A load balancing method and system based on vehicle-road cloud cooperation early prediction

Abstract

The present application relates to the technical field of vehicle-mounted computing resource scheduling, and specifically relates to a load balancing method and system based on vehicle-road cloud cooperation and advance prediction, comprising: acquiring real-time multi-dimensional road condition information in front of the vehicle through the whole vehicle cloud service; analyzing and extracting features of the information to generate a load influencing factor dataset; combining the current system resource state to predict future task load demand; when the demand exceeds a preset threshold, generating a scheduling strategy and adjustment instruction; dynamically adjusting the system resources according to the instruction to realize advance load balancing. The present application realizes predictive resource allocation by fusing cloud road conditions and local states, actively implements scheduling before the system resource is tight, guarantees the real-time performance and certainty of key tasks such as advanced auxiliary driving, significantly improves the utilization rate of computing resources, and provides a fundamental breakthrough for system stability and safety in high-dynamic driving environments.

Description

Technical Field

[0001] This invention relates to the field of vehicle-mounted computing resource scheduling technology, specifically to a load balancing method and system based on vehicle-road-cloud collaborative advance prediction. Background Technology

[0002] With the deepening development of the "software-defined vehicle" concept, automotive electronic and electrical architecture is undergoing a profound transformation from traditional distributed control to centralized computing. Under this trend, the functions of in-vehicle systems are becoming increasingly complex, especially in scenarios such as advanced driver assistance systems and autonomous driving, which place more stringent demands on the underlying operating system. The operating system not only needs to process massive amounts of data from multiple sensors (such as LiDAR, cameras, millimeter-wave radar, etc.), but also needs to complete the calculation and response of critical control tasks (such as emergency braking, lane avoidance, etc.) in a very short time. Any performance fluctuation may affect the real-time performance and safety of driving control.

[0003] Currently, most vehicle system load balancing technologies rely on dynamic monitoring mechanisms within the operating system. These mechanisms collect real-time data on CPU and memory usage and then adjust for detected load imbalances. While this approach can alleviate resource contention to some extent, its response is significantly delayed, making it ill-suited for the highly dynamic and real-time demands of driving environments. Furthermore, traditional load balancing strategies are typically limited to internal system states and fail to integrate external business information such as cloud-based traffic updates. This results in a lack of proactive resource allocation capabilities, impacting the certainty of critical task execution and leading to insufficient utilization of computing resources. Summary of the Invention

[0004] To address the above technical problems, this invention provides a technical solution for a load balancing method and system based on vehicle-road-cloud collaborative early prediction.

[0005] The technical problem solved by this invention can be achieved by the following technical solutions: A load balancing method based on vehicle-road-cloud collaborative early prediction includes: Step S1: Establish a communication connection between the vehicle and the vehicle cloud service, and obtain multi-dimensional road conditions and environmental information within a predetermined distance in front of the vehicle in real time from the vehicle cloud service. Step S2: Analyze and extract features from the multidimensional road condition and environmental information to generate a load impact factor dataset; Step S3: Based on the load impact factor dataset and the current resource usage status of the system, predict the task load demand within a predetermined time period in the future. Step S4: Compare the task load requirement with a preset load threshold. If the task load requirement exceeds the load threshold, generate a corresponding task scheduling strategy and resource adjustment instruction, and execute step S5; otherwise, return to step S2 for periodic detection. Step S5: Based on the task scheduling strategy and resource adjustment instructions, dynamically adjust the system resources to achieve early load balancing.

[0006] Preferably, step S2 is executed by the load model parser, and includes: Step S21: Perform data cleaning and verification on the multidimensional road condition and environmental information, including missing value processing, outlier value processing, and duplicate value processing. Step S22: Perform feature extraction and engineering processing on the cleaned data. The engineering processing includes feature scaling, discretization binning, and time feature processing. Step S23: The extracted features are classified and integrated according to preset influence dimensions to generate a structured load influence factor dataset; wherein, the influence dimensions include computational load, real-time demand, and resource consumption.

[0007] Preferably, step S3 is performed by a task load pre-allocator, including: Step S31: Query the current resource usage status of the system, including the number of tasks on each CPU computing core, CPU operating frequency, and memory usage. Step S32: Perform a fusion analysis of the load impact factor dataset and the current resource usage status of the system; Step S33: Based on the results of the fusion analysis, predict the task load requirements within the next 2 to 5 seconds.

[0008] Preferably, step S5 is executed through a load balancer and a resource manager, and includes: Based on the task scheduling strategy and resource adjustment instructions, the predicted trends of external road conditions and environment are analyzed. If the external road conditions and environment become more complex, resource reservation and load balancing operations are performed. If the external road conditions and environment become less complex, power consumption optimization operations are performed.

[0009] Preferably, in step S5, a data analysis model is used to conduct an in-depth analysis of the changing trends of the external road conditions and environment; The analysis model is trained based on historical vehicle data and corresponding multidimensional road condition and environmental data. By fusing multidimensional road condition and environmental information features, it outputs classification results to characterize the complexity of road conditions.

[0010] Preferably, the resource reservation and load balancing operation includes: Step S51: Increase the CPU frequency for running high-priority tasks through the resource manager, and reserve and lock memory resources for the high-priority tasks; Step S52: Determine whether the load of the CPU computing core where the high-priority task is located exceeds a preset threshold. If it does not exceed the threshold, end the current operation. If it does exceed the threshold, proceed to step S53. Step S53: Using the load balancer and considering the current external environment, identify low-priority tasks from the CPU computing cores within the next 2 to 5 seconds, and determine the target core for task migration. Step S54: Analyze the working frequency of the target core. If the working frequency of the target core is lower than a preset level, increase the working frequency of the target core through the resource manager and migrate the low-priority tasks to the target core through the load balancer.

[0011] Preferably, the power consumption optimization operation includes: The resource manager obtains the current operating frequency of all CPUs, identifies the CPU computing cores where low-priority tasks reside within the next 2 to 5 seconds, and reduces the operating frequency of those CPU computing cores.

[0012] Preferably, the priority of system tasks is pre-configured by the user according to the needs of different in-vehicle business applications.

[0013] Preferably, in step S1, the multidimensional road condition and environment information includes at least one of the following: road congestion ahead, traffic density, average vehicle speed, lane changes, weather conditions, intersection signal status, obstacle type, and road construction information.

[0014] This invention also provides a load balancing system based on vehicle-road-cloud cooperative advance prediction, which implements the load balancing method based on vehicle-road-cloud cooperative advance prediction as described above, including: The whole vehicle cloud service is used to push multi-dimensional road condition and environmental information to vehicles in real time based on cloud big data and roadside perception network; The load model parser, connected to the vehicle cloud service, is used to analyze and extract features from the pushed multi-dimensional road condition and environmental information to generate a load impact factor dataset. A task load pre-allocator, connected to the load model parser, is used to make predictions based on the load influencing factor dataset and generate task scheduling strategies and resource adjustment instructions. A load balancer, connected to the task load pre-allocator, is used to execute the task scheduling strategy to enable task migration between multiple CPU computing cores. The resource manager, connected to the task load pre-allocator, is used to execute the resource adjustment instructions to adjust the CPU operating frequency and memory resources.

[0015] The beneficial effects of this invention are as follows: By deeply integrating multi-dimensional road condition information from the cloud with the local system status, it can accurately predict future task load requirements and proactively implement scheduling strategies before system resources become scarce. This predictive resource allocation not only ensures the real-time performance and determinism of critical tasks such as advanced driver assistance systems, but also significantly improves the utilization rate of computing resources, providing a fundamental breakthrough for system stability and safety in highly dynamic driving environments. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a block diagram of step S2 of the present invention; Figure 3 This is a block diagram of step S3 of the present invention; Figure 4 This is a flowchart of the task scheduling and resource adjustment process of the present invention; Figure 5 This is a flowchart of the resource reservation and load balancing operation of the present invention; Figure 6 This is a system architecture diagram of the present invention; Figure 7 This is a high load balancing block diagram of existing technologies; Figure 8 This is the pre-prediction equilibrium block diagram of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0020] Reference Figure 1 This invention provides a load balancing method based on vehicle-road-cloud collaborative early prediction, comprising: Step S1: Establish a communication connection between the vehicle and the vehicle cloud service, and obtain multi-dimensional road conditions and environmental information within a predetermined distance in front of the vehicle in real time from the vehicle cloud service. Step S2: Analyze and extract features from the multidimensional road condition and environmental information to generate a load impact factor dataset; Step S3: Based on the load impact factor dataset and the current resource usage status of the system, predict the task load demand within a predetermined time period in the future. Step S4: Compare the task load requirement with a preset load threshold. If the task load requirement exceeds the load threshold, generate a corresponding task scheduling strategy and resource adjustment instruction, and execute step S5; otherwise, return to step S2 for periodic detection. Step S5: Based on the task scheduling strategy and resource adjustment instructions, dynamically adjust the system resources to achieve early load balancing.

[0021] Specifically, in this embodiment of the invention, in response to the technical problems of traditional load balancing technology in high-dynamic driving environments, such as slow response and lack of foresight, resulting in the inability to guarantee the real-time performance of critical tasks and low resource utilization, a vehicle-road-cloud collaborative mechanism is introduced. This mechanism utilizes the rich prior information provided by the cloud to predict and actively schedule the computing load in advance, avoiding performance fluctuations and security risks caused by passive responses within the system. This achieves a paradigm shift from "passive response" to "active guarantee," ensuring the determinism of critical task execution and optimizing overall resource efficiency.

[0022] Specifically, the Vehicle Cloud Service (VCS), as the core of the vehicle-road-cloud collaborative system, is responsible for aggregating, processing, and distributing real-time and predictive information from roadside facilities (RSUs), traffic management centers, and other vehicles (V2X).

[0023] Accordingly, this real-time and predictive information includes road congestion ahead, traffic density, average vehicle speed, lane changes, weather conditions, intersection signal status, obstacle types, and road construction information. By analyzing this information, the system can accurately predict the evolution of driving scenarios. For example, recognizing the feature of "a sudden increase in traffic density and a decrease in average vehicle speed 2 kilometers ahead" can predict a significant increase in the data processing load of the perception module; while the information of "approaching an intersection without traffic lights" indicates an increase in the computational complexity of the decision-making and planning module.

[0024] Based on these quantitative predictions, the system can reserve and allocate computing resources smoothly and in advance before the vehicle actually enters these high-load scenarios, thereby completely avoiding task delays or performance drops caused by resource competition in traditional solutions, and providing deterministic and highly reliable underlying computing support for high-level autonomous driving functions.

[0025] In a preferred embodiment of the present invention, the priority of system tasks is pre-configured by the user according to the needs of different in-vehicle business applications.

[0026] Specifically, in this embodiment of the invention, the system divides tasks into multiple key levels according to pre-configured priority rules. For example, tasks involving direct vehicle control, such as emergency braking and lane keeping, are set as high priority to ensure absolute priority of computing resources; while non-real-time tasks such as infotainment and data logging are set as low priority. When performing forward-looking resource allocation, the system prioritizes the resource needs of high-priority tasks and can intelligently restrict or delay the execution of low-priority tasks when resource shortages are predicted, thereby achieving optimal allocation of computing resources under complex road conditions and ensuring the core requirements of driving safety and system functionality.

[0027] As a preferred embodiment of the present invention, refer to Figure 2 Step S2 is executed by the Load Model Parser (LMP) and includes: Step S21: Perform data cleaning and verification on the multidimensional road condition and environmental information, including handling missing values, outlier values, and duplicate values. Step S22: Perform feature extraction and engineering processing on the cleaned data. The engineering processing includes feature scaling, discretization binning, and time feature processing. Step S23: The extracted features are classified and integrated according to preset influence dimensions to generate a structured load influence factor dataset; wherein, the influence dimensions include computational load, real-time demand, and resource consumption, etc.

[0028] Specifically, in this embodiment of the invention, LMP transforms raw, heterogeneous cloud information into load impact factor characteristics that the system can understand by constructing a standardized data processing pipeline. Its workflow is as follows: During the data cleaning phase, LMP uses interpolation to complete missing traffic flow density data and identifies and removes abnormal speed readings caused by sensor errors based on statistical models. For example, when network fluctuations cause the loss of traffic flow density data for a certain road segment, LMP will perform linear interpolation based on data from adjacent road segments to generate reasonable complete values. Simultaneously, for obvious errors such as instantaneous speeds of 500 km / h caused by radar data anomalies, the system will identify them as outliers based on a historical normal data distribution model and remove them. This step ensures the integrity and reliability of the input data, laying a solid foundation for accurate prediction.

[0029] During the feature engineering phase, LMP discretizes the continuous "average vehicle speed" into bins, such as 0-20 km / h representing "severe congestion," and extracts temporal features such as "time until the next traffic light." For example, when the system detects a traffic light 300 meters ahead that is currently red, LMP combines the vehicle's current position, speed, and traffic light timing to accurately calculate the key time feature of "25 seconds remaining until the green light turns on." This process not only reduces the complexity of model calculations but also transforms abstract physical quantities into feature identifiers that directly reflect the system's load status.

[0030] Ultimately, LMP categorizes features such as "traffic density" and "number of obstacles" into "computational load," features such as "traffic light state switching countdown" into "real-time demand," and features such as "perception algorithm mode (e.g., anti-interference mode required in severe weather)" into "resource consumption," thus building a complete and structured input foundation for subsequent accurate load prediction. This multi-dimensional feature classification system enables the prediction model to deeply understand the differentiated impact of different road conditions on system resources, achieving a leap from perception to cognition. This provides accurate and quantifiable decision-making basis for forward-looking resource scheduling, fundamentally improving the system's intelligence and response efficiency.

[0031] As a preferred embodiment of the present invention, refer to Figure 3 Step S3 is executed by the Task Load Predictor (TLP) and includes: Step S31: Query the current resource usage status of the system, including the number of tasks on each CPU computing core, CPU operating frequency, and memory usage. Step S32: Perform a fusion analysis of the load impact factor dataset and the current resource usage status of the system; Step S33: Based on the results of the fusion analysis, predict the task load requirements within the next 2 to 5 seconds.

[0032] Specifically, because critical tasks such as advanced driver assistance systems (ADAS) have extremely high real-time requirements for system response, their decision-making and control cycles are typically on the order of hundreds of milliseconds. Therefore, predicting the load window for the next 2 to 5 seconds provides the system with sufficient reaction time to smoothly adjust and pre-allocate resources. This effectively addresses unexpected road conditions while avoiding the uncertainties associated with overly long-term predictions. In this embodiment of the invention, TLP achieves accurate prediction through a lightweight time-series prediction model. This model takes fused structured data (load influencing factors and real-time resource status) as input, and its core logic lies in identifying the mapping relationship between external road condition characteristics and internal resource consumption. For example, when "computational load" factors (such as traffic density) increase sharply and the system detects that the perception task queue begins to grow, the model predicts that the load on the visual computing module will continue to rise in the next few seconds. At the same time, if "real-time demand" factors (such as a traffic light about to turn red) indicate an increase in urgency, the model will further increase the predicted computational resource demand for the decision-making and planning tasks. Ultimately, TLP outputs a quantitative, task-type-specific future resource demand, providing precise data support for subsequent threshold comparisons and resource scheduling.

[0033] As a preferred embodiment of the present invention, refer to Figure 4 Step S5 is executed through a load balancer (LB) and a resource manager (RM), and includes: Based on the task scheduling strategy and resource adjustment instructions, the predicted trends of external road conditions and environment are analyzed. If the external road conditions and environment become more complex, resource reservation and load balancing operations are performed. If the external road conditions and environment become less complex, power consumption optimization operations are performed.

[0034] Specifically, in this embodiment of the invention, the LB and RM work together to form the decision-making and execution center of the system. Based on the task scheduling strategy and resource adjustment instructions issued by the TLP, and supplemented by a lightweight data analysis model, they perform in-depth analysis of the predicted external road conditions and environmental change trends.

[0035] The analysis model is trained based on historical vehicle data and corresponding multidimensional road condition and environmental data. By fusing multidimensional road condition and environmental information features, it outputs a classification result that characterizes the complexity of road conditions.

[0036] Reference Figure 4 and Figure 5 The resource reservation and load balancing operations include: Step S51: Increase the CPU frequency for running high-priority tasks through the resource manager, and reserve and lock memory resources for the high-priority tasks; Step S52: Determine whether the load of the CPU computing core where the high-priority task is located exceeds a preset threshold. If it does not exceed the threshold, end the current operation. If it does exceed the threshold, proceed to step S53. Step S53: Using the load balancer and considering the current external environment, identify low-priority tasks from the CPU computing cores within the next 2 to 5 seconds, and determine the target core for task migration. Step S54: Analyze the working frequency of the target core. If the working frequency of the target core is lower than a preset level, increase the working frequency of the target core through the resource manager and migrate the low-priority tasks to the target core through the load balancer.

[0037] Reference Figure 4 The power consumption optimization operation includes: The resource manager obtains the current operating frequency of all CPUs, identifies the CPU computing cores where low-priority tasks reside within the next 2 to 5 seconds, and reduces the operating frequency of those CPU computing cores.

[0038] As can be seen, through the above-mentioned collaborative control mechanism, the load balancer (LB) and resource manager (RM) have achieved a paradigm shift from "resource competition" to "resource planning", thus achieving a dynamic optimal balance between system performance and energy efficiency.

[0039] This design enables the system to proactively shape its internal resource state based on deterministic predictions of the external environment. When the environment becomes complex, a closed-loop operation of "resource reservation - load assessment - frequency boost - task migration" builds a protected, high-performance execution environment for core safety tasks, ensuring the functional safety and real-time performance of the vehicle. When the environment becomes less demanding, the frequency of non-critical cores can be reduced in a timely manner to avoid unnecessary idling of computing resources, significantly optimizing the overall power consumption of the system and extending the driving range of electric vehicles.

[0040] Reference Figure 6 The present invention also provides a load balancing system based on vehicle-road-cloud cooperative advance prediction, which implements the load balancing method based on vehicle-road-cloud cooperative advance prediction as described above, including: Vehicle Cloud Service (VCS) is used to push multi-dimensional road condition and environmental information to vehicles in real time based on cloud big data and roadside perception networks. The Load Model Parser (LMP) is connected to the vehicle cloud service and is used to parse and extract features from the pushed multi-dimensional road condition and environmental information to generate a load impact factor dataset. The Task Load Predictor (TLP), connected to the load model parser, is used to make predictions based on the load influencing factor dataset and generate task scheduling strategies and resource adjustment instructions. A load balancer (LB) is connected to the task load pre-allocator and is used to execute the task scheduling strategy to enable the migration of tasks between multiple CPU computing cores. The Resource Manager (RM) is connected to the task load pre-allocator and is used to execute the resource adjustment instructions to adjust the CPU operating frequency and memory resources.

[0041] Specifically, in this embodiment of the invention, the system constructs a complete "perception-prediction-decision-execution" closed loop through the collaborative work of five core modules: VCS, LMP, TLP, LB, and RM. VCS provides global environmental perception input, LMP is responsible for transforming raw information into load characteristics understandable to the system, TLP performs forward-looking load prediction and strategy generation based on feature data, and finally, LB and RM are responsible for the fine-grained scheduling of computing tasks and hardware resources, respectively. This architecture achieves end-to-end linkage from cloud information to onboard computing resources, enabling the system to proactively and accurately allocate resources according to future driving scenarios, thereby ensuring the real-time performance of critical tasks, the overall efficiency of the system, and the economy of operation under complex road conditions.

[0042] In summary, this invention provides a load balancing method and system based on vehicle-road-cloud collaboration for advance prediction. By introducing a vehicle-road-cloud collaborative mechanism, a complete technical closed loop of "perception-prediction-decision-execution" is constructed. Figure 7 As shown, in high-load scenarios, existing technologies result in some CPU cores experiencing severe task overload, while other cores remain idle. This passive response mode inevitably leads to a lag in load balancing. In contrast, as... Figure 8 As shown, this invention, through advance acquisition and intelligent analysis of cloud-based traffic information, can pre-allocate resources and intelligently schedule tasks before vehicles enter high-load areas, making the load distribution of each computing core more balanced and reasonable. This fundamentally solves the response delay problem of traditional load balancing technology, realizes a paradigm shift from passive response to proactive protection, and significantly improves the real-time performance, safety, and resource utilization efficiency of the vehicle system in complex driving environments.

[0043] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A load balancing method based on vehicle-road-cloud collaborative early prediction, characterized in that, include: Step S1: Establish a communication connection between the vehicle and the vehicle cloud service, and obtain multi-dimensional road conditions and environmental information within a predetermined distance in front of the vehicle in real time from the vehicle cloud service. Step S2: Analyze and extract features from the multidimensional road condition and environmental information to generate a load impact factor dataset; Step S3: Based on the load impact factor dataset and the current resource usage status of the system, predict the task load demand within a predetermined time period in the future. Step S4: Compare the task load requirement with a preset load threshold. If the task load requirement exceeds the load threshold, generate a corresponding task scheduling strategy and resource adjustment instruction, and execute step S5; otherwise, return to step S2 for periodic detection. Step S5: Based on the task scheduling strategy and resource adjustment instructions, dynamically adjust the system resources to achieve early load balancing. Step S2 is executed by the load model parser and includes: Step S21: Perform data cleaning and verification on the multidimensional road condition and environmental information, including missing value processing, outlier value processing, and duplicate value processing. Step S22: Perform feature extraction and engineering processing on the cleaned data. The engineering processing includes feature scaling, discretization binning, and time feature processing. Step S23: The extracted features are classified and integrated according to preset influence dimensions to generate a structured load influence factor dataset; wherein, the influence dimensions include computational load, real-time demand, and resource consumption.

2. The load balancing method based on vehicle-road-cloud collaborative early prediction according to claim 1, characterized in that, Step S3 is executed through the task load pre-allocator, including: Step S31: Query the current resource usage status of the system, including the number of tasks on each CPU computing core, CPU operating frequency, and memory usage. Step S32: Perform a fusion analysis of the load impact factor dataset and the current resource usage status of the system; Step S33: Based on the results of the fusion analysis, predict the task load requirements within the next 2 to 5 seconds.

3. The load balancing method based on vehicle-road-cloud collaborative early prediction according to claim 1, characterized in that, Step S5 is executed through a load balancer and a resource manager, and includes: Based on the task scheduling strategy and resource adjustment instructions, the predicted trends of external road conditions and environment are analyzed. If the external road conditions and environment become more complex, resource reservation and load balancing operations are performed. If the external road conditions and environment become less complex, power consumption optimization operations are performed.

4. The load balancing method based on vehicle-road-cloud collaborative early prediction according to claim 3, characterized in that, In step S5, a data analysis model is used to conduct an in-depth analysis of the changing trends of the external road conditions and environment. The analysis model is trained based on historical vehicle data and corresponding multidimensional road condition and environmental data. By fusing multidimensional road condition and environmental information features, it outputs classification results to characterize the complexity of road conditions.

5. A load balancing method based on vehicle-road-cloud collaborative early prediction as described in claim 3, characterized in that, The resource reservation and load balancing operations include: Step S51: Increase the CPU frequency for running high-priority tasks through the resource manager, and reserve and lock memory resources for the high-priority tasks; Step S52: Determine whether the load of the CPU computing core where the high-priority task is located exceeds a preset threshold. If it does not exceed the threshold, end the current operation. If it does exceed the threshold, proceed to step S53. Step S53: Using the load balancer and considering the current external environment, identify low-priority tasks from the CPU computing cores within the next 2 to 5 seconds, and determine the target core for task migration. Step S54: Analyze the working frequency of the target core. If the working frequency of the target core is lower than a preset level, increase the working frequency of the target core through the resource manager and migrate the low-priority tasks to the target core through the load balancer.

6. The load balancing method based on vehicle-road-cloud collaborative early prediction according to claim 5, characterized in that, The power consumption optimization operation includes: The resource manager obtains the current operating frequency of all CPUs, identifies the CPU computing cores where low-priority tasks reside within the next 2 to 5 seconds, and reduces the operating frequency of those CPU computing cores.

7. A load balancing method based on vehicle-road-cloud collaborative early prediction as described in claim 6, characterized in that, The priority of system tasks is pre-configured by the user according to the needs of different in-vehicle business applications.

8. The load balancing method based on vehicle-road-cloud collaborative early prediction according to claim 1, characterized in that, In step S1, the multidimensional road condition and environment information includes at least one of the following: road congestion ahead, traffic density, average vehicle speed, lane line changes, weather conditions, intersection signal status, obstacle type, and road construction information.

9. A load balancing system based on vehicle-road-cloud collaborative early prediction, characterized in that, An application implementing a load balancing method based on vehicle-road-cloud collaboration for advance prediction as described in any one of claims 1-8 includes: The whole vehicle cloud service is used to push multi-dimensional road condition and environmental information to vehicles in real time based on cloud big data and roadside perception network; The load model parser, connected to the vehicle cloud service, is used to analyze and extract features from the pushed multi-dimensional road condition and environmental information to generate a load impact factor dataset. A task load pre-allocator, connected to the load model parser, is used to make predictions based on the load influencing factor dataset and generate task scheduling strategies and resource adjustment instructions. A load balancer, connected to the task load pre-allocator, is used to execute the task scheduling strategy to enable task migration between multiple CPU computing cores. The resource manager, connected to the task load pre-allocator, is used to execute the resource adjustment instructions to adjust the CPU operating frequency and memory resources.

Images