 synchronously converts the parsed control intentions for different actuators into actual control signals, which are then sent to the corresponding controllers. Simultaneous control of at least two different actuators is a result of the multiple types of control variables distributed from the cloud, and is the technological foundation for achieving complex and coordinated autonomous driving behavior. For example, simultaneously steering and accelerating to overtake requires coordinated control.

[0047] As one possible implementation, the vehicle actuators include a steering system, a power system, and a braking system; S240 may include: generating control signals for coordinated control of the vehicle's steering system, power system, and braking system based on at least two types of control quantities, to perform at least two of the following operations: adjusting the vehicle's perception focus direction and recognition priority; updating the vehicle's driving route and speed planning; adjusting the vehicle's steering angle, throttle opening, and braking force ratio; triggering the vehicle's emergency braking or evasive steering operation; and adjusting the vehicle's lane keeping, following distance, or traffic signal response behavior.

[0048] For example, if the analysis results simultaneously include instructions to adjust following distance and perform slight braking, the generated control signals need to work in tandem with the powertrain and braking systems. The effect on the powertrain is to reduce torque output, and the effect on the braking system is to apply slight braking force. This synergistic action achieves smooth deceleration and increases following distance. This collaborative approach allows the multi-dimensional AI calculation results from the cloud to be executed collaboratively on the vehicle, enabling safe, smooth, and efficient autonomous driving behavior.

[0049] In some embodiments, after generating control signals based on the analysis results, the cloud-vehicle collaboration method may further include: after executing the control signals, transmitting vehicle state data back to the cloud, so that the cloud can optimize and iterate the model used to generate control quantities based on the vehicle state data. This step can establish a feedback loop from execution results to model optimization. This feedback loop enables the cloud-based AI model to continuously improve based on the actual control effect. That is to say, the process is not the end point after the vehicle executes the control instructions from the cloud. The vehicle-side execution module collects the vehicle state data after execution, such as the vehicle's actual driving trajectory, yaw rate, acceleration, whether it is stable, and whether it successfully avoids obstacles, etc. This vehicle state data is encapsulated and transmitted back to the cloud processing module. The cloud can obtain the execution feedback of its decisions and controls in the actual physical world, thereby providing valuable real data for the training and optimization of the cloud-based AI model.

[0050] In other embodiments, the vehicle status data can be transmitted to the cloud periodically, upon the occurrence of a specific event, or at the request of the cloud. Specific events may include driver takeover and severe vehicle vibration. Alternatively, the vehicle may send status reports to the cloud at fixed time intervals, including information such as vehicle pose, speed, and control signal following error.

[0051] As one possible implementation, after receiving the returned vehicle status data, the cloud processing module combines it with previously issued control variables and the sensor data received at that time to form a training data sample. These samples are stored in a cloud database and used to periodically or triggerically retrain or fine-tune the AI ​​model that generates control variables. For example, if data from a large number of vehicles in a certain curve scenario shows a systematic deviation between the actual trajectory and the planned trajectory, the cloud can use this data to fine-tune the planning or control model for that scenario to correct the deviation. Therefore, the cloud-based AI model can possess the ability to learn online and continuously evolve, constantly improving the accuracy and adaptability of the model's generated control variables, helping to solve the problem that static models struggle to adapt to dynamically changing long-tail scenarios.

[0052] Exemplary Method 2 Figure 3 This is a schematic flowchart of a cloud processing method provided in an exemplary embodiment of this application. This method is applied to the cloud and works in conjunction with the vehicle-side method to form a complete cloud-vehicle collaborative processing flow.

[0053] See Figure 3 S310: Receives sensor data from the vehicle.

[0054] The S310 serves as the starting point for cloud-based computation. The cloud processing module receives sensor data uploaded from the vehicle via a communication interface. This sensor data reflects the vehicle's real-time status and environment, providing the fundamental input for the cloud-based AI model to perform inference and computation.

[0055] See Figure 3 S320: Processes sensor data to generate at least two different types of control variables.

[0056] In this architecture, control variables are used to directly control vehicle actuators and can be encapsulated into a format that can be parsed by the vehicle. The cloud leverages its powerful computing resources and vast storage of AI models, map data, and traffic information to perform in-depth processing on the received sensor data. This processing typically involves the serial or parallel operation of multiple AI models. For example, a perception model identifies obstacles, a decision-making model assesses risks, a planning model calculates trajectories, and a control model generates specific parameters. The generated control variables, at least two different types, are the output of these models working together. By placing core, complex computations in the cloud, which can flexibly allocate computing power according to vehicle needs, a single vehicle can utilize the cloud's cluster computing power to complete ultra-large-scale AI calculations. Furthermore, cloud computing power can be elastically expanded according to business needs. Therefore, this architecture can overcome the vehicle's computing power bottleneck and adapt to complex AI models.

[0057] As one possible implementation, the control parameters are generated by adjusting the internal parameters of the model in the cloud based on scenario data. Scenario data includes one or more of the following: real-time weather information, road type information, real-time traffic flow information, and traffic event information. For example, when generating path planning and control parameters, the cloud not only relies on current sensor data but also queries real-time traffic flow information. If congestion occurs ahead of the planned path, the cloud model can dynamically adjust the weight parameters used to evaluate path costs. After adjustment, it generates a recommended detour route and corresponding speed control parameters. As another example, when rain is detected, the model used in the cloud to generate vehicle attitude control parameters can automatically increase the internal safety parameters related to the road surface adhesion coefficient, thereby generating more conservative steering and braking control parameters. Therefore, by introducing real-time scenario data as the basis for model parameter tuning, the cloud AI model is no longer static but can dynamically adjust its processing methods according to the specific environment, thereby generating control strategies and parameters that are more closely aligned with the actual scenario, improving the adaptability and safety of the autonomous driving system in different environments. Furthermore, the cloud AI model possesses strong environmental adaptability and scenario generalization capabilities, and the generated control parameters are accurate, safe, and practical.

[0058] Understandably, to improve the accuracy and efficiency of the model, optimization algorithms can be employed. By training and adjusting the model parameters, it can be made to better adapt to different real-world application scenarios such as urban roads, highways, and rural roads. Furthermore, the model can generate adaptive calculation results for different road conditions, weather, and traffic flow, which enhances the practicality and accuracy of the calculation results.

[0059] See also Figure 3 S330: Sends at least two control quantities to the vehicle end so that the vehicle end can simultaneously control at least two different vehicle actuators after parsing.

[0060] As one possible implementation, S330 may include: the cloud selecting a target link from multiple preset communication links based on the vehicle's real-time location information and network status information; the cloud switching the current communication link to the target link; and sending control signals through the target link.

[0061] For example, a communication link manager is maintained in the cloud. This manager continuously monitors the signal strength, latency, and bandwidth information reported by the vehicle across different networks. When a vehicle moves at high speed from the coverage edge of one base station to another, the manager predicts an impending degradation in the current link quality. It can proactively switch the link used to transmit high-priority control commands from the current base station to a new base station with a stronger signal. High-priority control commands can include emergency braking commands. By dynamically selecting the optimal communication path through real-time link switching, the system effectively combats transmission instability caused by vehicle movement and network fluctuations, ensuring low-latency and highly reliable transmission of critical control commands. Furthermore, the cloud's distributed communication scheduling capabilities can dynamically switch transmission links based on the vehicle's real-time location and network status information, avoiding single-point-of-failure issues in onboard device communication and improving data transmission stability. This communication transmission system ensures that the vehicle can receive decision-making and control calculation results from the cloud within milliseconds, providing communication assurance for real-time adjustments to driving strategies and behaviors.

[0062] As one possible implementation, different strategies can be adopted for transmitting control signals. For example, latency-sensitive control signals (such as emergency braking commands) can be transmitted via a high-priority, low-latency link, while latency-insensitive control signals (such as path planning updates) can be transmitted via a standard link.

[0063] In some embodiments, the cloud processing module can employ a multi-node redundant computing architecture, configured to switch computing tasks to other available nodes for continued execution when a single computing node fails. That is, the cloud processing module does not run on a single physical server; it can be deployed in a cluster of multiple computing nodes. Computing nodes can be servers, virtual machines, or containers. These nodes can reside in different racks within the same data center or be distributed across data centers in different regions for disaster recovery. The cluster management system assigns a primary computing node to each received data processing request from a vehicle. Simultaneously, the task status and necessary data for the request are synchronized to one or more backup nodes. The principle behind this design is to utilize redundant resources to ensure service continuity. The advantage is that when the monitoring system detects that the primary computing node has failed due to hardware failure, software crash, or network interruption, the cluster management system can quickly and seamlessly switch unfinished computing tasks to pre-prepared backup nodes for continued execution. For the vehicle, this switching process may be imperceptible, or only result in a very brief computing delay. The vehicle will not lose its autonomous driving capabilities due to the failure of a single node in the cloud. In other words, the cloud has multi-node redundancy computing capabilities, and if one cloud computing node fails, it can quickly switch to other nodes. At the same time, the lightweight AI processing unit on the vehicle has basic emergency handling capabilities. Combined with real-time fault scheduling in the cloud, it can avoid system failure due to single device failure. Compared with the single point of failure mode of pure vehicle-mounted solutions, the fault tolerance and safety are greatly improved.

[0064] Exemplary vehicle Figure 4 This is a schematic diagram of the structure of a vehicle provided in an exemplary embodiment of this application. This application also provides... Figure 4 The vehicle shown. Vehicle 4 includes a powertrain system 41, a steering system 42, and a braking system 43. The vehicle also includes a cloud-to-vehicle collaborative system 44. The cloud-to-vehicle collaborative system 44 and... Figure 1 The cloud-vehicle collaborative system 3 shown represents the same system. The vehicle-side execution module in the cloud-vehicle collaborative system 44 is used to control at least two vehicle actuators in the powertrain 41, steering system 42, and braking system 43 based on control signals. This vehicle embodiment demonstrates the implementation of the aforementioned system and method in a specific product. As part of the vehicle's electronic and electrical architecture, the vehicle-side execution module is responsible for receiving control signals from the cloud. Ultimately, the vehicle-side execution module drives the traditional vehicle actuators, enabling the vehicle to achieve advanced autonomous driving capabilities beyond the limitations of local computing power.

[0065] Exemplary device Figure 5 This is a schematic diagram of the structure of a cloud-vehicle collaborative device provided in an exemplary embodiment of this application, as shown below. Figure 5 As shown, the cloud-vehicle collaborative device 5 includes: a data transmission module 51 for transmitting sensor data from the vehicle to the cloud; a control quantity receiving module 52 for receiving at least two different types of control quantities from the cloud; wherein the control quantities are generated by the cloud based on sensor data processing and are used to directly control the vehicle's actuators; a parsing module 53 for parsing at least two different types of control quantities and generating parsing results; and a signal generation and control module 54 for generating control signals based on the parsing results to simultaneously control at least two different actuators of the vehicle.

[0066] In some embodiments, the vehicle actuators include a steering system, a power system, and a braking system. The signal generation and control module 54 can be configured to generate control signals for coordinating the control of the vehicle's steering system, power system, and braking system based on at least two types of control quantities, to perform at least two of the following operations: adjusting the vehicle's perception focus direction and recognition priority; updating the vehicle's driving route and speed planning; adjusting the vehicle's steering angle, throttle opening, and braking force ratio; triggering the vehicle's emergency braking or evasive steering operation; and adjusting the vehicle's lane keeping, following distance, or traffic signal response behavior.

[0067] In some embodiments, the cloud-vehicle collaboration device 5 can also be configured to: after executing a control signal, transmit vehicle status data back to the cloud so that the cloud can optimize and iterate the model used to generate control quantities based on the vehicle status data.

[0068] Figure 6 This is a schematic diagram of the structure of a cloud-vehicle collaborative device provided in an exemplary embodiment of this application, as shown below. Figure 6 As shown, the cloud processing device 6 includes: a data receiving module 61 for receiving sensor data from the vehicle; a control quantity generation module 62 for processing the sensor data to generate at least two different types of control quantities; the control quantities are used to directly control the vehicle actuators; and a sending module 63 for sending the at least two control quantities to the vehicle so that the vehicle can simultaneously control at least two different vehicle actuators after parsing.

[0069] In some embodiments, the cloud processing device 6 can be configured such that the control quantity is generated by adjusting the internal parameters of the generation model based on scene data in the cloud; wherein the scene data includes one or more of real-time weather information, road type information, real-time traffic flow information, and traffic event information.

[0070] In some embodiments, the sending module 63 can be configured such that: the cloud selects a target link from multiple preset communication links based on the vehicle's real-time location information and network status information; the cloud switches the current communication link to the target link and sends a control signal through the target link.

[0071] Exemplary electronic devices An electronic device includes: a processor; a memory for storing processor-executable instructions; and a processor for executing the cloud-to-vehicle collaboration method of the embodiments provided in this application.

[0072] Below, for reference Figure 7 This application describes an electronic device according to embodiments thereof. The electronic device may be either or both of a first device and a second device, or a standalone device independent of them, which may communicate with the first device and the second device to receive acquired input signals from them.

[0073] Figure 7 A block diagram of an electronic device according to an embodiment of this application is illustrated.

[0074] like Figure 7 As shown, the electronic device 70 includes one or more processors 71 and a memory 72.

[0075] The processor 71 may be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and may control other components in the electronic device 70 to perform desired functions.

[0076] The memory 72 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 71 may execute the program instructions to implement the cloud-to-vehicle collaboration methods of the various embodiments of this application described above, and / or other desired functions. Various contents such as input signals, signal components, and noise components may also be stored in the computer-readable storage medium.

[0077] In one example, the electronic device 70 may also include an input device 73 and an output device 74, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0078] When the electronic device is a standalone device, the input device 73 can be a communication network connector for receiving the acquired input signals from the first device and the second device.

[0079] In addition, the input device 73 may also include, for example, a keyboard, a mouse, etc.

[0080] The output device 74 can output various information to the outside, including determined distance information, direction information, etc. The output device 74 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0081] Of course, for the sake of simplicity, Figure 7 Only some of the components of the electronic device 70 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 70 may include any other suitable components depending on the specific application.

[0082] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0083] A computer-readable storage medium stores a computer program for executing the cloud-vehicle collaboration method of the embodiments provided in this application.

[0084] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0085] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A cloud-vehicle collaboration method, characterized in that, Applications in vehicles include: Send vehicle-side sensor data to the cloud; Receive at least two different types of control quantities from the cloud, wherein the control quantities are generated by the cloud based on the sensor data processing and are used to directly control the vehicle actuators; Analyze at least two different types of the control variables and generate analysis results; Based on the analysis results, control signals are generated to simultaneously control at least two different vehicle actuators.

2. The cloud-vehicle collaboration method according to claim 1, characterized in that, The vehicle actuators include a steering system, a power system, and a braking system; The step of generating control signals based on the analysis results to simultaneously control at least two different vehicle actuators includes: Based on at least two types of the control quantities, control signals are generated for the coordinated control of the vehicle steering system, powertrain system, and braking system to perform at least two of the following operations: Adjust the vehicle's perception focus direction and recognition priority; Update vehicle route and speed planning; Adjust the vehicle's steering angle, throttle opening, and brake force ratio; Triggering emergency braking or evasive steering maneuvers of the vehicle; Adjust the vehicle's lane keeping, following distance, or traffic signal response behavior.

3. The cloud-vehicle collaboration method according to claim 1, characterized in that, After generating the control signal based on the parsing result, the cloud-vehicle collaboration method further includes: After executing the control signal, vehicle status data is transmitted back to the cloud so that the cloud can optimize and iterate the model used to generate the control quantity based on the vehicle status data.

4. A cloud processing method, characterized in that, include: Receive sensor data from the vehicle; Based on the sensor data, at least two different types of control quantities are generated; The control quantity is used to directly control the vehicle actuators; At least two control quantities are sent to the vehicle terminal so that the vehicle terminal can simultaneously control at least two different vehicle actuators after parsing.

5. The cloud processing method according to claim 4, characterized in that, The control quantity is generated by adjusting the internal parameters of the generation model based on the scene data in the cloud; wherein, the scene data includes one or more of the following: real-time weather information, road type information, real-time traffic flow information, and traffic event information.

6. The cloud processing method according to claim 4, characterized in that, Sending at least two control quantities to the vehicle includes: The cloud selects the target link from multiple preset communication links based on the vehicle's real-time location information and network status information; The cloud switches the current communication link to the target link and sends the control quantity through the target link.

7. A cloud-vehicle collaborative system, applicable to the cloud-vehicle collaborative method according to any one of claims 1-3, characterized in that, include: A cloud processing module, configured to receive sensor data from the vehicle; Based on the received sensor data, the system processes the data to generate at least two different types of control quantities; these control quantities are used to directly control the vehicle actuators; and the at least two control quantities are sent to the vehicle. The vehicle-side execution module is communicatively connected to the cloud processing module. The vehicle-side execution module is configured to: send sensor data from the vehicle to the cloud; receive and parse at least two control quantities from the cloud processing module and generate a parsing result; and generate a control signal based on the parsing result to simultaneously control the at least two different vehicle actuators.

8. The cloud-to-vehicle collaborative system according to claim 7, characterized in that, The control quantity is a set of instructions or parameters that can be directly mapped and converted into underlying drive signals by the vehicle-side execution module; the underlying drive signals are used to drive at least two vehicle actuators in the vehicle's steering system, drive system, and braking system.

9. The cloud-to-vehicle collaborative system according to claim 7, characterized in that, The cloud processing module adopts a multi-node redundant computing architecture, which is configured to switch computing tasks to other available nodes to continue execution when a single computing node fails.

10. A vehicle, characterized in that, include: Powertrain, steering system, and braking system; as well as The cloud-vehicle collaborative system as described in any one of claims 7-9, wherein the vehicle-side execution module of the cloud-vehicle collaborative system is communicatively connected to the power system, the steering system and the braking system, and is used to control at least two vehicle actuators in the power system, the steering system and the braking system based on control signals.

Images