Dynamic load distribution method, device and equipment based on vehicle body controller and medium

Abstract

The present application relates to a kind of dynamic load distribution method based on vehicle body controller, device, equipment and medium, method includes: by distributed sensor network real-time acquisition vehicle body mechanical load, electrical load and thermal load data, generate multi-source initial load data of heterogeneous;Initial load data is standardized, and three-dimensional data cube type standard load data set is constructed;By feature space mapping extraction comprehensive representation feature, dynamically generate fusion load feature set containing threshold range in combination with historical data;Based on fusion feature, load correlation model is constructed, and steering and brake control parameter combination is generated by mapping;According to real-time load threshold out-of-range state, dynamically adjust task priority, for chassis system distribution real-time computing core resource and generate optimization allocation scheme.The method solves the multi-source load data fusion failure in the prior art, resource allocation rigidification and safety task response delay problem, realizes the dynamic load optimization of vehicle body control system under complex working condition.

Description

Technical Field

[0001] This invention relates to the field of vehicle body dynamic load control technology, specifically to a dynamic load distribution method, device, equipment, and medium based on a vehicle body controller. Background Technology

[0002] In modern intelligent vehicle systems, the body controller, as the core hub of the vehicle's electronic architecture, undertakes the crucial task of coordinating the powertrain, chassis system, and various electronic devices. With the increasing complexity of onboard functional modules, the electrical, mechanical, and thermal loads generated during vehicle operation exhibit multi-source heterogeneous characteristics. The real-time collaborative processing capability of this load data directly impacts vehicle handling performance and safety. Current industry-standard load allocation methods primarily rely on static resource scheduling using preset rules, which struggles to cope with the dynamic fluctuations in vehicle load states under complex operating conditions. Especially in extreme scenarios such as high-speed cornering and emergency braking, the coupling effects of various load parameters often lead to control system response lags and even trigger abnormal execution of critical safety functions.

[0003] Traditional load distribution schemes have significant limitations: First, they lack effective means of integrating monitoring data from multiple dimensions such as mechanical vibration, electrical fluctuations, and temperature changes, making it impossible for the control system to establish a precise mapping relationship between load status and execution parameters. Second, the resource allocation mechanism is rigid; when scenarios such as a surge in steering assist demand or sudden changes in regenerative braking occur, it cannot dynamically adjust the allocation strategy of processor core resources based on real-time load characteristics. Third, the priority management of critical safety tasks and non-real-time tasks lacks quantitative basis, which may cause chassis control response delays when system resources are strained. These problems are particularly prominent in the high-power electronic and electrical architecture of new energy vehicles, manifesting as safety hazards such as steering system response timeouts and inaccurate regenerative braking torque distribution, severely restricting further improvements in vehicle dynamic control performance. Summary of the Invention

[0004] Based on this, the purpose of the present invention is to provide a dynamic load allocation method, device, equipment and medium based on a vehicle body controller that can integrate multi-source load characteristics in real time and dynamically optimize processor resource allocation strategies.

[0005] The objective of this invention is achieved through the following solution:

[0006] In a first aspect, the present invention provides a dynamic load distribution method based on a vehicle body controller, comprising the following steps:

[0007] S1: Data is collected from various components of the vehicle body based on a distributed sensor network, including current and voltage data of electrical loads, force and acceleration data of mechanical loads, and temperature and heat flow data of thermal loads, to generate initial load data;

[0008] S2: Preprocess the initial load data to eliminate the dimensional differences between mechanical load, electrical load and thermal load, and generate a standard load dataset;

[0009] S3: Perform unified feature space mapping on the standard load dataset, extract comprehensive characterization features of mechanical load, electrical load and thermal load, and generate feature threshold range based on historical data distribution to generate a fused load feature set containing fused feature vector and preset threshold range;

[0010] S4: Construct a load association model of load features and control parameters based on the fused load feature set, and input the fused feature vector into the load association model for parameter mapping processing to generate a combination of mapped control parameters;

[0011] S5: Based on the preset threshold range of the fused load feature set, dynamically adjust the load of the mapping control parameter combination, adjust the task execution order and resource allocation ratio based on the preset load priority rules, generate an optimized load allocation scheme, and use the optimized load allocation scheme to instruct the CPU core allocation strategy and control the body controller to make real-time parameter adjustments to the power system and chassis system.

[0012] Secondly, the present invention provides a dynamic load distribution device based on a vehicle body controller, the device being configured with the following modules:

[0013] The data acquisition module is used to collect data from various components of the vehicle body based on a distributed sensor network. It collects current and voltage data of electrical loads, force and acceleration data of mechanical loads, and temperature and heat flow data of thermal loads to generate initial load data.

[0014] The data preprocessing module is used to preprocess the initial load data, eliminate the dimensional differences between mechanical load, electrical load and thermal load, and generate a standard load dataset.

[0015] The fusion load feature generation module is used to perform unified feature space mapping on the standard load dataset, extract comprehensive characterization features of mechanical load, electrical load and thermal load, and generate feature threshold ranges based on historical data distribution, generating a fusion load feature set containing fusion feature vectors and preset threshold ranges.

[0016] The load association model mapping module is used to construct a load association model of load features and control parameters based on the fused load feature set, and input the fused feature vector into the load association model for parameter mapping processing to generate a combination of mapped control parameters;

[0017] The dynamic load optimization and allocation module is used to dynamically adjust the load of the mapping control parameter combination based on the preset threshold range of the fused load feature set, adjust the task execution order and resource allocation ratio based on the preset load priority rules, and generate an optimized load allocation scheme. The optimized load allocation scheme is used to instruct the CPU core allocation strategy and control the body controller to adjust the parameters of the power system and chassis system in real time.

[0018] Thirdly, this application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the above-mentioned dynamic load distribution methods based on the vehicle body controller.

[0019] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-described dynamic load distribution methods based on a vehicle body controller.

[0020] In summary,

[0021] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0022] Figure 1 A flowchart illustrating a dynamic load distribution method based on a vehicle body controller, provided for an embodiment of this application;

[0023] Figure 2 A schematic diagram illustrating the process of generating an optimized load allocation scheme provided in an embodiment of this application;

[0024] Figure 3 A schematic diagram illustrating the process of updating the load association model provided in an embodiment of this application;

[0025] Figure 4 This is a schematic diagram of a dynamic load distribution device based on a vehicle body controller, provided as another embodiment of this application. Detailed Implementation

[0026] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0028] In one embodiment, such as Figure 1 As shown, a dynamic load distribution method based on a vehicle body controller is provided. This embodiment illustrates the method by applying it to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0029] S1: Data is collected from various components of the vehicle body based on a distributed sensor network, including current and voltage data of electrical loads, force and acceleration data of mechanical loads, and temperature and heat flow data of thermal loads, to generate initial load data.

[0030] Specifically, in the vehicle's electronic architecture, a pre-deployed distributed sensor network covers various key components of the vehicle body, such as the motor and battery in the powertrain system, and the steering mechanism, braking system, and various electronic devices in the chassis system. For the electrical load, high-precision current transformers and voltage sensors are used to collect the current intensity flowing through each electrical component and the voltage value across the component in real time. The sampling frequency can be set according to actual needs, typically between hundreds of hertz and thousands of hertz, to ensure that subtle changes in the electrical load can be captured. For the mechanical load, corresponding force sensors and acceleration sensors are installed. Force sensors, for example, are installed at key connection points in the suspension system to measure the forces exerted on the vehicle from the road surface during driving. Acceleration sensors are deployed at key components of the vehicle body and chassis to acquire vehicle acceleration information. This data reflects the dynamic loads borne by the vehicle's mechanical structure under different driving conditions. Regarding the thermal load, temperature sensors and heat flux sensors are configured. Temperature sensors are distributed around components that easily generate heat, such as motors and battery packs, to monitor their temperature changes in real time. Heat flux sensors are used to detect the rate of heat transfer between the components and the surrounding environment during operation, i.e., heat flux density.

[0031] Each sensor collects corresponding physical quantities in real time at a fixed sampling frequency. The sampling frequency is set according to different load types and vehicle operating conditions. For example, under dynamic conditions such as rapid acceleration and deceleration, where electrical, mechanical, and thermal loads change rapidly, the sampling frequency is increased to 1000Hz to ensure data timeliness and integrity. Under steady-state conditions such as constant speed driving, the sampling frequency is set to 100Hz to reduce data processing pressure and power consumption. The collected data is transmitted to the body control module (BCM) via vehicle network buses such as CAN bus and FlexRay bus. The data transmission rate of CAN bus can reach up to 1Mbps, and the data transmission rate of FlexRay bus can reach up to 10Mbps, ensuring fast and stable data transmission, thereby generating initial load data containing the original data of electrical, mechanical, and thermal loads.

[0032] S2: Preprocess the initial load data to eliminate the dimensional differences between mechanical load, electrical load and thermal load, and generate a standard load dataset.

[0033] Specifically, because the data for mechanical load, electrical load, and thermal load originate from different physical quantities, each with its own unique units and dimensions—such as amperes for current, volts for voltage, newtons for force, meters per second squared for acceleration, degrees Celsius for temperature, and watts per square meter for heat flux—this difference makes direct comprehensive analysis and processing of the data difficult. Therefore, the system employs a unified normalization operation for all types of data. Specifically, based on the historical statistical characteristics of each data type, the system determines a reasonable mapping interval and accurately maps the original data to this interval using linear or nonlinear transformation functions. For example, for current data, the system references the current fluctuation range of the vehicle's electrical system under normal operating conditions and converts it into a standardized value between 0 and 1. Similarly, for the force and acceleration data of mechanical loads, corresponding standardized values ​​are assigned based on the mechanical performance indicators determined during the vehicle design phase. During this process, the system identifies and removes abnormal data points that significantly deviate from the normal range. These abnormal data points may originate from erroneous readings caused by sensor malfunctions or temporary interference.

[0034] S3: Perform unified feature space mapping on the standard load dataset, extract comprehensive characterization features of mechanical load, electrical load and thermal load, and generate feature threshold range based on historical data distribution, generating a fused load feature set containing fused feature vector and preset threshold range.

[0035] Specifically, for mechanical loads, the system can use short-time Fourier transform to extract their time-frequency domain features, including the frequency distribution of force signals and the amplitude modulation features of acceleration signals. For electrical loads, the system can obtain the time-domain statistical features of current and voltage through sliding window mean and variance calculations, and simultaneously use wavelet packet decomposition to extract their frequency-domain energy distribution features. Thermal load data is analyzed based on the heat conduction equation to extract physical features such as the rate of temperature change and heat flux density gradient. The extracted multi-dimensional original feature vectors are then processed using principal component analysis dimensionality reduction algorithms to select key features that are highly sensitive to changes in load state and have strong correlations, forming a preliminary fusion feature vector. Based on this, the system integrates the distribution characteristics of historical load data and uses Gaussian mixture model clustering analysis to determine the feature threshold range of each load type under different operating conditions. For example, for the power steering electrical load of new energy vehicles under high-speed conditions, based on the current peak distribution of historical high-speed driving scenarios, the current threshold of the 95% confidence interval is set as [15A, 35A]. Correspondingly, the steering knuckle torque threshold on the mechanical load side is set as [800N, 1200N], and the motor controller temperature threshold on the thermal load side is set as [65℃, 85℃].

[0036] S4: Construct a load association model of load features and control parameters based on the fused load feature set, and input the fused feature vector into the load association model for parameter mapping processing to generate a combination of mapped control parameters.

[0037] Specifically, the load correlation model is used to establish a quantitative mapping relationship between fused load features and vehicle body control parameters. The model structure adopts a deep neural network architecture. The input layer nodes correspond to the dimensions of the fused load feature vector, including key features such as the force frequency and acceleration amplitude of mechanical loads, the mean current and voltage variance of electrical loads, and the temperature gradient of thermal loads. The hidden layer uses a combination of multiple nonlinear activation functions such as ReLU and sigmoid to mine the implicit correlation patterns between load features and control parameters. The output layer nodes are mapped to the core control parameters of the vehicle power system and chassis system, such as electronic throttle opening, electric power steering ratio, and regenerative braking energy recovery ratio.

[0038] During model training, the system uses a large amount of historical load data and corresponding actual control parameter samples as a basis, and applies a gradient descent optimization algorithm to adjust the network weights, so that the error loss function between the model's output mapped control parameter combination and the optimal control parameters under actual operating conditions gradually converges to a minimum. When the real-time acquired fused feature vector is input into this load-related model, it generates a mapped control parameter combination that accurately matches the current load state through fast forward propagation calculation of the deep neural network. This provides real-time, intelligent parameter decision support for vehicle dynamic control, effectively improving the vehicle's adaptive control performance under complex operating conditions.

[0039] S5: Based on the preset threshold range of the fused load feature set, dynamically adjust the load of the mapping control parameter combination, adjust the task execution order and resource allocation ratio based on the preset load priority rules, generate an optimized load allocation scheme, and use the optimized load allocation scheme to instruct the CPU core allocation strategy and control the body controller to make real-time parameter adjustments to the power system and chassis system.

[0040] Specifically, relying on the preset threshold range contained in the fused load feature set, the system dynamically adjusts and optimizes the combination of control parameters generated by mapping. Specifically, the system compares the current mapped control parameters with the threshold boundaries defined by preset load priority rules in real time. For example, in an emergency braking scenario, if the thermal load temperature of the braking system exceeds the preset high-temperature threshold, and the mechanical load value fed back by the brake pressure sensor is close to its limit, the highest-priority braking safety task protection mechanism is triggered. The system reorders the task execution order of the powertrain and chassis systems according to preset load priority rules, prioritizing the energy supply and control response of the brake assist system, and dynamically downgrading or suspending non-critical comfort electrical loads such as seat heating and window defrosting. Regarding resource allocation ratio adjustment, by dynamically adjusting the CPU core allocation strategy, more computing core resources are tilted towards tasks responsible for braking control and vehicle stability system calculations. For example, 30% of the CPU core resources originally allocated to the in-vehicle entertainment system are urgently allocated to the chassis control module to ensure the real-time performance and stability of critical safety control tasks. The optimized load distribution scheme is output in the form of system commands, directly instructing the body controller to adjust the parameters of the power system and chassis system in real time, in order to adjust the engine torque output, optimize the motor speed, adjust the suspension damping coefficient, and distribute the braking torque, thereby achieving load balance and performance optimization of the vehicle under complex dynamic conditions.

[0041] In summary, the dynamic load distribution method based on the vehicle body controller provided in this application embodiment achieves the desired results.

[0042] In one embodiment, S1 of the dynamic load distribution method based on a vehicle body controller provided by the present invention specifically includes the following steps:

[0043] S11: Perform stress distribution monitoring on the vehicle body structural components. Collect vibration spectrum data of key suspension nodes through strain sensors on the longitudinal beams of the frame to generate raw datasets of mechanical loads.

[0044] Specifically, the system precisely deploys strain sensors on key structural components such as the longitudinal beams of the vehicle frame. These sensors sample vibration spectrum data from critical suspension nodes at high frequencies, including structural vibration information caused by factors such as uneven road surfaces, changes in vehicle speed, and uneven load distribution during vehicle operation. The sensors convert the collected strain signals into electrical signals and transmit them to the signal acquisition module of the body controller via shielded cables to ensure signal stability and anti-interference during transmission. The system performs preliminary filtering on these raw strain signals to remove high-frequency noise and low-frequency drift, retaining effective frequency components related to the vehicle's dynamic response. After filtering, the strain signals are further converted into stress values. Using material mechanics models and vehicle structural parameters, the stress distribution of structural components such as the longitudinal beams of the vehicle frame under different operating conditions is calculated. This stress data not only reflects the mechanical load state of the vehicle during operation but also provides a foundation for subsequent mechanical load feature extraction and analysis.

[0045] S12: Real-time parameter acquisition and processing of the vehicle's high-voltage electrical system, obtaining battery management system discharge current ripple and motor controller phase voltage fluctuation data via CAN bus, and generating raw electrical load dataset.

[0046] Specifically, the system utilizes the efficient data transmission capabilities of the CAN bus to acquire real-time data on the discharge current ripple of the battery management system and the phase voltage fluctuations of the motor controller. The battery management system monitors the discharge current of the battery pack through high-precision current sensors. The amplitude and frequency of the ripple current reflect the dynamic characteristics of the battery during discharge, including the internal chemical reaction rate, changes in contact resistance, and fluctuations in the external load. The motor controller monitors the phase voltage fluctuations in real-time through voltage sensors. This fluctuation data contains important information such as the motor's operating status, the switching characteristics of the controller's power module, and the dynamic response of the vehicle's powertrain.

[0047] The system digitizes the collected current ripple and phase voltage fluctuation data, converting analog signals to digital signals via an analog-to-digital converter, and then performs data packetization and frame synchronization to ensure reliable data transmission on the CAN bus. Upon receiving this data, the system uses signal processing algorithms to perform preliminary analysis, extracting key characteristic parameters such as the amplitude and frequency of the current ripple, and the amplitude and phase of the phase voltage fluctuation. These parameters not only reflect the operating status of the high-voltage electrical system but are also closely related to the vehicle's power output, energy recovery, and electrical load distribution.

[0048] S13: Perform temperature field scanning processing on the thermally sensitive areas of the power battery pack, and capture the surface temperature gradient distribution of the motor controller and the change in heat flux density of the power module using an on-board infrared thermal imager to generate the original dataset of thermal load.

[0049] Specifically, infrared thermal imagers capture infrared radiation information from key heat source areas such as the surface of the motor controller and power modules, accurately reflecting the temperature distribution and changes in these areas. The system calibrates and verifies the onboard infrared thermal imager to ensure the accuracy and reliability of its measurement results. During vehicle operation, the infrared thermal imager continuously scans the heat-sensitive areas of the power battery pack at set time intervals, capturing temperature gradient distribution and heat flux density changes. The temperature gradient distribution reflects the direction and rate of heat transfer within the battery pack, while the heat flux density change reflects the amount of heat generated during charging and discharging and the efficiency of the cooling system. The system digitally processes the acquired infrared thermal image data, extracting key characteristic parameters of the temperature field, such as the highest temperature point, temperature change rate, and heat flux density distribution map, using image processing algorithms. These parameters not only intuitively reflect the thermal state of the battery pack but also provide important information for the optimized control of the thermal management system.

[0050] During processing, the system performs noise filtering and image enhancement on the infrared thermal image data to improve data quality and readability. Finally, the system integrates these processed thermal load data into a raw thermal load dataset. This dataset records in detail the thermal state changes of the power battery pack under different operating conditions, providing crucial thermal load information for subsequent thermal load feature extraction and the formulation of dynamic load allocation strategies.

[0051] S14: Based on the precise time protocol, the original datasets of mechanical load, electrical load and thermal load are spatiotemporally aligned. The time base of the mechanical load data, electrical load data and thermal load data is unified by the timestamp calibration algorithm to generate the initial load data.

[0052] Specifically, to achieve the fusion and analysis of mechanical load, electrical load, and thermal load data, the system can use a precise time protocol to perform spatiotemporal alignment processing on the raw datasets of various loads. The raw datasets of mechanical load, electrical load, and thermal load originate from different sensor networks and monitoring systems, and these data may have certain deviations in terms of acquisition time and spatial location.

[0053] Specifically, the system uses a precision time protocol and a high-precision clock to synchronize and calibrate the time of each data acquisition node. During time alignment, the system assigns a unique timestamp identifier to each data acquisition node, ensuring that each data point accurately corresponds to a specific time point. Simultaneously, the system employs a timestamp calibration algorithm to uniformly adjust the time base of different load data. This algorithm calculates calibration parameters by analyzing time deviations and frequency differences between different datasets and performs interpolation or sampling processing on the data, thereby unifying the time base of mechanical load data, electrical load data, and thermal load data to the same time scale. After spatiotemporal alignment processing, the system generates initial load data. This dataset not only retains the integrity and accuracy of the original data but also achieves temporal and spatial consistency for different load types. This spatiotemporal alignment processing lays a solid foundation for subsequent multi-source load data fusion and comprehensive analysis, enabling the system to more accurately grasp the overall load status of the vehicle under complex operating conditions and providing a more reliable basis for the formulation of dynamic load allocation strategies.

[0054] In one embodiment, step S2 of the dynamic load distribution method based on a vehicle body controller provided by the present invention specifically includes the following steps:

[0055] S21: Perform amplitude normalization processing on the mechanical load data of the initial load data, and scale the suspension stress data to the range of 0-1 according to the upper limit of the material yield strength to generate normalized mechanical load data.

[0056] Specifically, the system analyzes suspension stress data by comparing the stress data experienced by the suspension system with the upper limit of the material's yield strength. Following a specific mathematical ratio, these stress data are scaled to a range of 0-1. In this process, the system, based on the theoretical foundations of materials mechanics and the physical properties of the materials used in the vehicle's suspension system, precisely determines the parameters required for normalization. For example, for suspension components made of high-strength alloy steel, the upper limit of its yield strength can reach XXXMPa. The system divides the actual measured suspension stress value by this upper limit to obtain the normalized stress value. This processing method not only eliminates data inconsistencies caused by differences in size and material between different suspension components but also ensures that the normalized mechanical load data reflects the true load condition of the suspension system on a uniform scale.

[0057] S22: Perform dynamic range compression processing on the electrical load data of the initial load data, normalize the current data according to the maximum discharge current of the battery, scale the voltage data according to the rated voltage of the drive motor, and generate standard electrical load data.

[0058] Specifically, the system monitors the current ripple data provided by the battery management system in real time and compares this data with the battery's maximum discharge current value. The current value is then mapped to a standardized range through scaling. Simultaneously, the system performs similar processing on the phase voltage fluctuation data of the motor controller, normalizing the voltage data according to the rated voltage of the drive motor to ensure that the voltage data can also be analyzed on the same scale. This dynamic range compression processing not only effectively reduces the fluctuation amplitude of electrical load data but also retains key feature information such as the current ripple frequency and voltage fluctuation amplitude. During processing, the system verifies the normalized data in real time to ensure its accuracy and reliability. The processed standard electrical load data not only reflects the electrical load status of the battery and motor under different operating conditions but also provides standardized data support for subsequent electrical load feature extraction and the formulation of dynamic load allocation strategies.

[0059] S23: Perform temperature rise characteristic conversion processing on the thermal load data of the initial load data, calculate the relative heat flux density based on the highest withstand temperature of the power module, and generate relative heat flux density data.

[0060] Specifically, the system uses an onboard infrared thermal imager to capture temperature gradient distribution and heat flux density changes in key heat source areas such as the power battery pack and motor controller. This data includes the vehicle's thermal state information, such as the rate of temperature change and heat flux density distribution map. The system compares this thermal load data with the maximum tolerable temperature of the power module and calculates the relative heat flux density using a mathematical model. This process not only eliminates the difference in temperature dimensions between different heat source areas but also intuitively reflects the relative relationship between the current thermal load state and the ultimate performance limit of the power module.

[0061] During processing, the system performs real-time calibration and verification of the relative heat flux density data to ensure its accuracy and consistency. The generated relative heat flux density data not only reflects the thermal state of key components such as the power battery pack and motor controller, but also provides crucial information for the optimized control of the thermal management system. In this way, the system can effectively prevent battery overheating and thermal runaway risks, ensuring the safe and stable operation of the power battery system in new energy vehicles. Simultaneously, it provides critical thermal load information for subsequent thermal load feature extraction and the formulation of dynamic load allocation strategies.

[0062] S24: Perform matrix reconstruction processing on normalized mechanical load data, standard electrical load data and relative heat flux density data, integrate stress data, current and voltage data and heat flux data into a three-dimensional data cube according to time series, and generate a standard load dataset. The standard load dataset is used to indicate the real-time load status of various components of the vehicle body.

[0063] Specifically, the system creates a three-dimensional data cube structure. The first dimension is the time series, used to record how the data changes over time; the second dimension is different types of load data, used to distinguish the characteristics of various load types; and the third dimension stores the specific values ​​of each type of load data at each time point. Through matrix operations and data filling algorithms, the system sequentially fills the three-dimensional data cube with normalized mechanical load data, standard electrical load data, and relative heat flux density data, ensuring that each data point accurately reflects its actual value at a specific time and under a specific load type. The generated standard load dataset not only integrates the real-time load status of various vehicle components but also provides efficient data structure support for subsequent data analysis and mining through the three-dimensional data cube format.

[0064] In one embodiment, S3 of the dynamic load distribution method based on a vehicle body controller provided by the present invention specifically includes the following steps:

[0065] S31: Perform multimodal feature extraction processing on the standard load dataset, calculate the covariance matrix of mechanical vibration frequency domain energy distribution, electrical load time domain ripple coefficient and thermal gradient spatial change rate, and generate feature correlation matrix.

[0066] Specifically, the system uses the Fast Fourier Transform (FFT) algorithm to perform frequency domain transformation on the mechanical vibration signal and calculate the frequency domain energy distribution of the mechanical vibration. The system decomposes the vibration signal into multiple frequency components and calculates the energy proportion of each frequency component, thereby obtaining the energy distribution reflecting the energy transfer and dissipation characteristics of the mechanical system at different frequencies. For electrical load data, the system uses wavelet transform to decompose and reconstruct its time-domain signal and calculate the time-domain ripple coefficient of the electrical load. Through multi-scale analysis, wavelet transform can capture the small fluctuations and changes of the electrical load in the time domain, and the ripple coefficient quantifies the amplitude and frequency of these fluctuations, characterizing the stability and smoothness of the current and voltage in the electrical system. When processing thermal load data, the system calculates the covariance matrix of the spatial rate of change of the thermal gradient based on the spatial difference algorithm. In the vehicle thermal field model, the system performs spatial difference calculation on the temperature data of key heat source parts such as the power module and their surrounding areas to obtain the thermal gradient vector, and then solves the covariance matrix. This reveals the changing trend and correlation of the thermal load in the spatial dimension and reflects the propagation and distribution law of heat between different components.

[0067] S32: Perform principal component screening on the feature correlation matrix, retain the principal component directions with eigenvalues ​​greater than 1 based on the Kaiser criterion, calculate the feature subspace with a cumulative variance contribution rate greater than 95%, and generate a dimension-reduced feature projection matrix.

[0068] Specifically, based on the Kaiser criterion, the system automatically selects principal component directions with eigenvalues ​​greater than 1. These principal component directions represent the main information and trends contained in the data, effectively removing the influence of noise and redundant information. The system further calculates the cumulative variance contribution rate of the selected principal components, ensuring it is greater than 95%. This process gradually accumulates the variance contribution of the principal components until a preset 95% threshold is reached, thus determining the final feature subspace. Determining the feature subspace not only preserves the variation information of the original data to the greatest extent but also significantly reduces the data dimensionality, improving the efficiency and accuracy of subsequent data processing. The system generates a dimensionality-reduced feature projection matrix based on this feature subspace.

[0069] S33: Perform feature space mapping processing on the initial load data. Based on the dimension-reduced feature projection matrix, project the suspension vibration spectrum, battery current ripple and power module temperature gradient to a unified feature space, calculate the similarity between feature vectors, and generate a fused feature vector.

[0070] Specifically, the system takes heterogeneous data from multiple sources, such as suspension vibration spectrum, battery current ripple, and power module temperature gradient, as input. It then performs a linear transformation using a dimensionality-reduced feature projection matrix, projecting these original feature data into a unified feature space. In this process, data from different physical quantities and dimensions are transformed into feature vectors within the same feature space, enabling the originally dispersed and heterogeneous load data to be compared and analyzed within the same mathematical framework.

[0071] Furthermore, the system calculates the similarity between the projected feature vectors. This similarity calculation can employ cosine similarity, obtained by calculating the ratio of the dot product of two feature vectors to the product of their magnitudes. This step not only reveals the inherent correlation and similarity between different load characteristics but also provides a quantitative basis for subsequent feature fusion. Finally, based on the similarity calculation results, the system fuses similar features to generate a fused feature vector. This fused feature vector integrates key feature information from mechanical, electrical, and thermal loads, forming a comprehensive representation of the vehicle's overall load state.

[0072] S34: Retrieve the historical load feature set stored during the vehicle's historical operation from the on-board storage unit as a historical data sample. Based on the ISO 26262 functional safety standard, perform dynamic distribution analysis on the historical load feature set and the fused feature vector, update the feature distribution boundary of the fused feature vector, and determine the preset threshold range of the current feature vector with three times the standard deviation to generate the fused load feature set.

[0073] Specifically, the system performs rigorous dynamic distribution analysis on historical load feature sets and fused feature vectors based on the ISO 26262 functional safety standard. The ISO 26262 standard provides functional safety guidelines for automotive electrical / electronic systems, ensuring that the system fully considers safety and reliability requirements when processing data. During the dynamic distribution analysis, the system employs a Gaussian mixture model and kernel density estimation to model and fit the distribution characteristics of the historical load feature set and fused feature vectors. The Gaussian mixture model captures the multimodal distribution characteristics of the data, while kernel density estimation provides a non-parametric probability density estimate. The combination of these two methods allows the system to accurately depict the dynamic distribution of load features. Based on the analysis results, the system updates the feature distribution boundary of the fused feature vectors. This process uses statistical methods to redefine the reasonable fluctuation range of the features, eliminating the influence of outlier data points, making the feature distribution boundary more closely match actual operating conditions. Finally, the system determines the preset threshold range of the current feature vectors using three standard deviations. The three standard deviations principle is based on the normal distribution characteristics, ensuring that under normal operating conditions, the feature vector values ​​have a high probability of falling within this range, thus providing a scientific basis for anomaly detection.

[0074] In one embodiment, S4 of the dynamic load distribution method based on a vehicle body controller provided by the present invention specifically includes the following steps:

[0075] S41: Perform nonlinear modeling on the feature vectors and control parameters of the fused load feature set, construct a support vector regression architecture based on radial basis kernel function, set regularization parameters to constrain model complexity, and construct an initial correlation model between load features and control parameters.

[0076] Specifically, the system constructs a Support Vector Regression (SVR) architecture based on the radial basis function kernel to capture the nonlinear relationship between load features and control parameters. In the SVR model, the radial basis function (RBF) acts as the kernel function, mapping the original feature vectors to a high-dimensional space. This makes it easier to find a linear hyperplane to fit the data in the high-dimensional space, thus achieving effective modeling of complex nonlinear relationships. The system sets a regularization parameter C and a kernel parameter γ to constrain model complexity and prevent overfitting. The regularization parameter C controls the balance between the model's fit to the training data and its generalization ability. A larger C value makes the model focus more on fitting the training data, potentially leading to overfitting; a smaller C value makes the model more generalizable but may lead to underfitting. The kernel parameter γ determines the width of the RBF, affecting the model's sensitivity to local features of the data. By adjusting these two parameters, the system constructs an initial correlation model between load features and control parameters.

[0077] S42: Perform grid search optimization on the hyperparameters of the initial correlation model, evaluate the prediction accuracy of different (C,γ) combinations using cross-validation in the parameter space, select the parameter configuration with the minimum mean square error, and generate the optimized correlation model.

[0078] Specifically, the system divides the possible values ​​of C and γ into multiple discrete points, forming a grid-like parameter combination space. For example, C may take the value of... γ can take the value of For each (C,γ) combination, the system uses cross-validation to evaluate its prediction accuracy. Cross-validation divides the training dataset into k subsets, alternately using k-1 subsets as the training set and the remaining subset as the validation set, and performs model training and validation k times. Finally, the average of the k validation results is taken as the performance index of the parameter combination.

[0079] The system selects the parameter configuration with the minimum mean square error (MSE) as the optimal hyperparameter combination. In this way, the system ensures that the initial correlation model achieves the best predictive performance in the parameter space, generating an optimized correlation model. The optimized model can more accurately predict vehicle control parameters, such as the response time of the power steering system and the torque distribution ratio of the regenerative braking system.

[0080] S43: Perform incremental learning processing on the training sample set of the optimized correlation model, realize online iterative update of model parameters on the domain controller platform of the AUTOSAR architecture, use the exponential decay forgetting factor mechanism to eliminate outdated data, and generate the final load correlation model.

[0081] Specifically, after generating the optimized correlation model, the system air conditioning adopts an online iterative update method. When new fused load feature set sample data is obtained, it is included in the training sample set, triggering the model parameter update process. To prevent information loss caused by the complete replacement of old data by new data, the system introduces an exponential decay forgetting factor mechanism. The forgetting factor is a parameter between 0 and 1, usually set to a value close to 1, such as 0.9 or 0.99.

[0082] When updating model parameters, the weights of older data gradually decay exponentially with respect to the forgetting factor. For example, assuming the current iteration number is t and the forgetting factor is α, then in the t-th iteration, the weight of the i-th historical data sample is α^{ti}. In this way, the influence of outdated data gradually weakens over time but is not completely discarded. Through this method, the system achieves a smooth transition and dynamic update of model parameters, generating the final load-related model. This model can predict vehicle control parameters and adapt to real-time changes in vehicle dynamic operating conditions, ensuring that the system accurately assesses and effectively controls the vehicle's load state throughout its entire lifecycle, based on the latest data and model, thereby improving the vehicle's performance and safety at different operating stages.

[0083] S44: Perform control parameter mapping processing on the fused feature vector of the fused load feature set, predict the response time of the power steering system and the torque distribution ratio of the regenerative braking system through the load correlation model, and generate a combination of mapped control parameters.

[0084] Specifically, the system inputs the fused feature vector into an optimized and incrementally learned load correlation model. Through the model's nonlinear mapping capability, it predicts key control parameters such as the response time of the power steering system and the torque distribution ratio of the regenerative braking system. The system uses algorithms such as matrix multiplication and kernel function calculation to achieve model prediction. During the prediction process, the system dynamically adjusts the model's internal parameters based on the current fused feature vector to adapt to constantly changing load conditions.

[0085] For example, when a vehicle is cornering at high speed, the system can predict the appropriate steering assist response time based on the current mechanical and thermal load characteristics, ensuring vehicle stability and safety. Similarly, for regenerative braking systems, the system can predict the optimal torque distribution ratio based on the real-time state of electrical and thermal loads, improving braking energy recovery efficiency. These predictions generate a combination of mapped control parameters to guide the body controller in real-time adjustments to the powertrain and chassis systems.

[0086] In one embodiment, such as Figure 2 As shown, S5 of the dynamic load distribution method based on the vehicle body controller provided by the present invention specifically includes the following steps:

[0087] S51: Perform real-time out-of-bounds detection processing on the preset threshold range of the fused load feature set. When the suspension vibration energy feature value exceeds the upper limit of the preset threshold, recalculate the feature offset, call the sliding time window algorithm to confirm whether the duration of the abnormality exceeds the preset warning threshold, and generate an abnormality detection report containing abnormal trigger signals and feature offset data.

[0088] Specifically, the system implements a real-time out-of-bounds detection mechanism for the preset threshold range of the fused load feature set. Taking the suspension vibration energy feature value as an example, the system sets an upper limit for the threshold based on the ISO 26262 functional safety standard. When the suspension vibration energy feature value is detected to exceed the preset threshold, an abnormal response procedure is immediately initiated. The ISO 26262 standard specifies the functional safety requirements for automotive electrical / electronic systems, with ASIL-D being the highest level, requiring the system to possess extremely high safety and reliability. Preferably, the feature offset is recalculated using the following formula:

[0089]

[0090] in, Indicates the feature offset. This represents the current characteristic value of suspension vibration energy. The system sets a preset upper limit for the threshold. By calculating the offset, the system accurately quantifies the degree to which the feature value exceeds the limit.

[0091] Subsequently, the system invokes a sliding time window algorithm to evaluate the duration of the abnormal state. In the time series data, data is extracted using a window length L and a step size S. For each data point within the window... Calculate the duration T of the anomaly:

[0092]

[0093] The system will calculate the duration of the anomaly. T With preset warning threshold Compare them. If T > If the anomaly is detected, an anomaly detection report will be generated. The report includes anomaly trigger signals and characteristic offset data. The anomaly trigger signals are represented in binary form, with 1 indicating an anomaly has occurred and 0 indicating normal operation.

[0094] S52: Perform safety-critical task reordering on anomaly detection reports. Based on ISO 26262 ASIL-D level, place electronic stability program control tasks at the top of the real-time execution queue and electric power steering tasks in the second priority queue, generating a reordered task execution sequence.

[0095] Specifically, the system performs safety-critical task rearrangement on anomaly detection reports based on ISO 26262 ASIL-D level. In the real-time execution queue, the electronic stability program control task is placed first, with a priority weight set to [value missing]. The system then sets the priority weight of the electric power steering task to... The system reorders the task execution sequence according to priority weights, and calculates the execution order of tasks using a scheduling algorithm. Task scheduling follows a priority queue rule, with tasks having higher priority weights being executed first. Specifically, the system assigns a priority weight value to each task in the task queue; the higher the weight value, the higher the priority of the task in the queue. The high priority of electronic stability program control tasks ensures the dynamic stability of the vehicle under abnormal conditions, while the secondary priority of electric power steering tasks ensures timely response to steering operations.

[0096] S53: Based on the reordering of task execution sequences, the multi-core processor resources are dynamically partitioned to allocate computing resources of ARM Cortex-R52 real-time cores to the chassis control system and background resources of Cortex-A78 performance cores to the in-vehicle infotainment system, generating a core-task binding mapping table.

[0097] Specifically, the Cortex-R52 processor, with its superior real-time performance and functional safety, is suitable for chassis control systems requiring high-precision, low-latency control tasks, such as electronic stability programs and braking systems, ensuring these critical tasks can be executed within strict time constraints. Simultaneously, the system allocates background resources from the Cortex-A78 performance cores to the in-vehicle infotainment system. The Cortex-A78 processor excels at handling complex computational tasks and provides high performance, making it suitable for infotainment systems with relatively lower real-time requirements but higher processing power demands, such as multimedia playback and navigation functions. Specifically, the system calculates the computational resources required for each task using the following formula:

[0098]

[0099] in, This represents the amount of computing resources required for task i. This indicates the execution time of task i. This represents the deadline for task i. By calculating the resource requirements of each task, the system allocates computing resources of the ARM Cortex-R52 real-time core to the chassis control system, ensuring that it can handle critical control tasks in real time, such as electronic stability program control and electric power steering. Simultaneously, background resources of the Cortex-A78 performance core are allocated to the in-vehicle infotainment system to ensure its normal operation without affecting critical tasks.

[0100] For in-vehicle infotainment systems, based on their energy efficiency requirements, the system uses a dynamic partitioning algorithm to generate a core-task binding map according to task priority and resource needs. This map details the binding relationship between each task and the processor core, as well as the resource allocation of each core. For example, it records the binding relationship between the chassis control system task and the ARM Cortex-R52 real-time core, along with the processor clock cycles and memory resources allocated to that task. In this way, the system achieves fine-grained management of multi-core processor resources, improving resource utilization efficiency and ensuring the real-time execution of critical tasks and the smooth operation of non-critical tasks.

[0101] S54: Perform instruction encoding processing on the core-task binding mapping table, use the AUTOSAR COM module to encapsulate the task execution sequence, core binding relationship and resource quota parameters, and generate an optimized load distribution scheme including a time-triggered mechanism.

[0102] Specifically, the system generates an optimized load balancing scheme that includes a time-triggered mechanism by defining task priorities, allocating time slices, and setting resource limits. The specific algorithm formula is as follows:

[0103]

[0104] in, Indicates the total scheduling weight. Indicates the priority of task i. This represents the weight of task i. This formula is used to calculate the total weight of task scheduling, ensuring that high-priority tasks can be executed in a timely manner. The system dynamically adjusts the execution order and resource allocation of tasks based on their priority and weight. Simultaneously, the system employs a time-triggered mechanism, setting task execution cycles and deadlines to ensure tasks are completed within the specified time. Optimized load distribution schemes ensure stable vehicle operation under various working conditions by rationally allocating processor resources, improving the system's real-time performance and reliability.

[0105] Specifically, the system encapsulates the task execution sequence, core binding relationships, and resource quota parameters into instruction packets, which are then transmitted and executed through the AUTOSAR COM module. For example, for electronic stabilization program control tasks, the system sets their priority to the highest, allocates a larger time slice and resource quota, and ensures timely execution within each control cycle through a time-triggered mechanism.

[0106] In real-world vehicle operating environments, this computer-based dynamic load allocation method can significantly improve vehicle performance and safety. Taking new energy vehicles as an example, when a vehicle is cornering at high speed, the system can detect potential roll risks in a timely manner by monitoring the vibration energy characteristics of the suspension system in real time. Once an anomaly is detected, the system immediately prioritizes ESP and EPS tasks according to a preset safety strategy, allocating them with high-performance real-time processor core resources. This rapid system response effectively prevents vehicle skidding, improves driving stability, and ensures passenger safety. Simultaneously, by rationally allocating computing resources, the system ensures that non-critical tasks such as the in-vehicle infotainment system can also operate normally without affecting vehicle safety, enhancing the user's driving experience. This intelligent dynamic load allocation method not only optimizes the resource utilization efficiency of the vehicle's electronic systems but also enhances the vehicle's adaptability and reliability under complex operating conditions, providing crucial technical support for the development of intelligent vehicles.

[0107] In one embodiment, the dynamic load distribution method based on a vehicle body controller provided by the present invention further includes the following steps:

[0108] S61: Collects response delay and control accuracy data of the vehicle body actuators after implementing the optimized load distribution scheme through the vehicle bus, calculates the deviation between the actual value and the expected value of the system coordination efficiency index, and generates an execution status deviation report.

[0109] Specifically, the system utilizes in-vehicle communication networks such as CAN bus or FlexRay bus to acquire the response timestamps and control accuracy parameters of various actuators (such as the power steering motor, brake controller, and suspension adjuster) in real time. The response delay of the power steering motor is defined as the time interval from receiving the control command to actually starting to adjust the steering assist; control accuracy is measured by the difference between the actual steering angle and the target steering angle. For the suspension adjuster, the response delay is the time difference from receiving the command to starting to adjust the suspension height or damping; control accuracy is determined by the deviation between the actual suspension parameters and the target parameters.

[0110] The system calculates the deviation between the actual and expected values ​​of the system coordination efficiency index. The coordination efficiency index is a parameter that comprehensively evaluates the cooperation effect between various vehicle systems, and its calculation formula is as follows:

[0111]

[0112] in, As a collaborative efficiency indicator, and These are the actuator control accuracy weight and the response delay weight, respectively, both of which are preset according to the importance of the actuator in the vehicle's dynamic performance. To score the control accuracy, a score is given based on how close the actual control accuracy of the actuator is to the ideal accuracy. This is a latency penalty factor that increases with increasing response latency, negatively impacting collaborative efficiency. The system compares the calculated actual collaborative efficiency index with the expected value based on historical best performance or factory-set standards, generating an execution status deviation report. This report details the response latency, control accuracy data of each actuator, and the overall collaborative efficiency index deviation.

[0113] S62: Perform incremental learning processing on the execution status deviation report and the fused load feature set, use the stochastic gradient descent algorithm to calculate the parameter update amount of the load correlation model, and generate model parameter adjustment suggestions.

[0114] Specifically, the system performs incremental learning processing on the execution state deviation report and the fused load feature set, and uses the stochastic gradient descent algorithm to calculate the parameter update amount of the load correlation model, generating model parameter adjustment suggestions. In detail, the system uses the deviation amount from the execution state deviation report and the feature vectors from the fused load feature set as input to construct a loss function to measure the difference between the model's predicted values ​​and the actual values. The loss function is defined as follows:

[0115]

[0116] in, This is the actual value. Let be the predicted value, and n be the number of samples. The system minimizes this loss function using the stochastic gradient descent algorithm to calculate the update amount of the model parameters. The core idea of ​​the stochastic gradient descent algorithm is to randomly select a sample, calculate the gradient of that sample, and use this gradient to update the model parameters. The parameter update formula is:

[0117]

[0118] in, For model parameters, For learning rate, This represents the gradient of the loss function with respect to the model parameters. The system updates the model parameters iteratively until the loss function converges to a small value. Finally, the system generates model parameter tuning suggestions, which detail the amount and direction of the model parameter updates.

[0119] S63: The load association model and model parameter adjustment suggestions are updated online. The parameter update amount is applied in real time during vehicle operation to adjust the model weight parameters and generate an updated load association model.

[0120] Specifically, the system reads the parameter update amounts from the model parameter adjustment suggestions and adjusts the weight parameters of the load association model accordingly. The updated model better reflects the actual operating state of the vehicle, improving prediction accuracy. The system applies parameter updates in real time during vehicle operation, ensuring the model remains in optimal condition through an online update mechanism. The updated load association model not only improves the system's collaborative efficiency but also enhances its adaptability to dynamic changes, providing more precise support for intelligent vehicle control.

[0121] After each update, the system performs real-time model validation to ensure that the updated model can accurately predict the actuator's response latency and control precision. The validation process can be performed using a pre-reserved validation dataset, calculating the prediction error of the updated model on the validation set. If the prediction error is within an acceptable range, the updated model is accepted; otherwise, the system will revert to the previous model parameters and recalculate the update parameters.

[0122] Through this online update mechanism, the system ensures that the load association model is always in optimal condition, accurately reflecting the actual operating conditions of the vehicle, thereby improving the accuracy and efficiency of dynamic load allocation. In the actual operating environment of new energy vehicles, this real-time update mechanism can significantly improve vehicle performance and safety. For example, under complex road conditions, the system can quickly adjust the load allocation strategy based on the real-time updated model, optimizing vehicle handling performance and energy efficiency. In long-term use, the system's adaptive capabilities ensure that the vehicle maintains optimal operating condition, reducing maintenance costs and enhancing the user experience.

[0123] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0124] Based on the same inventive concept, this application also provides a vehicle body controller-based dynamic load distribution device for implementing the above-described vehicle body controller-based dynamic load distribution method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more vehicle body controller-based dynamic load distribution device embodiments provided below can be found in the limitations of the vehicle body controller-based dynamic load distribution method described above, and will not be repeated here.

[0125] Preferably, such as Figure 4 As shown, the present invention provides a dynamic load distribution device 700 based on a vehicle body controller, which is configured with the following modules:

[0126] The data acquisition module 710 is used to acquire data from various components of the vehicle body based on a distributed sensor network, including current and voltage data of electrical loads, force and acceleration data of mechanical loads, and temperature and heat flow data of thermal loads, and to generate initial load data.

[0127] The data preprocessing module 720 is used to preprocess the initial load data, eliminate the dimensional differences between mechanical load, electrical load and thermal load, and generate a standard load dataset.

[0128] The fusion load feature generation module 730 is used to perform unified feature space mapping on the standard load dataset, extract comprehensive characterization features of mechanical load, electrical load and thermal load, and generate feature threshold range based on historical data distribution, thereby generating a fusion load feature set containing fusion feature vector and preset threshold range.

[0129] The load association model mapping module 740 is used to construct a load association model of load features and control parameters based on the fused load feature set, and input the fused feature vector into the load association model for parameter mapping processing to generate a combination of mapped control parameters.

[0130] The dynamic load optimization and allocation module 750 is used to dynamically adjust the load of the mapping control parameter combination based on the preset threshold range of the fused load feature set, adjust the task execution order and resource allocation ratio based on the preset load priority rules, and generate an optimized load allocation scheme. The optimized load allocation scheme is used to instruct the CPU core allocation strategy and control the body controller to adjust the parameters of the power system and chassis system in real time.

[0131] Preferably, the data acquisition module 710 provided in this application is configured with the following units:

[0132] The mechanical load data acquisition unit is used to monitor and process the stress distribution of the vehicle body structural components. It collects vibration spectrum data of key suspension nodes through strain sensors on the longitudinal beams of the frame and generates a raw mechanical load dataset.

[0133] The electrical load data acquisition unit is used to acquire and process parameters of the vehicle's high-voltage electrical system in real time. It obtains data on the discharge current ripple of the battery management system and the phase voltage fluctuation of the motor controller through the CAN bus, and generates the original electrical load dataset.

[0134] The thermal load data acquisition unit is used to scan the temperature field of the thermally sensitive area of ​​the power battery pack. It captures the surface temperature gradient distribution of the motor controller and the change in heat flux density of the power module through the vehicle-mounted infrared thermal imager, and generates the original thermal load dataset.

[0135] The multi-load data spatiotemporal alignment unit is used to perform spatiotemporal alignment of the original datasets of mechanical load, electrical load, and thermal load based on a precise time protocol. It unifies the time base of mechanical load data, electrical load data, and thermal load data through a timestamp calibration algorithm to generate initial load data.

[0136] Preferably, the data preprocessing module 720 provided in this application is configured with the following units:

[0137] The mechanical load amplitude normalization unit is used to normalize the mechanical load data of the initial load data, and scale the suspension stress data to the range of 0-1 according to the upper limit of the material yield strength to generate normalized mechanical load data.

[0138] The electrical load dynamic compression unit is used to perform dynamic range compression processing on the electrical load data of the initial load data. Based on the normalized current data of the maximum discharge current of the battery, the voltage data is scaled according to the rated voltage of the drive motor to generate standard electrical load data.

[0139] The heat load temperature rise conversion unit is used to process the initial load data for temperature rise characteristic conversion, calculate the relative heat flux density based on the highest withstand temperature of the power module, and generate relative heat flux density data.

[0140] The multi-load data matrix reconstruction unit is used to perform matrix reconstruction processing on normalized mechanical load data, standard electrical load data and relative heat flux density data. It integrates stress data, current and voltage data and heat flux data into a three-dimensional data cube according to the time series to generate a standard load dataset. The standard load dataset is used to indicate the real-time load status of various components of the vehicle body.

[0141] Preferably, the fusion load feature generation module 730 provided in this application is configured with the following units:

[0142] Multimodal feature correlation analysis unit: used to perform multimodal feature extraction processing on standard load datasets, calculate the covariance matrix of mechanical vibration frequency domain energy distribution, electrical load time domain ripple coefficient and thermal gradient spatial change rate, and generate feature correlation matrix;

[0143] Principal component dimensionality reduction projection unit: used to perform principal component screening on the feature correlation matrix, retain the principal component directions with eigenvalues ​​greater than 1 based on the Kaiser criterion, calculate the feature subspace with a cumulative variance contribution rate greater than 95%, and generate the dimensionality reduction feature projection matrix.

[0144] Unified Feature Space Fusion Unit: Used to perform feature space mapping processing on the initial load data. Based on the dimension-reduced feature projection matrix, the suspension vibration spectrum, battery current ripple and power module temperature gradient are projected to the unified feature space, the similarity between feature vectors is calculated, and a fused feature vector is generated.

[0145] Historical data threshold modeling unit: used to retrieve the historical load feature set stored during the vehicle's historical operation from the on-board storage unit as historical data samples, perform dynamic distribution analysis on the historical load feature set and fused feature vector based on the ISO 26262 functional safety standard, update the feature distribution boundary of the fused feature vector, and determine the preset threshold range of the current feature vector with three times the standard deviation to generate the fused load feature set.

[0146] Preferably, the load association model mapping module 740 provided in this application is configured with the following units:

[0147] The nonlinear modeling building unit is used to perform nonlinear modeling processing on the feature vectors and control parameters of the fused load feature set, construct a support vector regression architecture based on the radial basis kernel function, set regularization parameters to constrain model complexity, and construct an initial correlation model between load features and control parameters.

[0148] The hyperparameter grid optimization unit is used to perform grid search optimization on the model hyperparameters of the initial correlation model. In the parameter space, the cross-validation method is used to evaluate the prediction accuracy of different (C,γ) combinations, select the parameter configuration with the minimum mean square error, and generate the optimized correlation model.

[0149] The incremental learning update unit is used to perform incremental learning processing on the training sample set of the optimized correlation model. It realizes online iterative update of model parameters on the domain controller platform of the AUTOSAR architecture, and uses an exponential decay forgetting factor mechanism to eliminate outdated data and generate the final load correlation model.

[0150] The control parameter mapping unit is used to perform control parameter mapping processing on the fused feature vector of the fused load feature set, predict the response time of the power steering system and the torque distribution ratio of the regenerative braking system through the load correlation model, and generate a combination of mapped control parameters.

[0151] Preferably, the dynamic load optimization and allocation module 750 provided in this application is configured with the following units:

[0152] The load anomaly detection unit is used to perform real-time out-of-bounds detection processing on the preset threshold range of the fused load feature set. When the suspension vibration energy feature value exceeds the upper limit of the preset threshold, the feature offset is recalculated, and the sliding time window algorithm is called to confirm whether the duration of the anomaly exceeds the preset warning threshold. An anomaly detection report containing the anomaly trigger signal and feature offset data is generated.

[0153] The safety task reordering unit processes the anomaly detection report and reorders the safety-critical tasks. Based on the ISO 26262ASIL-D level, it places the electronic stability program control task at the top of the real-time execution queue and the electric power steering task in the second priority queue, generating a reordered task execution sequence.

[0154] The processor resource partitioning unit is used to dynamically partition multi-core processor resources based on the reordered task execution sequence, allocate computing resources of ARM Cortex-R52 real-time cores to the chassis control system, allocate background resources of Cortex-A78 performance cores to the in-vehicle infotainment system, and generate a core-task binding mapping table.

[0155] The load distribution encoding unit is used to encode instructions into the core-task binding mapping table. It uses an AUTOSAR COM module to encapsulate the task execution sequence, core binding relationship, and resource quota parameters to generate an optimized load distribution scheme that includes a time-triggered mechanism.

[0156] In one embodiment, the dynamic load distribution device 700 based on the vehicle body controller provided in this application is further configured with the following units:

[0157] The execution status monitoring unit is used to collect response delay and control accuracy data of the body actuators after the optimized load distribution scheme is executed via the vehicle bus, calculate the deviation between the actual value and the expected value of the system coordination efficiency index, and generate an execution status deviation report.

[0158] The incremental learning optimization unit is used to perform incremental learning processing on the execution status deviation report and the fused load feature set. It uses the stochastic gradient descent algorithm to calculate the parameter update amount of the load correlation model and generate model parameter adjustment suggestions.

[0159] The online model update unit is used to update the load association model and model parameter adjustment suggestions online. During vehicle operation, the parameter update amount is applied in real time to adjust the model weight parameters and generate an updated load association model.

[0160] In one embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described dynamic load distribution method based on the vehicle body controller.

[0161] In one embodiment, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described dynamic load distribution method based on a vehicle body controller.

[0162] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0163] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0164] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dynamic load distribution method based on a vehicle body controller, characterized in that, Includes the following steps: S1: Data is collected from various components of the vehicle body based on a distributed sensor network, including current and voltage data of electrical loads, force and acceleration data of mechanical loads, and temperature and heat flow data of thermal loads, to generate initial load data; S2: Preprocess the initial load data to eliminate the dimensional differences between mechanical load, electrical load and thermal load, and generate a standard load dataset; S3: Perform unified feature space mapping on the standard load dataset, extract comprehensive characterization features of mechanical load, electrical load and thermal load, and generate feature threshold range based on historical data distribution to generate a fused load feature set containing fused feature vector and preset threshold range; S4: Construct a load association model of load features and control parameters based on the fused load feature set, and input the fused feature vector into the load association model for parameter mapping processing to generate a combination of mapped control parameters; S5: Based on the preset threshold range of the fused load feature set, dynamically adjust the load of the mapping control parameter combination, adjust the task execution order and resource allocation ratio based on the preset load priority rules, and generate an optimized load allocation scheme. The optimized load allocation scheme is used to instruct the CPU core allocation strategy and control the body controller to adjust the parameters of the power system and chassis system in real time. S61: Collect response delay and control accuracy data of the body actuators after executing the optimized load distribution scheme through the vehicle bus, calculate the deviation between the actual value and the expected value of the system coordination efficiency index, and generate an execution status deviation report. S62: Perform incremental learning processing on the execution status deviation report and the fused load feature set, use the stochastic gradient descent algorithm to calculate the parameter update amount of the load correlation model, and generate model parameter adjustment suggestions; S63: The load association model and the model parameter adjustment suggestions are updated online. During vehicle operation, the parameter update amount is applied in real time to adjust the model weight parameters and generate an updated load association model. The step of constructing a load correlation model based on the fused load feature set and control parameters, and inputting the fused feature vector into the load correlation model for parameter mapping processing to generate a mapping control parameter combination includes: S41: Perform nonlinear modeling on the feature vectors and control parameters of the fused load feature set, construct a support vector regression architecture based on radial basis kernel function, set regularization parameters to constrain model complexity, and construct an initial correlation model between load features and control parameters; S42: Perform grid search optimization on the model hyperparameters of the initial association model, evaluate the prediction accuracy of different (C,γ) combinations in the parameter space using cross-validation, select the parameter configuration with the minimum mean square error, and generate an optimized association model. S43: Perform incremental learning processing on the training sample set of the optimized correlation model, realize online iterative update of model parameters on the domain controller platform of AUTOSAR architecture, use the exponential decay forgetting factor mechanism to eliminate old data, and generate the final load correlation model. S44: Perform control parameter mapping processing on the fused feature vector of the fused load feature set, predict the response time of the steering assist system and the torque distribution ratio of the regenerative braking system through the load association model, and generate a combination of mapping control parameters; The process of dynamically adjusting the mapping control parameter combination based on a preset threshold range of the fused load feature set, adjusting the task execution order and resource allocation ratio based on preset load priority rules, and generating an optimized load allocation scheme includes: S51: Perform real-time out-of-bounds detection processing on the preset threshold range of the fused load feature set. When the suspension vibration energy feature value exceeds the upper limit of the preset threshold, recalculate the feature offset, call the sliding time window algorithm to confirm whether the duration of the anomaly exceeds the preset warning threshold, and generate an anomaly detection report containing the anomaly trigger signal and feature offset data; the calculation formula for the feature offset is: ； in, Indicates the feature offset. This represents the current characteristic value of suspension vibration energy. The upper limit of the preset threshold; S52: Perform safety-critical task reordering on the anomaly detection report, and based on the ISO 26262 ASIL-D level, place the electronic stability program control task at the top of the real-time execution queue and the electric power steering task in the second priority queue, generating a reordered task execution sequence. S53: Based on the reordered task execution sequence, the multi-core processor resources are dynamically partitioned to allocate computing resources of ARM Cortex-R52 real-time cores to the chassis control system and background resources of Cortex-A78 performance cores to the in-vehicle infotainment system, and a core-task binding mapping table is generated. S54: Perform instruction encoding processing on the core-task binding mapping table, and use the AUTOSAR COM module to encapsulate the task execution sequence, core binding relationship and resource quota parameters to generate an optimized load allocation scheme that includes a time triggering mechanism.

2. The method according to claim 1, characterized in that, S1 includes: S11: Perform stress distribution monitoring on the vehicle body structural components, collect vibration spectrum data of key suspension nodes through strain sensors on the frame longitudinal beams, and generate raw dataset of mechanical load. S12: Real-time parameter acquisition and processing of the vehicle's high-voltage electrical system; obtaining battery management system discharge current ripple and motor controller phase voltage fluctuation data via CAN bus to generate raw electrical load dataset. S13: Perform temperature field scanning on the thermally sensitive areas of the power battery pack, capture the surface temperature gradient distribution of the motor controller and the change in heat flux density of the power module using an on-board infrared thermal imager, and generate the original dataset of thermal load. S14: Based on the precision time protocol, the original datasets of the mechanical load, electrical load, and thermal load are spatiotemporally aligned. The time base of the mechanical load data, electrical load data, and thermal load data is unified by the timestamp calibration algorithm to generate initial load data.

3. The method according to claim 1, characterized in that, S2 includes: S21: Perform amplitude normalization processing on the mechanical load data of the initial load data, and scale the suspension stress data to the range of 0-1 according to the upper limit of the material yield strength to generate normalized mechanical load data. S22: Perform dynamic range compression processing on the electrical load data of the initial load data, normalize the current data according to the maximum discharge current of the battery, and scale the voltage data according to the rated voltage of the drive motor to generate standard electrical load data. S23: Perform temperature rise characteristic conversion processing on the thermal load data of the initial load data, calculate the relative heat flux density based on the highest withstand temperature of the power module, and generate relative heat flux density data. S24: Perform matrix reconstruction processing on the normalized mechanical load data, the standard electrical load data, and the relative heat flux density data, and integrate the stress data, current and voltage data, and heat flux data into a three-dimensional data cube according to the time series to generate a standard load dataset. The standard load dataset is used to indicate the real-time load status of each component of the vehicle body.

4. The method according to claim 1, characterized in that, S3 includes: S31: Perform multimodal feature extraction processing on the standard load dataset, calculate the covariance matrix of mechanical vibration frequency domain energy distribution, electrical load time domain ripple coefficient and thermal gradient spatial change rate, and generate feature correlation matrix; S32: Perform principal component screening on the feature correlation matrix, retain the principal component directions with eigenvalues ​​greater than 1 based on the Kaiser criterion, calculate the feature subspace with a cumulative variance contribution rate greater than 95%, and generate a dimension-reduced feature projection matrix. S33: Perform feature space mapping processing on the initial load data, project the suspension vibration spectrum, battery current ripple and power module temperature gradient to a unified feature space based on the dimensionality reduction feature projection matrix, calculate the similarity between feature vectors, and generate a fused feature vector. S34: Retrieve the historical load feature set stored during the vehicle's historical operation from the on-board storage unit as a historical data sample. Perform dynamic distribution analysis on the historical load feature set and the fused feature vector based on the ISO 26262 functional safety standard, update the feature distribution boundary of the fused feature vector, and determine the preset threshold range of the current feature vector with three times the standard deviation to generate the fused load feature set.

5. A dynamic load distribution device based on a vehicle body controller, characterized in that, The device includes: The data acquisition module is used to collect data from various components of the vehicle body based on a distributed sensor network. It collects current and voltage data of electrical loads, force and acceleration data of mechanical loads, and temperature and heat flow data of thermal loads to generate initial load data. The data preprocessing module is used to preprocess the initial load data, eliminate the dimensional differences between mechanical load, electrical load and thermal load, and generate a standard load dataset. The fusion load feature generation module is used to perform unified feature space mapping on the standard load dataset, extract comprehensive characterization features of mechanical load, electrical load and thermal load, and generate feature threshold range based on historical data distribution, thereby generating a fusion load feature set containing fusion feature vector and preset threshold range. The load association model mapping module is used to construct a load association model of load features and control parameters based on the fused load feature set, and input the fused feature vector into the load association model for parameter mapping processing to generate a combination of mapped control parameters; The dynamic load optimization and allocation module is used to dynamically adjust the load of the mapping control parameter combination based on the preset threshold range of the fused load feature set, adjust the task execution order and resource allocation ratio based on the preset load priority rules, and generate an optimized load allocation scheme. The optimized load allocation scheme is used to instruct the CPU core allocation strategy and control the body controller to adjust the parameters of the power system and chassis system in real time. The execution status monitoring unit is used to collect response delay and control accuracy data of the body actuators after the optimized load distribution scheme is executed via the vehicle bus, calculate the deviation between the actual value and the expected value of the system coordination efficiency index, and generate an execution status deviation report. The incremental learning optimization unit is used to perform incremental learning processing on the execution status deviation report and the fused load feature set, and to calculate the parameter update amount of the load association model using the stochastic gradient descent algorithm to generate model parameter adjustment suggestions. The online model update unit is used to update the load association model and the model parameter adjustment suggestions online. During vehicle operation, the updated parameters are applied in real time to adjust the model weight parameters, generating an updated load association model. The load association model mapping module includes: The nonlinear modeling construction unit is used to perform nonlinear modeling processing on the feature vectors and control parameters of the fused load feature set, construct a support vector regression architecture based on the radial basis kernel function, set regularization parameters to constrain model complexity, and construct an initial correlation model between load features and control parameters. The hyperparameter grid optimization unit is used to perform grid search optimization on the model hyperparameters of the initial association model. In the parameter space, the cross-validation method is used to evaluate the prediction accuracy of different (C,γ) combinations, select the parameter configuration with the minimum mean square error, and generate an optimized association model. The incremental learning update unit is used to perform incremental learning processing on the training sample set of the optimized correlation model, realize online iterative update of model parameters on the domain controller platform of the AUTOSAR architecture, and use the exponential decay forgetting factor mechanism to eliminate outdated data and generate the final load correlation model. The control parameter mapping unit is used to perform control parameter mapping processing on the fused feature vector of the fused load feature set, predict the response time of the power steering system and the torque distribution ratio of the regenerative braking system through the load association model, and generate a combination of mapped control parameters. The dynamic load optimization and allocation module includes: The load anomaly detection unit is used to perform real-time out-of-bounds detection processing on the preset threshold range of the fused load feature set. When the suspension vibration energy feature value exceeds the upper limit of the preset threshold, the feature offset is recalculated, and a sliding time window algorithm is called to confirm whether the duration of the anomaly exceeds the preset warning threshold. An anomaly detection report containing the anomaly trigger signal and feature offset data is generated. The formula for calculating the feature offset is: ； in, Indicates the feature offset. This represents the current characteristic value of suspension vibration energy. The upper limit of the preset threshold; The safety task reordering unit is used to reorder the safety-critical tasks of the anomaly detection report. Based on the ISO26262 ASIL-D level, the electronic stability program control task is placed at the top of the real-time execution queue, and the electric power steering task is placed in the second priority queue, generating a reordered task execution sequence. The processor resource partitioning unit is used to dynamically partition multi-core processor resources based on the reordered task execution sequence, allocate computing resources of ARM Cortex-R52 real-time cores to the chassis control system, allocate background resources of Cortex-A78 performance cores to the in-vehicle infotainment system, and generate a core-task binding mapping table. The load allocation encoding unit is used to perform instruction encoding processing on the core-task binding mapping table. It uses an AUTOSAR COM module to encapsulate the task execution sequence, core binding relationship and resource quota parameters to generate an optimized load allocation scheme that includes a time-triggered mechanism.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 4.

Images