A vehicle control compensation method and device based on cross-system coupling relationship

Abstract

The application discloses a vehicle control compensation method and device based on cross-system coupling relationship, and relates to the technical field of vehicles.The method comprises the following steps: collecting multi-source operation data related to a steering system, a driving system, a braking system and a vehicle body posture in real time during vehicle operation; inputting the multi-source operation data into a benchmark health model to obtain a predicted behavior signal of the vehicle in a healthy state; comparing the predicted behavior signal with a corresponding actually collected behavior signal, calculating a coupling deviation of the predicted behavior signal and the behavior signal, and extracting a space-time feature of the coupling deviation; performing trend analysis on the coupling deviation based on the space-time feature of the coupling deviation, combining a preset threshold and a pattern matching algorithm, and identifying an implicit and gradual degradation mode of the vehicle; and when the degradation trend is identified, generating a control instruction to control and compensate the vehicle.The application solves the problems that the timeliness and accuracy of the vehicle control compensation mode in the prior art are low.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and in particular to a vehicle control compensation method and apparatus based on cross-system coupling relationships. Background Technology

[0002] As the automotive industry undergoes a profound transformation towards intelligent and electric vehicles, vehicle control precision, driving safety, and reliability have become core R&D objectives. A modern vehicle is a complex coupled system comprised of multiple subsystems, including steering, drive, and braking systems. The dynamic interactions between these subsystems directly impact the vehicle's overall control performance and driving stability. For example, during steering, changes in drive torque output affect steering response speed, while vehicle body deviations in turn influence braking force distribution. This cross-system coupling effect persists throughout the vehicle's entire operating condition.

[0003] Existing vehicle control technologies mostly adopt the approach of "independent control of a single subsystem" or "cooperative control based on preset rules," which has significant limitations. On the one hand, they ignore the time-varying and complex dynamic coupling relationships between subsystems and only perform closed-loop control on the operating parameters of a single system. This makes it difficult to cope with dynamic changes in operating conditions under the coupling of multiple systems, and it is easy for control deviations to accumulate. On the other hand, existing fault diagnosis and compensation strategies mostly focus on "explicit faults" (such as sudden faults such as sensor failure and actuator jamming), and lack effective means to identify "implicit progressive degradation" caused by factors such as component wear, aging, and environmental corrosion during vehicle operation (such as slow increase in steering clearance, gradual decrease in drive motor efficiency, and gradual change in braking friction coefficient).

[0004] Latent, gradual degradation is characterized by slow changes and insidious effects. Initially, it may not cause obvious fault symptoms, but as degradation accumulates, it gradually disrupts the normal coupling between systems, leading to decreased vehicle control precision, increased response delays, and even cascading failures, seriously threatening driving safety. Current technologies lack in-depth fusion analysis of multi-source operational data, making it impossible to accurately capture subtle changes in coupling relationships and identify degradation trends in advance. Furthermore, compensation strategies are mostly "reactive compensation," meaning remedial measures are only initiated after a fault occurs, failing to achieve "predictive compensation" based on degradation trends and making it difficult to fundamentally avoid the risk of control performance deterioration.

[0005] Therefore, current cross-system vehicle control compensation methods suffer from low timeliness and accuracy. Summary of the Invention

[0006] In view of this, the purpose of the present invention is to provide a vehicle control compensation method and apparatus based on cross-system coupling relationship, which aims to solve the problems of low timeliness and accuracy of existing vehicle control compensation methods.

[0007] This invention proposes a vehicle control compensation method based on cross-system coupling relationships, the method comprising:

[0008] During vehicle operation, multi-source operational data related to the steering system, drive system, braking system, and vehicle posture are collected in real time.

[0009] Multi-source operational data is input into a pre-trained baseline health model that represents dynamic coupling relationships to obtain predicted behavior signals of the vehicle in a healthy state.

[0010] By comparing the predicted behavior signal with the corresponding actual collected behavior signal, the coupling deviation between the predicted behavior signal and the behavior signal is calculated and the spatiotemporal features of the coupling deviation are extracted.

[0011] Based on the spatiotemporal characteristics of coupling bias, trend analysis of coupling bias is performed, and combined with preset thresholds and pattern matching algorithms, the implicit progressive degradation mode of vehicles is identified.

[0012] When a degradation trend is detected, a control command is generated, instructing the drive or braking system to apply predictive compensation torque to control and compensate the vehicle.

[0013] Furthermore, in the above-mentioned vehicle control compensation method based on cross-system coupling, the training process of the baseline health model includes:

[0014] Collect historical operating data of the vehicle under different operating conditions. The historical operating data includes multi-source data related to the steering system, drive system, braking system and vehicle posture.

[0015] The collected historical operational data is preprocessed, including data cleaning to remove noise and outliers, and data normalization to unify data units, to ensure the consistency and usability of data quality.

[0016] Key features are extracted from the preprocessed historical operating data, including steering torque characteristics of the steering system, driving torque characteristics of the drive system, braking torque characteristics of the braking system, and pitch and roll angles of the vehicle body.

[0017] Based on the extracted key features, an initial benchmark health model representing the dynamic coupling relationship is constructed. This model adopts a deep learning neural network architecture, takes multi-source key features as input, and takes the predicted behavior signal of the vehicle in a healthy state as output.

[0018] The initial baseline health model is trained using historical operational data under labeled health conditions. During the training process, the backpropagation algorithm is used to continuously adjust the model parameters, so that the error between the predicted behavior signal output by the model and the behavior signal under the actual labeled health conditions gradually decreases.

[0019] During the training process, cross-validation is used to divide the historical running data into training set, validation set and test set. The model is trained using the training set, the performance of the model is evaluated and tuned using the validation set, and finally the performance of the trained model is evaluated using the test set.

[0020] When the model's performance metrics on the test set meet the preset requirements, the model is determined to be a pre-trained benchmark healthy model representing dynamic coupling relationships.

[0021] Furthermore, in the above-mentioned vehicle control compensation method based on cross-system coupling relationship, the step of comparing the predicted behavior signal with the corresponding actually acquired behavior signal, calculating the coupling deviation between the predicted behavior signal and the behavior signal, and extracting the spatiotemporal features of the coupling deviation includes:

[0022] Simultaneous preprocessing is performed on the predicted behavior signal and the actual behavior signal. The preprocessing includes time dimension alignment and normalization.

[0023] Based on the aligned predicted behavior signal and the actual behavior signal, the single-system deviation values ​​of the predicted behavior signal and the actual behavior signal are calculated respectively. Then, through absolute value transformation and normalization, the standardized single-system deviation is obtained.

[0024] Based on the cross-system coupling characteristics of vehicles, a coupling weight matrix is ​​constructed. The element values ​​of this matrix represent the coupling correlation strength between different systems. The deviation of each single system is used as input, and weighted fusion calculation is performed in combination with the coupling weight matrix to obtain the coupling deviation of each cross-system combination.

[0025] The coupling deviations of all cross-system combinations are fused by weighted summation to obtain a global coupling deviation that can characterize the cross-system coupling state of the whole vehicle. The global coupling deviation is then preprocessed to obtain a coupling deviation sequence.

[0026] Based on the coupling bias sequence, time domain and spatial domain features are extracted respectively. The time domain features and spatial domain features are normalized and fused to form a complete coupling bias spatiotemporal feature set.

[0027] Furthermore, in the above-mentioned vehicle control compensation method based on cross-system coupling relationships, the step of constructing a coupling weight matrix based on the cross-system coupling characteristics of the vehicle, wherein the element values ​​of the matrix characterize the coupling strength between different systems, includes:

[0028] Based on vehicle dynamics principles and cross-system interaction mechanisms, a set of core influencing factors is identified that affect the coupling strength between the steering system, drive system, braking system, and vehicle posture system. The set of core influencing factors includes dynamic operating condition parameters, system operating state parameters, and environmental disturbance parameters.

[0029] Based on historical operational data, parameter sequences corresponding to each core influencing factor are extracted. A dual-quantitative analysis method coupling mutual information entropy and grey relational degree is adopted to calculate the initial coupling correlation degree of any two systems under different combinations of influencing factors.

[0030] Construct an initial coupling weight matrix framework with a preset dimension. Set the diagonal elements of the matrix to zero and set the initial values ​​of the off-diagonal elements to the initial coupling correlation degree of the corresponding system combination.

[0031] Furthermore, the above-mentioned vehicle control compensation method based on cross-system coupling relationships, after the step of constructing an initial coupling weight matrix framework of a preset dimension, with the diagonal elements of the matrix set to zero and the initial values ​​of the off-diagonal elements set to the initial coupling correlation degree of the corresponding system combination, further includes:

[0032] Based on the dynamic operating condition parameters, system operating status parameters, and environmental interference parameters in the real-time collected operating data, an operating condition adaptation factor calculation model is constructed. The model takes the deviation values ​​between the real-time parameters of each core influencing factor and the standard operating condition parameters as input, and outputs the operating condition adaptation factor corresponding to each system combination through a radial basis neural network. This factor represents the degree of gain or attenuation of the coupling correlation strength between the current operating condition and the standard operating condition.

[0033] The off-diagonal elements in the initial coupling weight matrix are multiplied with the corresponding system combination's operating condition adaptation factor to obtain the dynamically adjusted coupling correlation strength value. The initial coupling weight matrix is ​​then updated to obtain the final vehicle cross-system coupling weight matrix.

[0034] Furthermore, in the aforementioned vehicle control compensation method based on cross-system coupling relationships, the step of calculating the initial coupling correlation degree of any two systems under different combinations of influencing factors using a dual-quantitative analysis method that couples mutual information entropy and grey relational degree includes:

[0035] First, the degree of nonlinear correlation between the operating parameters of the two systems is calculated by mutual information entropy to obtain the information entropy correlation value. Then, the degree of trend fit between the parameter sequences of the two systems is calculated by grey relational degree to obtain the grey relational value.

[0036] Based on the entropy weight method, the weight ratios of information entropy correlation value and gray correlation value are allocated, and the initial coupling correlation degree under the combination of various influencing factors is obtained by weighted fusion.

[0037] Furthermore, in the aforementioned vehicle control compensation method based on cross-system coupling relationships, the step of analyzing the trend of coupling deviation based on the spatiotemporal characteristics of coupling deviation, and identifying the vehicle's implicit progressive degradation mode by combining a preset threshold with a pattern matching algorithm, includes:

[0038] The extracted spatiotemporal feature set of coupling deviation is standardized and dimensionality reduced. Principal component analysis algorithm is used to remove redundant features and retain key spatiotemporal features that can characterize the core change law of coupling deviation, resulting in a simplified core feature sequence.

[0039] Based on the simplified core feature sequence, a time series trend analysis model is constructed. A combination of sliding window weighted fitting and exponential smoothing prediction is used to track the changing trend of coupling bias in real time and make short-term predictions.

[0040] Based on the health operation data of vehicles under different operating conditions, a multi-level threshold system for operating condition adaptation is constructed. The multi-level thresholds include warning thresholds, degradation confirmation thresholds, and emergency intervention thresholds. The actual rate of change and predicted change of the coupling deviation obtained from trend analysis are compared with the multi-level thresholds under the corresponding operating conditions to preliminarily determine whether there is a potential degradation trend.

[0041] Collect historical data on different types of implicit progressive degradation throughout the entire life cycle of a vehicle, extract spatiotemporal feature templates of coupling deviations corresponding to each degradation type, label the degradation level and evolution cycle, and construct a standardized implicit degradation pattern library.

[0042] The dynamic time warping algorithm is used as the core pattern matching algorithm to calculate the similarity between the core feature sequence of the current coupling deviation and each degenerate feature template in the pattern library. A similarity threshold is set. When the similarity is higher than the threshold, it is determined that the current coupling deviation trend matches the corresponding degenerate pattern. At the same time, the threshold comparison results in the trend analysis are combined to cross-validate the matching effectiveness.

[0043] Based on the matching results after cross-validation, the system outputs the vehicle's current implicit progressive degradation type, degradation level, and predicted degradation evolution cycle, thus completing the identification of implicit progressive degradation patterns.

[0044] Another object of the present invention is to provide a vehicle control compensation device based on cross-system coupling relationship, the device comprising:

[0045] The acquisition module is used to collect multi-source operating data related to the steering system, drive system, braking system and vehicle posture in real time during vehicle operation;

[0046] The input module is used to input multi-source operational data into a pre-trained benchmark health model that represents dynamic coupling relationships, so as to obtain the predicted behavior signal of the vehicle in a healthy state.

[0047] The calculation module is used to compare the predicted behavior signal with the corresponding actual collected behavior signal, calculate the coupling deviation between the predicted behavior signal and the behavior signal, and extract the spatiotemporal features of the coupling deviation.

[0048] The identification module is used to perform trend analysis on coupling deviation based on the spatiotemporal characteristics of coupling deviation, and identify the vehicle's implicit progressive degradation mode by combining preset thresholds and pattern matching algorithms.

[0049] The compensation module is used to generate control commands when a degradation trend is detected, instructing the drive system or braking system to apply predictive compensation torque to control and compensate the vehicle.

[0050] Another object of the present invention is to provide a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0051] Another object of the present invention is to provide an electronic device including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0052] This invention collects multi-source operational data related to the steering system, drive system, braking system, and vehicle posture in real time during vehicle operation. This multi-source operational data is input into a pre-trained benchmark health model representing dynamic coupling relationships to obtain predicted behavioral signals under healthy conditions. The coupling deviation is calculated by comparing the predicted behavioral signals with the actual collected behavioral signals, and spatiotemporal features are extracted. Based on these spatiotemporal features, trend analysis combined with preset thresholds and pattern matching algorithms identifies implicit progressive degradation patterns. When a degradation trend is identified, a control command is generated to apply predictive compensation torque to the drive system or braking system. Multi-source operational data acquisition overcomes the limitations of monitoring single system parameters, comprehensively covering cross-system coupling and correlation information. The benchmark health model accurately represents the time-varying and complex dynamic coupling relationships between subsystems, and the spatiotemporal feature analysis of coupling deviations accurately captures subtle changes in coupling relationships. Early identification of implicit progressive degradation patterns is achieved through trend analysis and pattern matching algorithms, while the application of predictive compensation torque abandons the reactive compensation mode, realizing early intervention based on degradation trends. Ultimately, it effectively improves the timeliness and accuracy of cross-system vehicle control compensation, solving the problems in existing technologies such as neglecting the dynamic coupling relationship of subsystems, which easily leads to the accumulation of control deviations, difficulty in identifying implicit gradual degradation, and poor timeliness of compensation strategies. Attached Figure Description

[0053] Figure 1 This is a flowchart of the vehicle control compensation method based on cross-system coupling relationship in the first embodiment of the present invention;

[0054] Figure 2 This is a structural block diagram of a vehicle control compensation device based on cross-system coupling relationship in the third embodiment of the present invention.

[0055] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0056] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated 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 so that this disclosure will be thorough and complete.

[0057] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0058] 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.

[0059] Example 1

[0060] Please see Figure 1 The figure shows a vehicle control compensation method based on cross-system coupling relationship in the first embodiment of the present invention, the method including steps S10 to S14.

[0061] Step S10: During vehicle operation, collect multi-source operating data related to the steering system, drive system, braking system, and vehicle posture in real time.

[0062] This involves acquiring comprehensive operational information covering key vehicle subsystems and overall attitude, providing a data foundation for subsequent coupling relationship analysis. Real-time acquisition means the data acquisition frequency must match the rate of change in vehicle operating conditions, typically set to a high frequency of 100 Hz to 1000 Hz to ensure the capture of instantaneous changes under dynamic operating conditions. Steering system-related data may include steering angle, steering torque, and steering motor speed; drive system-related data may include drive motor output torque, motor speed, and battery output power; braking system-related data may include brake pedal travel, braking torque, and brake disc temperature; and vehicle attitude-related data may include vehicle pitch angle, roll angle, longitudinal acceleration, and lateral acceleration. The acquisition of multi-source operational data can be achieved by deploying corresponding sensors in each system. For example, a steering angle sensor collects steering angle, a torque sensor collects steering torque, and an inertial measurement unit collects vehicle attitude parameters. The acquired data is transmitted to the control unit via the vehicle bus for further processing.

[0063] Step S11: Input multi-source operational data into a pre-trained baseline health model representing dynamic coupling relationships to obtain the predicted behavior signal of the vehicle in a healthy state.

[0064] The model achieves accurate prediction of vehicle behavior under healthy conditions through pre-trained models. Representing dynamic coupling relationships means that the model does not model each system in isolation, but focuses on learning the interaction patterns between them, such as the matching relationship between steering angle changes and drive torque adjustments, and the correlation between body roll angle and braking force distribution. The baseline health model is built based on deep learning algorithms and has predictive capabilities after being trained on a large amount of health-state data. After inputting multi-source operational data, the model outputs predicted behavior signals that represent the behavioral parameters the vehicle should exhibit under the current input conditions when it is in a fully healthy state. These parameters include theoretical values ​​of drive torque and body roll angle under healthy conditions, based on the current steering angle and vehicle speed, providing a benchmark for subsequent comparison with actual behavior signals.

[0065] Specifically, historical operating data of the vehicle under different operating conditions is collected. This historical operating data includes multi-source data related to the steering system, drive system, braking system, and vehicle posture. The core of this step is to obtain rich and comprehensive training data to ensure that the model can adapt to diverse vehicle operating scenarios. Different operating conditions cover common vehicle driving states, such as constant speed driving, acceleration driving, deceleration driving, turning driving, and climbing driving, as well as different road surface conditions such as asphalt roads, cement roads, and gravel roads, and different environmental conditions such as high temperature, low temperature, rainy days, and sunny days. The collected historical data must be operating data when the vehicle is in a healthy state to avoid interference from fault state data on model training. The data type must be consistent with the data type of the real-time multi-source operating data to ensure the consistency of model input and lay the foundation for the subsequent model generalization ability.

[0066] The collected historical operational data undergoes preprocessing, including data cleaning to remove noise and outliers, and data normalization to unify data dimensions, ensuring data consistency and usability. This step is crucial for improving data quality and preventing poor-quality data from affecting model training performance. During data cleaning, noise mainly originates from electromagnetic interference during sensor acquisition and can be removed using methods such as smoothing filtering. Outliers mainly arise from abnormal data points caused by sensor malfunctions or data transmission errors and can be identified and removed using methods such as the 3σ criterion or box plots. Data normalization converts data with different dimensions to the same numerical range. For example, steering angle is converted from degrees to a normalized value of 0 to 1, and drive torque is converted from Newton-meters to a normalized value of 0 to 1. Common normalization methods include min-max normalization or z-score standardization. Normalization avoids the influence of data with different dimensions on model training weights, improving model convergence speed and prediction accuracy.

[0067] Key features are extracted from preprocessed historical operating data. These features include steering torque characteristics of the steering system, driving torque characteristics of the drive system, braking torque characteristics of the braking system, and vehicle attitude parameters such as pitch and roll angles. The core of this step is to sift through massive amounts of data to extract essential information that characterizes the operating states of each system and the coupling relationships between systems, thereby reducing the complexity of model training. The extraction of key features requires consideration of vehicle dynamics principles, selecting parameters sensitive to system operating states and coupling relationships. For example, steering torque features may include peak value, mean, and rate of change of steering torque; driving torque features may include response time and fluctuation amplitude of driving torque. These features directly reflect the operating states of each system, while the correlations between features indirectly characterize cross-system coupling characteristics, providing core input for the model to learn coupling relationships.

[0068] Based on the extracted key features, an initial baseline health model representing the dynamic coupling relationship is constructed. This model employs a deep learning neural network architecture, using multi-source key features as input and the predicted behavior signal of the vehicle in a healthy state as output. This step is the core of model construction. The deep learning neural network architecture has powerful nonlinear fitting capabilities and can effectively learn the complex dynamic coupling relationships between various systems. Commonly used architectures include recurrent neural networks or hybrid architectures of convolutional neural networks and recurrent neural networks. The model input consists of the extracted multi-source key features, such as combining steering torque features, driving torque features, braking torque features, and vehicle posture features to form a multi-dimensional input vector; the output is the predicted behavior signal in a healthy state, such as the theoretical values ​​of vehicle roll angle and driving torque predicted based on the input features, realizing the modeling of the mapping relationship from input features to healthy behavior signals.

[0069] The initial baseline health model is trained using labeled historical operational data under healthy states. During training, the backpropagation algorithm is used to continuously adjust the model's parameters, gradually reducing the error between the model's predicted behavior signal and the actual behavior signal under labeled healthy states. This step is the core process of model optimization. The labeled health state data refers to the actual health behavior signal corresponding to each input feature, forming input-output sample pairs. The backpropagation algorithm is a commonly used algorithm for training deep learning models. Its core logic is to calculate the model's output error and propagate the error back to each layer of the model through gradient descent, adjusting the weights and bias parameters of each layer. For example, when there is an error between the model's predicted driving torque and the actual healthy driving torque, the parameters of each convolutional or recurrent layer in the network are adjusted through backpropagation to gradually reduce the error, making the model's prediction result continuously approach the actual health state.

[0070] During training, cross-validation is employed, dividing historical data into training, validation, and test sets. The model is trained using the training set, its performance is evaluated and fine-tuned using the validation set, and finally, the trained model is evaluated using the test set. The core objective of this step is to prevent overfitting and ensure the model has good generalization ability. A typical data partitioning ratio is 70% training set, 15% validation set, and 15% test set. The training set is used to learn model parameters; the validation set is used to evaluate model performance during training; training stops when the model's error on the validation set no longer decreases or even increases to avoid overfitting; and hyperparameters such as the learning rate and number of network layers can be adjusted using the validation set; the test set is used to evaluate the performance of the finally trained model. Test set data is not used in model training and fine-tuning to ensure the objectivity of the evaluation results.

[0071] When the model's performance metrics on the test set meet the preset requirements, the model is determined to be a pre-trained benchmark health model representing dynamic coupling relationships. This step is a crucial stage in model acceptance. Performance metrics typically include prediction error and accuracy, such as preset prediction error of less than 5% and accuracy of more than 95%. When the error between the model's predicted behavioral signals and actual health behavioral signals on the test set is less than a preset value, and the accuracy meets the preset standard, it indicates that the model has the ability to accurately represent cross-system dynamic coupling relationships and can be used as a pre-trained model for health status prediction.

[0072] Step S12: Compare the predicted behavior signal with the corresponding actual collected behavior signal, calculate the coupling deviation between the predicted behavior signal and the behavior signal, and extract the spatiotemporal features of the coupling deviation.

[0073] The process involves using deviation analysis to capture the difference between the vehicle's actual operating state and its health state, particularly the differences at the cross-system coupling level. First, signal comparison is performed, ensuring the correspondence and synchronization of the two types of signals; that is, comparing the same type of behavioral parameters under the same time and operating conditions. The calculation of coupling deviation is not based on the deviation of a single system parameter, but rather on the overall deviation considering the coupling relationships between various systems. For example, when there is a small gap in the steering system, not only will the matching deviation between the steering angle and drive torque change, but it will also lead to a coordinated deviation between the vehicle's attitude parameters and the braking system parameters. Coupling deviation is a quantitative representation of this cross-system coordinated deviation. Spatiotemporal feature extraction then mines the changing patterns of deviation from both temporal and spatial dimensions. Temporal features can include the rate of change, cumulative amount, and periodicity of deviation, such as the rate of increase of a certain coupling deviation within 10 seconds. Spatial features can include the propagation characteristics of deviation between different systems and the correlation strength of deviations between systems, such as the degree of influence of steering system deviation on drive system and braking system deviations. The fusion of spatiotemporal features comprehensively characterizes the core characteristics of coupling deviation.

[0074] Step S13: Based on the spatiotemporal characteristics of coupling deviation, perform trend analysis on coupling deviation, and combine preset thresholds and pattern matching algorithms to identify the vehicle's implicit progressive degradation mode.

[0075] Trend analysis, based on historical and real-time spatiotemporal characteristics of coupling deviation, predicts the trend of deviation changes over a future period. For example, by fitting a deviation change curve, it predicts whether the coupling deviation will continue to increase within the next 20 seconds. The preset threshold is a critical value for deviation set based on vehicle health operation data. Different threshold standards can be set for different operating conditions; for example, the deviation threshold differs between high-speed driving and low-speed turning conditions. The pattern matching algorithm compares the spatiotemporal characteristics of the current coupling deviation with known implicit progressive degradation feature templates. Implicit progressive degradation patterns include types such as slowly increasing steering clearance and gradually decreasing drive motor efficiency. Each type corresponds to a specific spatiotemporal feature template for coupling deviation. For example, the characteristic corresponding to increased steering clearance is that the coupling deviation between the steering system and the vehicle attitude system increases linearly over time, and the rate at which the deviation propagates to the drive system gradually accelerates. By comparing the trend analysis results with the threshold, a preliminary judgment is made as to whether a degradation trend exists. Then, pattern matching is used to determine the specific degradation type, achieving accurate identification of implicit progressive degradation.

[0076] Specifically, the extracted spatiotemporal feature set of coupling bias is standardized and dimensionality reduced. Principal component analysis (PCA) is used to remove redundant features, retaining key spatiotemporal features that characterize the core variation patterns of coupling bias, resulting in a simplified core feature sequence. The core of this step is to optimize feature quality and reduce the complexity of subsequent analysis. Standardization transforms all features in the feature set to the same numerical range, avoiding the influence of dimensional differences between features on the analysis results. Dimensionality reduction is used to remove redundant features, which are those that contribute little to degenerate pattern recognition or are highly correlated with other features. For example, a correlation coefficient of 0.95 between two time-domain features can be considered redundant. PCA transforms multiple correlated features into a few uncorrelated principal component features through orthogonal transformation. These principal component features retain most of the information of the original features; for example, reducing the 20-dimensional spatiotemporal feature set to a 5-dimensional core feature set results in a core feature sequence that better focuses on the core variation patterns of coupling bias.

[0077] Based on the simplified core feature sequence, a time series trend analysis model is constructed. This model combines sliding window weighted fitting with exponential smoothing forecasting to track and predict the changing trend of coupling bias in real time and in the short term. The core of this step is to accurately grasp the changing patterns of coupling bias and predict future trends. The time series trend analysis model is specifically designed to process feature sequences that change over time. Sliding window weighted fitting uses a fixed-length sliding window (e.g., 10 data points) to weight and fit the core feature values ​​within the window, obtaining the trend curve. Real-time trend tracking is achieved by sliding the window. Exponential smoothing forecasting uses a weighted average of historical feature values, assigning higher weight to recent data, to predict feature values ​​for several future time points, such as predicting the core feature values ​​for the next five time points. The combination of these two methods allows for both real-time tracking of current trends and accurate prediction of future changes, providing a basis for judging degradation trends.

[0078] Based on vehicle health operation data under different operating conditions, a condition-adaptive multi-level threshold system is constructed. This system includes warning thresholds, degradation confirmation thresholds, and emergency intervention thresholds. The actual rate of change and predicted change of coupling deviation obtained from trend analysis are compared with the corresponding multi-level thresholds under different operating conditions to preliminarily determine whether a potential degradation trend exists. The core of this step is to establish scientific threshold standards to achieve preliminary screening of degradation trends. The condition-adaptive multi-level threshold system refers to setting corresponding multi-level thresholds for different operating conditions. For example, the warning thresholds for high-speed and low-speed operating conditions are different, avoiding misjudgments by a single threshold under different conditions. The warning threshold is the critical value for judging the possible existence of a degradation trend; when the actual rate of change or predicted change exceeds the warning threshold, a warning signal is issued. The degradation confirmation threshold is the critical value for confirming the existence of a degradation trend; exceeding this threshold preliminarily determines the existence of degradation. The emergency intervention threshold is the critical value for which immediate compensatory measures are required to ensure timely response to severe degradation. By comparing with the thresholds under the corresponding operating conditions, situations with potential degradation trends can be preliminarily screened, reducing the workload of subsequent pattern matching.

[0079] This process involves collecting historical data on different types of latent progressive degradation throughout the vehicle's entire lifecycle, extracting spatiotemporal feature templates of coupling deviations corresponding to each degradation type, labeling degradation severity levels and evolution cycles, and constructing a standardized latent degradation pattern library. The core of this step is establishing a reference template library for degradation patterns to provide a basis for pattern matching. Historical data throughout the vehicle's lifecycle includes various types of latent progressive degradation data that occur from vehicle manufacturing to scrapping, such as data on different degradation types like increased steering clearance and decreased drive motor efficiency. Feature templates are combinations of typical spatiotemporal features of coupling deviations extracted from various degradation data. For example, the feature template corresponding to increased steering clearance is a continuously increasing steering-body posture coupling deviation with a gradually accelerating rate of change. Degradation severity levels can be divided into slight degradation, moderate degradation, and severe degradation, and the evolution cycle refers to the time from the onset of degradation to its development into severe degradation. The standardized latent degradation pattern library organizes and stores information such as feature templates, severity levels, and evolution cycles for various degradation types, forming a structured reference database.

[0080] The Dynamic Time Warping (VTW) algorithm is used as the core pattern matching algorithm to calculate the similarity between the core feature sequence of the current coupling deviation and each degenerate feature template in the pattern library. A similarity threshold is set; when the similarity is higher than the threshold, the current coupling deviation trend is determined to match the corresponding degenerate pattern. Simultaneously, the threshold comparison results from trend analysis are combined to cross-validate the matching effectiveness. This step is the core of achieving accurate degenerate pattern matching, improving matching accuracy through similarity calculation and cross-validation. The VTW algorithm is an algorithm for time series similarity matching, capable of handling time series of different lengths and time scales. It is suitable for matching coupling deviation feature sequences with degenerate templates, such as calculating the similarity between the current feature sequence and the feature template of increased steering clearance in the pattern library. This algorithm maximizes the similarity between the two sequences by stretching or compressing the time axis. The similarity threshold is a preset judgment standard. For example, if the similarity threshold is set to 0.8, when the calculated similarity is higher than 0.8, it means that the current feature sequence is highly similar to the degenerate template. Cross-validation combines the threshold comparison results of the previous step. For example, if the similarity is higher than the threshold and the rate of change of coupling deviation exceeds the degradation confirmation threshold, the matching is confirmed to be valid, avoiding misjudgment by a single matching method.

[0081] Based on the matching results after cross-validation, the system outputs the vehicle's current implicit progressive degradation type, degradation level, and predicted degradation evolution cycle, thus completing the identification of implicit progressive degradation patterns. This step is the final output of degradation pattern identification, providing accurate degradation information for subsequent predictive compensation. For example, if cross-validation matches a degradation pattern of increased steering clearance, combined with the degree of change in the feature sequence, it can be determined as moderate degradation. Based on the evolution cycle data of this degradation type in the pattern library and the current trend prediction, the subsequent evolution cycle is output as 3 months. This information will be directly used to guide the generation of control commands, ensuring the pertinence and effectiveness of compensation measures.

[0082] Step S14: When a degradation trend is detected, a control command is generated, which instructs the drive system or braking system to apply predictive compensation torque to control and compensate the vehicle.

[0083] When the system confirms a latent degradation trend, it generates targeted control commands based on the type and degree of degradation. For example, if it detects an increasing steering clearance, it generates a control command to adjust the drive torque, outputting additional compensating torque through the drive system to offset the steering response delay caused by the steering clearance. If it detects a decreasing drive motor efficiency degradation trend, it applies a slight compensating torque through the braking system to adjust the vehicle's driving posture stability. The magnitude and timing of the predictive compensating torque are determined based on the degradation trend prediction results. For example, based on the rate of increase in deviation, it predicts the deviation value at a future moment and applies a corresponding amount of compensating torque in advance to ensure that the compensation effect accurately matches the degree of degradation, achieving dynamic correction of vehicle control performance and preventing further deterioration of the degradation trend.

[0084] In summary, the vehicle control compensation method based on cross-system coupling relationships in the above embodiments of the present invention collects multi-source operational data related to the steering system, drive system, braking system, and vehicle posture in real time during vehicle operation. This multi-source operational data is input into a pre-trained benchmark health model representing dynamic coupling relationships to obtain predicted behavioral signals under healthy conditions. The coupling deviation is calculated by comparing the predicted behavioral signals with the actual collected behavioral signals, and spatiotemporal features are extracted. Based on these spatiotemporal features, trend analysis combined with preset thresholds and pattern matching algorithms identifies implicit progressive degradation patterns. When a degradation trend is identified, a control command is generated to drive the drive system or braking system to apply predictive compensation torque. Multi-source operational data acquisition overcomes the limitations of monitoring single system parameters, fully covering cross-system coupling correlation information. The benchmark health model can accurately characterize the time-varying and complex dynamic coupling relationships between subsystems, and the spatiotemporal feature analysis of coupling deviation can accurately capture subtle changes in coupling relationships. Early identification of implicit progressive degradation patterns is achieved through trend analysis and pattern matching algorithms, while the application of predictive compensation torque abandons the reactive compensation mode, realizing early intervention based on degradation trends. Ultimately, it effectively improves the timeliness and accuracy of cross-system vehicle control compensation, solving the problems in existing technologies such as neglecting the dynamic coupling relationship of subsystems, which easily leads to the accumulation of control deviations, difficulty in identifying implicit gradual degradation, and poor timeliness of compensation strategies.

[0085] Example 2

[0086] This embodiment also proposes a vehicle control compensation method based on cross-system coupling relationship. The difference between the vehicle control compensation method based on cross-system coupling relationship in this embodiment and the vehicle control compensation method based on cross-system coupling relationship in Embodiment 1 is as follows:

[0087] The steps of comparing the predicted behavior signal with the corresponding actual acquired behavior signal, calculating the coupling deviation between the predicted behavior signal and the behavior signal, and extracting the spatiotemporal features of the coupling deviation include:

[0088] Simultaneous preprocessing is performed on the predicted behavior signal and the actual behavior signal. The preprocessing includes time dimension alignment and normalization.

[0089] Based on the aligned predicted behavior signal and the actual behavior signal, the single-system deviation values ​​of the predicted behavior signal and the actual behavior signal are calculated respectively. Then, through absolute value transformation and normalization, the standardized single-system deviation is obtained.

[0090] Based on the cross-system coupling characteristics of vehicles, a coupling weight matrix is ​​constructed. The element values ​​of this matrix represent the coupling correlation strength between different systems. The deviation of each single system is used as input, and weighted fusion calculation is performed in combination with the coupling weight matrix to obtain the coupling deviation of each cross-system combination.

[0091] The coupling deviations of all cross-system combinations are fused by weighted summation to obtain a global coupling deviation that can characterize the cross-system coupling state of the whole vehicle. The global coupling deviation is then preprocessed to obtain a coupling deviation sequence.

[0092] Based on the coupling bias sequence, time domain and spatial domain features are extracted respectively. The time domain features and spatial domain features are normalized and fused to form a complete coupling bias spatiotemporal feature set.

[0093] The process involves simultaneous preprocessing of the predicted and actual behavior signals. This preprocessing includes time-dimension alignment and normalization to ensure comparability between the two types of signals and prevent inaccurate deviation calculations due to time asynchrony or differences in dimensions. Time-dimension alignment matches the predicted and actual behavior signals according to their timestamps, ensuring that each predicted signal can be matched with its corresponding actual signal at that moment. For example, the predicted steering angle at the same millisecond is paired with the actual steering angle to avoid miscalculations caused by time differences. Normalization follows the same logic as data normalization, converting the two types of signals to the same numerical range to eliminate the impact of differences in dimensions. For example, both the predicted and actual driving torques are normalized to the range of 0 to 1 to ensure the rationality of the deviation calculation.

[0094] Based on the aligned predicted and actual behavior signals, the single-system deviation values ​​of the predicted and actual behavior signals are calculated separately. Then, through absolute value transformation and normalization, standardized single-system deviations are obtained, which form the basis for coupling deviation calculation. The quantification of single-system deviations is achieved first. The calculation of single-system deviation values ​​can use the difference method, i.e., the actual value of the corresponding behavior parameter of the same system minus the predicted value. For example, the steering angle deviation of the steering system is calculated as the actual steering angle minus the predicted steering angle, and the driving torque deviation of the drive system is calculated as the actual driving torque minus the predicted driving torque. Absolute value transformation converts the deviation values ​​to non-negative values, avoiding the cancellation of positive and negative deviations and more intuitively reflecting the magnitude of the deviation. Subsequent normalization transforms each single-system deviation to a uniform range; for example, the steering angle deviation and driving torque deviation are both normalized to the range of 0 to 1, laying the foundation for the subsequent fusion of cross-system coupling deviations.

[0095] Based on the cross-system coupling characteristics of vehicles, a coupling weight matrix is ​​constructed. The element values ​​of this matrix represent the coupling strength between different systems. The deviations of each individual system are used as input, and weighted fusion calculations are performed using the coupling weight matrix to obtain the coupling deviation of each cross-system combination. This is the core of the transformation from single-system deviations to coupling deviations, highlighting the impact of cross-system coupling relationships on the deviations. The coupling weight matrix is ​​a square matrix, with the number of rows and columns equal to the number of systems involved in the coupling analysis. For example, when including the steering system, drive system, braking system, and vehicle attitude system, the matrix is ​​a 4×4 square matrix. Matrix elements represent the coupling strength between two systems; the higher the strength, the larger the weight value. For example, the steering system and drive system have a high coupling strength under steering conditions, corresponding to large matrix element values; the steering system and braking system have a low coupling strength under straight-line driving conditions, corresponding to small matrix element values. Weighted fusion calculation involves multiplying the deviations of each individual system by their corresponding coupling weights and then summing the results to obtain the coupling deviation of the cross-system combination. For example, the coupling deviation of the steering-drive system combination is the steering system deviation multiplied by the steering-drive coupling weight plus the drive system deviation multiplied by the drive-steering coupling weight, thus quantifying the cross-system collaborative deviation.

[0096] A weighted summation method is used to fuse the coupling deviations of all cross-system combinations, ultimately obtaining a global coupling deviation that characterizes the cross-system coupling state of the entire vehicle. This global coupling deviation is then preprocessed to obtain a coupling deviation sequence. The core of this step is to integrate the local cross-system combination deviations into the global vehicle deviation, comprehensively reflecting the vehicle's coupling state. The weights in the weighted summation can be set according to the degree of influence of each cross-system combination on the vehicle's operational stability. For example, the steering-drive combination and the steering-body attitude combination have a greater impact on the vehicle's handling stability, and therefore have higher weights. The global coupling deviation is a comprehensive quantitative indicator that can reflect the degree of deviation in the coupling relationship between various systems. Preprocessing includes operations such as smoothing and denoising to remove random noise from the global coupling deviation, resulting in a smooth coupling deviation sequence, providing a stable data foundation for subsequent spatiotemporal feature extraction.

[0097] Based on the coupling deviation sequence, time-domain and spatial-domain features are extracted separately. These features are then normalized and fused to form a complete spatiotemporal feature set of coupling deviations. The core of this step is to comprehensively explore the changing patterns of coupling deviations, providing multi-dimensional feature support for degradation pattern recognition. Time-domain features are extracted from the temporal variation dimension of the deviation sequence, such as the mean, variance, maximum, minimum, rate of change, cumulative amount, and frequency of peak occurrence of the deviation. For example, the mean of the global coupling deviation within one minute reflects the overall level of coupling deviation during that time, and the rate of change reflects the speed of deviation change. Spatial-domain features are extracted from the distribution and propagation dimension of deviations across different systems, such as the proportion of cross-system combined deviations to the global deviation, and the time delay of deviation propagation from one system to other systems. For example, the proportion of steering-drive combined deviations to the global deviation reflects the degree of impact of the deviation in this combined coupling relationship on the entire vehicle. Normalization fusion transforms the time-domain and spatial-domain features to the same range and combines them to form a feature set, ensuring that each feature has a balanced weight in subsequent analysis.

[0098] For example, based on vehicle dynamics principles and cross-system interaction mechanisms, a set of core influencing factors affecting the coupling strength between the steering system, drive system, braking system, and vehicle posture system is identified. This set includes dynamic operating parameters, system operating state parameters, and environmental disturbance parameters. The core of this step is to identify the key variables affecting the coupling strength, providing direction for subsequent correlation calculations. Vehicle dynamics principles reveal the mechanical interactions between systems; for example, centrifugal force during steering affects vehicle posture and drive torque distribution, allowing for the identification of key influencing factors. Dynamic operating parameters include vehicle speed, steering angle, and acceleration; for example, higher vehicle speed leads to stronger coupling between the steering and drive systems. System operating state parameters include the output torque and operating temperature of each system; for example, increased drive motor temperature alters the coupling characteristics between the drive system and other systems. Environmental disturbance parameters include road surface adhesion coefficient and wind speed; for example, a decrease in road surface adhesion coefficient affects the coupling strength between the braking system and the vehicle posture system.

[0099] Based on historical operational data, parameter sequences corresponding to each core influencing factor are extracted. A dual-quantitative analysis method, combining mutual information entropy and grey relational analysis, is used to calculate the initial coupling correlation degree of any two systems under different combinations of influencing factors. This step is the core of quantifying the coupling correlation strength, improving the accuracy of correlation degree calculation through dual-quantitative analysis. The parameter sequence is a numerical sequence of each influencing factor over time extracted from historical data; for example, the vehicle speed sequence is a sequence of vehicle speed values ​​at different times. Mutual information entropy is used to calculate the degree of nonlinear correlation between the operating parameters of two systems, capturing both linear and nonlinear relationships. For example, mutual information entropy can be used to calculate the nonlinear correlation strength between the steering angle parameter sequence and the driving torque parameter sequence. Grey relational analysis is used to calculate the degree of trend fit between the parameter sequences of two systems, i.e., the similarity of the changing trends of the two sequences. For example, the closer the changing trends of the steering angle sequence and the vehicle roll angle sequence, the higher the grey relational degree. Dual-quantitative analysis fuses the calculation results of the two methods to obtain a more comprehensive and accurate initial coupling correlation degree.

[0100] Construct an initial coupling weight matrix framework with preset dimensions. The diagonal elements are set to zero, and the initial values ​​of the off-diagonal elements are set to the initial coupling correlation degree of the corresponding system combination. This step transforms the quantified coupling correlation degree into matrix form, forming the initial weight matrix. The preset dimension is determined by the number of systems participating in the coupling analysis; for example, when there are four systems, the matrix dimension is 4×4. The diagonal elements correspond to the coupling correlation between a system itself; since a single system does not have cross-system coupling, they are set to zero. The off-diagonal elements correspond to the coupling correlation between two different systems. For example, the element in the first row and second column of the matrix corresponds to the initial coupling correlation degree between the steering system and the drive system, and the element in the second row and first column corresponds to the initial coupling correlation degree between the drive system and the steering system, thus achieving a matrix-based representation of the coupling correlation strength between each system.

[0101] Based on dynamic operating condition parameters, system operating status parameters, and environmental disturbance parameters from real-time collected operational data, a working condition adaptation factor calculation model is constructed. This model takes the deviation between the real-time parameters of each core influencing factor and the standard working condition parameters as input, and outputs the working condition adaptation factor corresponding to each system combination through a radial basis function neural network. This factor characterizes the degree of gain or attenuation of the coupling strength between systems under the current working condition relative to the standard working condition. The core of this step is to quantify the impact of the difference between real-time and standard working conditions on the coupling strength. Standard working condition parameters are preset benchmark parameters, such as a vehicle speed of 60 km / h and a steering angle of 0 degrees under standard constant speed conditions. The input deviation values ​​are the differences between real-time parameters and standard parameters, such as the deviation between real-time vehicle speed and standard vehicle speed, or the deviation between real-time road surface adhesion coefficient and standard adhesion coefficient. Radial basis function neural networks are a type of neural network with powerful nonlinear mapping capabilities, suitable for factor calculation under complex operating conditions. The operating condition adaptation factor output by the neural network is a value greater than 0. When the factor is greater than 1, it indicates that the current operating condition enhances the coupling strength between the systems. When the factor is less than 1, it indicates that the current operating condition weakens the coupling strength. For example, under high-speed operating conditions, the operating condition adaptation factor of the steering-drive system combination may be 1.2, indicating that the high-speed operating condition enhances the coupling strength between the two.

[0102] The off-diagonal elements of the initial coupling weight matrix are multiplied by the corresponding system combination's operating condition adaptation factor to obtain the dynamically adjusted coupling strength value. This process updates the initial coupling weight matrix, resulting in the final vehicle cross-system coupling weight matrix. This step is the execution stage for dynamically updating the weight matrix, ensuring that the weight matrix matches the real-time operating conditions. The multiplication operation involves multiplying the initial coupling strength of each off-diagonal element by the corresponding operating condition adaptation factor to obtain the adjusted coupling strength. For example, if the initial coupling strength of the steering-drive system is 0.8 and the current operating condition adaptation factor is 1.2, the adjusted strength is 0.96. By updating all off-diagonal elements in the matrix in this way, the final coupling weight matrix can reflect the coupling strength between systems under the current operating conditions in real time, making subsequent coupling deviation calculations more consistent with the actual operating state.

[0103] For example, the nonlinear correlation between the operating parameters of the two systems is first calculated using mutual information entropy to obtain the information entropy correlation value. Then, the trend fit between the parameter sequences of the two systems is calculated using grey relational degree to obtain the grey relational value. This step is the core of the dual-quantification analysis, quantifying the degree of coupling correlation from two dimensions: nonlinear correlation and trend fit. The calculation of mutual information entropy is based on the principle of information theory. By calculating the joint entropy and the respective marginal entropy of the two parameter sequences, the mutual information entropy value is obtained. The larger the value, the stronger the nonlinear correlation between the parameters of the two systems. For example, the larger the mutual information entropy value between the steering torque sequence and the vehicle roll angle sequence, the closer the nonlinear correlation between the two. The calculation of grey relational degree involves initializing the two parameter sequences, calculating the correlation coefficient, and then calculating the mean of the correlation coefficient to obtain the grey relational value. The closer the value is to 1, the more consistent the changing trends of the two sequences are. For example, the grey relational value between the drive torque sequence and the vehicle speed sequence is 0.9, indicating that the changing trends of the two are highly consistent.

[0104] The initial coupling correlation degree is obtained by weighting and fusing information entropy correlation values ​​and gray correlation values ​​using the entropy weighting method, and then merging the weights based on these weights. This step is crucial for achieving the fusion of two-dimensional quantitative results, ensuring the objectivity of weight allocation through the entropy weighting method. The entropy weighting method determines weights based on the information entropy of the data itself; the lower the information entropy, the higher the discriminative power of the indicator, and the larger its weight. In practice, the information entropy of the information entropy correlation value and the gray correlation value are first calculated. Then, their weights are determined based on the information entropy. For example, if the information entropy of the information entropy correlation value is lower, a weight of 0.6 is assigned, and the weight of the gray correlation value is 0.4. The two correlation values ​​are then multiplied by their respective weights and summed to obtain the initial coupling correlation degree, thus achieving a comprehensive quantification of the coupling correlation strength between the two systems.

[0105] In the process of multimodal sensor data fusion for drive motors, due to factors such as the hardware performance and transmission lines of each sensor, some sensor channels may experience data transmission lag. This can lead to asynchronous data acquisition from these channels with other sensor data. If fusion is performed directly according to the initial weights, errors may be introduced, potentially even resulting in diagnostic errors. To address this issue, before obtaining the fused key feature parameters by weighted summation based on the initial weights, the initial weights of sensor channels with transmission lag are specially processed. First, a penalty factor is introduced to adjust the initial weights of sensor channels with transmission lag. However, simple penalty adjustment may cause excessively drastic weight changes. Therefore, an exponential moving average filter is further used to smooth the adjusted weights, making the weight changes more stable and continuous, avoiding the impact of sudden weight changes on the stability of data fusion. This effectively eliminates the negative impact of transmission lag, resulting in more accurate and reliable fused key feature parameters, thus improving the accuracy and stability of drive motor fault diagnosis.

[0106] In summary, the vehicle control compensation method based on cross-system coupling relationships in the above embodiments of the present invention collects multi-source operational data related to the steering system, drive system, braking system, and vehicle posture in real time during vehicle operation. This multi-source operational data is input into a pre-trained benchmark health model representing dynamic coupling relationships to obtain predicted behavioral signals under healthy conditions. The coupling deviation is calculated by comparing the predicted behavioral signals with the actual collected behavioral signals, and spatiotemporal features are extracted. Based on these spatiotemporal features, trend analysis combined with preset thresholds and pattern matching algorithms identifies implicit progressive degradation patterns. When a degradation trend is identified, a control command is generated to drive the drive system or braking system to apply predictive compensation torque. Multi-source operational data acquisition overcomes the limitations of monitoring single system parameters, fully covering cross-system coupling correlation information. The benchmark health model can accurately characterize the time-varying and complex dynamic coupling relationships between subsystems, and the spatiotemporal feature analysis of coupling deviation can accurately capture subtle changes in coupling relationships. Early identification of implicit progressive degradation patterns is achieved through trend analysis and pattern matching algorithms, while the application of predictive compensation torque abandons the reactive compensation mode, realizing early intervention based on degradation trends. Ultimately, it effectively improves the timeliness and accuracy of cross-system vehicle control compensation, solving the problems in existing technologies such as neglecting the dynamic coupling relationship of subsystems, which easily leads to the accumulation of control deviations, difficulty in identifying implicit gradual degradation, and poor timeliness of compensation strategies.

[0107] Example 3

[0108] Please see Figure 2 The figure shows a vehicle control compensation device based on cross-system coupling relationship proposed in the third embodiment of the present invention. The device includes:

[0109] The acquisition module 100 is used to collect multi-source operating data related to the steering system, drive system, braking system and vehicle posture in real time during vehicle operation.

[0110] The input module 200 is used to input multi-source operating data into a pre-trained benchmark health model that represents dynamic coupling relationships, so as to obtain the predicted behavior signal of the vehicle in a healthy state.

[0111] The calculation module 300 is used to compare the predicted behavior signal with the corresponding actual collected behavior signal, calculate the coupling deviation between the predicted behavior signal and the behavior signal, and extract the spatiotemporal features of the coupling deviation.

[0112] The identification module 400 is used to perform trend analysis on coupling deviation based on the spatiotemporal characteristics of coupling deviation, and identify the implicit progressive degradation pattern of the vehicle by combining a preset threshold and a pattern matching algorithm.

[0113] The compensation module 500 is used to generate control commands when a degradation trend is detected, instructing the drive system or braking system to apply predictive compensation torque to control and compensate the vehicle.

[0114] The functions or operation steps implemented by the above modules are largely the same as those in the above method embodiments, and will not be repeated here.

[0115] Example 4

[0116] In another aspect, the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of the method described in any one of Embodiments 1 to 2 above.

[0117] Example 5

[0118] In another aspect, the present invention provides an electronic device, the electronic device including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps of any one of the methods described in Embodiments 1 to 2 above.

[0119] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0120] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable instructions for implementing logical functions, and can be embodied in any computer-readable storage medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable storage medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0121] More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable storage media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0122] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0123] 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 the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0124] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A vehicle control compensation method based on cross-system coupling relationship, characterized in that, The method includes: During vehicle operation, multi-source operating data is collected in real time, including steering system data, drive system data, braking system data, and vehicle attitude-related data. Multi-source operational data is input into a pre-trained baseline health model that represents dynamic coupling relationships to obtain predicted behavior signals of the vehicle in a healthy state. By comparing the predicted behavior signal with the corresponding actual collected behavior signal, the coupling deviation between the predicted behavior signal and the actual behavior signal is calculated and the spatiotemporal features of the coupling deviation are extracted. Based on the spatiotemporal characteristics of coupling bias, trend analysis of coupling bias is performed, and combined with preset thresholds and pattern matching algorithms, the implicit progressive degradation mode of vehicles is identified. When a degradation trend is detected, control commands are generated to control the drive or braking system to apply predictive compensation torque to control and compensate the vehicle.

2. The vehicle control compensation method based on cross-system coupling relationship according to claim 1, characterized in that, The training process of the benchmark health model includes: Collect historical operating data of the vehicle under different operating conditions. The historical operating data includes multi-source data related to the steering system, drive system, braking system and vehicle posture. The collected historical operational data is preprocessed, including data cleaning to remove noise and outliers, and data normalization to unify data units, to ensure the consistency and usability of data quality. Key features are extracted from the preprocessed historical operating data, including steering torque characteristics of the steering system, driving torque characteristics of the drive system, braking torque characteristics of the braking system, and pitch and roll angles of the vehicle body. Based on the extracted key features, an initial benchmark health model representing the dynamic coupling relationship is constructed. This model adopts a deep learning neural network architecture, takes multi-source key features as input, and takes the predicted behavior signal of the vehicle in a healthy state as output. The initial baseline health model is trained using historical operational data under labeled health conditions. During the training process, the backpropagation algorithm is used to continuously adjust the model parameters, so that the error between the predicted behavior signal output by the model and the behavior signal under the actual labeled health conditions gradually decreases. During the training process, cross-validation is used to divide the historical running data into training set, validation set and test set. The model is trained using the training set, the performance of the model is evaluated and tuned using the validation set, and finally the performance of the trained model is evaluated using the test set. When the model's performance metrics on the test set meet the preset requirements, the model is determined to be a pre-trained benchmark healthy model representing dynamic coupling relationships.

3. The vehicle control compensation method based on cross-system coupling relationship according to claim 1, characterized in that, The steps of comparing the predicted behavior signal with the corresponding actual acquired behavior signal, calculating the coupling deviation between the predicted behavior signal and the behavior signal, and extracting the spatiotemporal features of the coupling deviation include: Simultaneous preprocessing is performed on the predicted behavior signal and the actual behavior signal. The preprocessing includes time dimension alignment and normalization. Based on the aligned predicted behavior signal and the actual behavior signal, the single-system deviation values ​​of the predicted behavior signal and the actual behavior signal are calculated respectively. Then, through absolute value transformation and normalization, the standardized single-system deviation is obtained. Based on the cross-system coupling characteristics of vehicles, a coupling weight matrix is ​​constructed. The element values ​​of this matrix represent the coupling correlation strength between different systems. The deviation of each single system is used as input, and weighted fusion calculation is performed in combination with the coupling weight matrix to obtain the coupling deviation of each cross-system combination. The coupling deviations of all cross-system combinations are fused by weighted summation to obtain a global coupling deviation that can characterize the cross-system coupling state of the whole vehicle. The global coupling deviation is then preprocessed to obtain a coupling deviation sequence. Based on the coupling bias sequence, time domain and spatial domain features are extracted respectively. The time domain features and spatial domain features are normalized and fused to form a complete coupling bias spatiotemporal feature set.

4. The vehicle control compensation method based on cross-system coupling relationship according to claim 3, characterized in that, The step of constructing a coupling weight matrix based on the cross-system coupling characteristics of vehicles, where the element values ​​of the matrix characterize the coupling strength between different systems, includes: Based on vehicle dynamics principles and cross-system interaction mechanisms, a set of core influencing factors is identified that affects the coupling strength between the steering system, drive system, braking system, and vehicle posture system. The set of core influencing factors includes dynamic operating condition parameters, system operating state parameters, and environmental disturbance parameters. Based on historical operational data, parameter sequences corresponding to each core influencing factor are extracted. A dual-quantitative analysis method coupling mutual information entropy and grey relational degree is adopted to calculate the initial coupling correlation degree of any two systems under different combinations of influencing factors. Construct an initial coupling weight matrix framework with a preset dimension. Set the diagonal elements of the matrix to zero and set the initial values ​​of the off-diagonal elements to the initial coupling correlation degree of the corresponding system combination.

5. The vehicle control compensation method based on cross-system coupling relationship according to claim 4, characterized in that, The step of constructing an initial coupling weight matrix framework of a preset dimension, setting the diagonal elements of the matrix to zero and the initial values ​​of the off-diagonal elements to the initial coupling correlation degree of the corresponding system combination, further includes: Based on the dynamic operating condition parameters, system operating status parameters, and environmental interference parameters in the real-time collected operating data, an operating condition adaptation factor calculation model is constructed. The model takes the deviation values ​​between the real-time parameters of each core influencing factor and the standard operating condition parameters as input, and outputs the operating condition adaptation factor corresponding to each system combination through a radial basis neural network. This factor represents the degree of gain or attenuation of the coupling correlation strength between the current operating condition and the standard operating condition. The off-diagonal elements in the initial coupling weight matrix are multiplied with the operating condition adaptation factors of the corresponding system combination to obtain the dynamically adjusted coupling correlation strength value. The initial coupling weight matrix is ​​then updated to obtain the final vehicle cross-system coupling weight matrix.

6. The vehicle control compensation method based on cross-system coupling relationship according to claim 5, characterized in that, The steps of using the dual-quantitative analysis method that couples mutual information entropy and grey relational degree to calculate the initial coupling correlation degree of any two systems under different combinations of influencing factors include: First, the degree of nonlinear correlation between the operating parameters of the two systems is calculated by mutual information entropy to obtain the information entropy correlation value. Then, the degree of trend fit between the parameter sequences of the two systems is calculated by grey relational degree to obtain the grey relational value. Based on the entropy weight method, the weight ratios of information entropy correlation value and gray correlation value are allocated, and the initial coupling correlation degree under the combination of various influencing factors is obtained by weighted fusion.

7. The vehicle control compensation method based on cross-system coupling relationship according to claim 1, characterized in that, The steps for identifying the implicit progressive degradation pattern of a vehicle by performing trend analysis on the coupling deviation based on the spatiotemporal characteristics of the coupling deviation, combined with a preset threshold and a pattern matching algorithm, include: The extracted spatiotemporal feature set of coupling deviation is standardized and dimensionality reduced. Principal component analysis algorithm is used to remove redundant features and retain key spatiotemporal features that can characterize the core change law of coupling deviation, resulting in a simplified core feature sequence. Based on the simplified core feature sequence, a time series trend analysis model is constructed. A combination of sliding window weighted fitting and exponential smoothing prediction is used to track the changing trend of coupling bias in real time and make short-term predictions. Based on the health operation data of vehicles under different operating conditions, a multi-level threshold system for operating condition adaptation is constructed. The multi-level thresholds include warning thresholds, degradation confirmation thresholds, and emergency intervention thresholds. The actual rate of change and predicted change of the coupling deviation obtained from trend analysis are compared with the multi-level thresholds under the corresponding operating conditions to preliminarily determine whether there is a potential degradation trend. Collect historical data on different types of implicit progressive degradation throughout the entire life cycle of a vehicle, extract spatiotemporal feature templates of coupling deviations corresponding to each degradation type, label the degradation level and evolution cycle, and construct a standardized implicit degradation pattern library. The dynamic time warping algorithm is used as the core pattern matching algorithm to calculate the similarity between the core feature sequence of the current coupling deviation and each degenerate feature template in the pattern library. A similarity threshold is set. When the similarity is higher than the threshold, it is determined that the current coupling deviation trend matches the corresponding degenerate pattern. At the same time, the threshold comparison results in the trend analysis are combined to cross-validate the matching effectiveness. Based on the matching results after cross-validation, the system outputs the vehicle's current implicit progressive degradation type, degradation level, and predicted degradation evolution cycle, thus completing the identification of implicit progressive degradation patterns.

8. A vehicle control compensation device based on cross-system coupling relationship, characterized in that, The device includes: The acquisition module is used to collect multi-source operating data in real time during vehicle operation. The multi-source operating data includes steering system data, drive system data, braking system data, and vehicle attitude-related data. The input module is used to input multi-source operational data into a pre-trained benchmark health model that represents dynamic coupling relationships, so as to obtain the predicted behavior signal of the vehicle in a healthy state. The calculation module is used to compare the predicted behavior signal with the corresponding actual collected behavior signal, calculate the coupling deviation between the predicted behavior signal and the actual behavior signal, and extract the spatiotemporal features of the coupling deviation. The identification module is used to perform trend analysis on coupling deviation based on the spatiotemporal characteristics of coupling deviation, and to identify the vehicle's implicit progressive degradation mode by combining a preset threshold with a pattern matching algorithm. The compensation module is used to generate control commands when a degradation trend is detected, and to control the drive system or braking system to apply predictive compensation torque to control and compensate the vehicle.

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

10. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the program, implements the steps of the method as described in any one of claims 1 to 7.

Images