A commercial vehicle chassis adaptive control method and system

Abstract

The application provides a kind of commercial vehicle chassis adaptive control method and system, the method includes: through the collaborative training of federal learning framework, the corresponding coupling feature vector is extracted;According to the coupling feature vector, the corresponding nonlinear sliding mode observer is constructed, the load mass change rate and the road adhesion coefficient are introduced as the gain adjustment factor of nonlinear sliding mode observer, and the corresponding observation equation is output;The joint estimation result between load total mass and road adhesion coefficient is calculated through the observation equation, the fuzzy reasoning and the defuzzification operation are combined, and the chassis comprehensive control strategy adapted to different working conditions is generated;The chassis comprehensive control strategy is issued to each actuator of chassis, and the output response data of actuator and the actual dynamics state of vehicle are synchronously collected, the deviation value of ideal state and actual state is calculated, and the adaptive correction of chassis comprehensive control strategy is completed synchronously according to the deviation value.The application can accurately control the vehicle chassis, and improves the control efficiency.

Description

Technical Field

[0001] This invention relates to the field of commercial vehicle technology, and in particular to an adaptive control method and system for commercial vehicle chassis. Background Technology

[0002] Commercial vehicles are core equipment in logistics, transportation, and engineering operations. Their driving safety is directly related to the safety of people and property, and their operational efficiency affects the development of the industry. As a key direction for the intelligent upgrading of commercial vehicles, chassis adaptive control can effectively suppress the risks of vehicle instability such as sideslip and fishtailing by dynamically adjusting the control strategy in real time through sensing the operating conditions. It is also suitable for complex and ever-changing operating scenarios such as switching between heavy load and no load and multiple road surfaces.

[0003] Existing commercial vehicle chassis control solutions mostly employ fixed parameter logic. Even those solutions that attempt adjustment rely solely on single parameters such as vehicle speed and yaw rate as core inputs, making true adaptive control difficult. Such solutions, like traditional ABS and ESP systems, can only meet the basic requirements of normal operating conditions with good road conditions and stable loads, and cannot adapt to complex, dynamically changing operating scenarios.

[0004] Furthermore, in actual commercial vehicle operation, load status (empty, fully loaded, off-center loaded) and road adhesion coefficient (dry asphalt, wet, icy / snowy roads) are dynamically changing, both of which significantly alter vehicle dynamics. Existing technologies lack joint estimation of these two key variables, leading to a mismatch between control thresholds and actual vehicle dynamic limits. Under extreme conditions such as heavy loads and low adhesion coefficients, control intervention may occur too early, too late, or even fail, failing to effectively suppress instability risks and severely restricting the improvement of chassis adaptive control capabilities. Therefore, achieving accurate joint estimation of load and road adhesion coefficient is a crucial technological bottleneck that urgently needs to be overcome in the development of adaptive control methods and systems for commercial vehicle chassis. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide an adaptive control method and system for commercial vehicle chassis to solve the problem that the existing technology lacks joint estimation of load state and road adhesion coefficient in the process of controlling commercial vehicle chassis, which leads to the risk of instability.

[0006] The first aspect of the present invention proposes:

[0007] An adaptive control method for a commercial vehicle chassis specifically includes the following steps:

[0008] The on-board sensor network of the target commercial vehicle collects load distribution data, dynamic state parameters and road image information during the vehicle's driving process, and performs collaborative training through a federated learning framework to extract the corresponding coupled feature vectors.

[0009] A corresponding nonlinear sliding mode observer is constructed based on the coupled feature vector. The load mass change rate and the road surface adhesion coefficient are introduced as gain adjustment factors of the nonlinear sliding mode observer to output the corresponding observation equation.

[0010] The joint estimation result between the total load mass and the road surface adhesion coefficient is calculated through the observation equation. Combined with fuzzy inference and defuzzification operation, a chassis integrated control strategy adapted to different working conditions is generated.

[0011] The integrated chassis control strategy is distributed to each actuator of the chassis, and the output response data of the actuator and the actual dynamic state of the vehicle are collected simultaneously to calculate the deviation between the ideal state and the actual state. The integrated chassis control strategy is then adaptively corrected based on the deviation.

[0012] The beneficial effects of this invention are as follows: This technical solution relies on the vehicle-mounted sensor network to comprehensively collect multi-dimensional data such as load distribution, dynamic state, and road surface images. Combined with federated learning and collaborative training, it ensures the accuracy and robustness of feature extraction, and achieves efficient joint estimation of load and road surface adhesion coefficient, completely breaking through the technical limitations of existing systems that rely on a single parameter. By dynamically adjusting the gain of the nonlinear sliding mode observer to adapt to changes in operating conditions, and by continuously optimizing the control strategy with fuzzy inference and adaptive correction mechanisms, it effectively avoids the risk of control failure under extreme operating conditions, significantly improves the stability, safety, and response speed of commercial vehicle chassis control, and greatly enhances the adaptability to complex and ever-changing operating scenarios.

[0013] Furthermore, the step of performing collaborative training through a federated learning framework to extract the corresponding coupled feature vectors includes:

[0014] Different working condition labels are added to the load distribution data, the dynamic state parameters, and the road surface image information, and corresponding domain data pools are constructed simultaneously based on the working condition labels.

[0015] By extracting the corresponding local initial features from the domain data pool through preset edge nodes, and simultaneously constructing screening criteria based on chassis dynamics prior constraints, redundant items in the local initial features are removed according to the screening criteria, and corresponding target features are generated simultaneously.

[0016] The target features are input into the dynamic simulation model to verify the representation capability, and the coupled feature vector is generated simultaneously based on the verification results.

[0017] Furthermore, the step of inputting the target features into the dynamic simulation model to verify the representation capability, and simultaneously generating the coupled feature vector based on the verification results, includes:

[0018] The target features are fused across dimensions and simultaneously subjected to noise reduction and enhancement processing through a variational autoencoder to generate the corresponding target feature set.

[0019] The representation errors of the target feature set in simulation and real vehicle scenarios are calculated respectively, and the corresponding error distribution is generated simultaneously. The corresponding comprehensive deviation is calculated based on the error distribution.

[0020] Based on chassis control requirements, dynamic weights are assigned to the comprehensive deviation and the target feature set through fuzzy decision-making, and simultaneously subjected to dimensionality reduction processing to generate the corresponding coupled feature vector.

[0021] Furthermore, the step of constructing a corresponding nonlinear sliding mode observer based on the coupled feature vector, and introducing the load mass change rate and road surface adhesion coefficient as gain adjustment factors for the nonlinear sliding mode observer to output the corresponding observation equation includes:

[0022] The coupled feature vector is subjected to variational mode decomposition to select key mode components that are compatible with the load mass change rate and the road surface adhesion coefficient, and weighted fusion is performed simultaneously to generate the corresponding fused feature vector.

[0023] Using the fused feature vector as input, a composite observer comprising fast terminal sliding mode and nonlinear integral sliding mode is constructed. Simultaneously, the load mass change rate and the road surface adhesion coefficient are used as linkage gain factors to dynamically adjust the gain of the composite observer.

[0024] The composite observer is iteratively optimized to output the corresponding observation equation.

[0025] Furthermore, the step of iteratively optimizing the composite observer to output the corresponding observation equation includes:

[0026] The preliminary observation output of the composite observer and the actual measured values ​​of the sensor are collected, and the corresponding observation error sequence is calculated simultaneously.

[0027] Variational mode decomposition is used to separate the corresponding systematic error, disturbance error and random error from the observation error sequence, so as to construct the target mapping relationship between the error components and the observer parameters and the linkage gain factor.

[0028] Based on the target mapping relationship, multi-vehicle data is federated to generate a corresponding extreme working condition dataset. The extreme working condition dataset is then synchronously input into the internal system of the composite observer to output the corresponding observation equation.

[0029] Furthermore, the step of calculating the joint estimation result between the total load mass and the road surface adhesion coefficient through the observation equation, and generating a chassis integrated control strategy adapted to different working conditions by combining fuzzy inference and defuzzification operations includes:

[0030] The joint estimation results are subjected to nonlinear feature enhancement processing using the kernel principal component analysis algorithm, and the vehicle driving parameters are simultaneously fused to construct the corresponding feature matrix.

[0031] Density peak clustering algorithm is used to identify and calibrate abnormal feature points under extreme conditions, so as to output the corresponding core control feature set;

[0032] The feature matrix and the core control feature set are parsed and processed to generate the corresponding chassis integrated control strategy.

[0033] Furthermore, the step of parsing the feature matrix and the core control feature set to generate the corresponding chassis integrated control strategy includes:

[0034] The feature matrix and the core control feature set are used as inputs to the graph nodes. The coupling relationship between each feature dimension is mined simultaneously through the graph neural network. Feature weights are dynamically allocated and fused based on the correlation strength to generate the corresponding control feature map.

[0035] Based on the control feature map, a corresponding adaptive neural fuzzy inference system is constructed, the fuzzy rule parameters are optimized, and the corresponding initial control quantity set is output synchronously.

[0036] A sliding mode variable structure algorithm is introduced to perform anti-interference correction on the initial control quantity, and combined with the physical constraint verification of each actuator, to generate the corresponding comprehensive chassis control strategy.

[0037] The second aspect of the present invention proposes:

[0038] An adaptive control system for a commercial vehicle chassis, wherein the system includes:

[0039] The acquisition module is used to collect load distribution data, dynamic state parameters and road image information during the vehicle's driving process based on the vehicle's on-board sensor network of the target commercial vehicle. It is then used for collaborative training through a federated learning framework to extract the corresponding coupled feature vectors.

[0040] The construction module is used to construct the corresponding nonlinear sliding mode observer based on the coupled feature vector, and introduces the load mass change rate and road surface adhesion coefficient as the gain adjustment factors of the nonlinear sliding mode observer to output the corresponding observation equation;

[0041] The calculation module is used to calculate the joint estimation result between the total load mass and the road surface adhesion coefficient through the observation equation, and generate a chassis integrated control strategy adapted to different working conditions by combining fuzzy inference and defuzzification operation.

[0042] The correction module is used to distribute the chassis integrated control strategy to each actuator of the chassis, synchronously collect the output response data of the actuator and the actual dynamic state of the vehicle, calculate the deviation value between the ideal state and the actual state, and synchronously complete the adaptive correction of the chassis integrated control strategy based on the deviation value.

[0043] Furthermore, the acquisition module is specifically used for:

[0044] Different working condition labels are added to the load distribution data, the dynamic state parameters, and the road surface image information, and corresponding domain data pools are constructed simultaneously based on the working condition labels.

[0045] By extracting the corresponding local initial features from the domain data pool through preset edge nodes, and simultaneously constructing screening criteria based on chassis dynamics prior constraints, redundant items in the local initial features are removed according to the screening criteria, and corresponding target features are generated simultaneously.

[0046] The target features are input into the dynamic simulation model to verify the representation capability, and the coupled feature vector is generated simultaneously based on the verification results.

[0047] Furthermore, the acquisition module is specifically used for:

[0048] The target features are fused across dimensions and simultaneously subjected to noise reduction and enhancement processing through a variational autoencoder to generate the corresponding target feature set.

[0049] The representation errors of the target feature set in simulation and real vehicle scenarios are calculated respectively, and the corresponding error distribution is generated simultaneously. The corresponding comprehensive deviation is calculated based on the error distribution.

[0050] Based on chassis control requirements, dynamic weights are assigned to the comprehensive deviation and the target feature set through fuzzy decision-making, and simultaneously subjected to dimensionality reduction processing to generate the corresponding coupled feature vector.

[0051] Furthermore, the building module is specifically used for:

[0052] The coupled feature vector is subjected to variational mode decomposition to select key mode components that are compatible with the load mass change rate and the road surface adhesion coefficient, and weighted fusion is performed simultaneously to generate the corresponding fused feature vector.

[0053] Using the fused feature vector as input, a composite observer comprising fast terminal sliding mode and nonlinear integral sliding mode is constructed. Simultaneously, the load mass change rate and the road surface adhesion coefficient are used as linkage gain factors to dynamically adjust the gain of the composite observer.

[0054] The composite observer is iteratively optimized to output the corresponding observation equation.

[0055] Furthermore, the building module is specifically used for:

[0056] The preliminary observation output of the composite observer and the actual measured values ​​of the sensor are collected, and the corresponding observation error sequence is calculated simultaneously.

[0057] Variational mode decomposition is used to separate the corresponding systematic error, disturbance error and random error from the observation error sequence, so as to construct the target mapping relationship between the error components and the observer parameters and the linkage gain factor.

[0058] Based on the target mapping relationship, multi-vehicle data is federated to generate a corresponding extreme working condition dataset. The extreme working condition dataset is then synchronously input into the internal system of the composite observer to output the corresponding observation equation.

[0059] Furthermore, the calculation module is specifically used for:

[0060] The joint estimation results are subjected to nonlinear feature enhancement processing using the kernel principal component analysis algorithm, and the vehicle driving parameters are simultaneously fused to construct the corresponding feature matrix.

[0061] Density peak clustering algorithm is used to identify and calibrate abnormal feature points under extreme conditions, so as to output the corresponding core control feature set;

[0062] The feature matrix and the core control feature set are parsed and processed to generate the corresponding chassis integrated control strategy.

[0063] Furthermore, the calculation module is specifically used for:

[0064] The feature matrix and the core control feature set are used as inputs to the graph nodes. The coupling relationship between each feature dimension is mined simultaneously through the graph neural network. Feature weights are dynamically allocated and fused based on the correlation strength to generate the corresponding control feature map.

[0065] Based on the control feature map, a corresponding adaptive neural fuzzy inference system is constructed, the fuzzy rule parameters are optimized, and the corresponding initial control quantity set is output synchronously.

[0066] A sliding mode variable structure algorithm is introduced to perform anti-interference correction on the initial control quantity, and combined with the physical constraint verification of each actuator, to generate the corresponding comprehensive chassis control strategy.

[0067] The third aspect of the present invention proposes:

[0068] A computer includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the commercial vehicle chassis adaptive control method as described above.

[0069] The fourth aspect of the present invention proposes:

[0070] A readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the commercial vehicle chassis adaptive control method as described above.

[0071] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0072] Figure 1 A flowchart of the commercial vehicle chassis adaptive control method provided in the first embodiment of the present invention;

[0073] Figure 2 This is a structural block diagram of the commercial vehicle chassis adaptive control system provided in the sixth embodiment of the present invention.

[0074] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0075] 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.

[0076] 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.

[0077] 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.

[0078] Please see Figure 1 The diagram shows the adaptive control method for commercial vehicle chassis provided in the first embodiment of the present invention. The adaptive control method for commercial vehicle chassis provided in this embodiment can automatically adjust the control strategy of the commercial vehicle chassis, thereby improving the control efficiency of the chassis.

[0079] Specifically, this embodiment provides:

[0080] An adaptive control method for a commercial vehicle chassis specifically includes the following steps:

[0081] Step S10: Based on the vehicle-mounted sensor network of the target commercial vehicle, load distribution data, dynamic state parameters and road image information during the vehicle's driving process are collected, and collaborative training is carried out through a federated learning framework to extract the corresponding coupled feature vectors.

[0082] It should be noted that, firstly, addressing the shortcomings of traditional chassis control data acquisition which is "single-dimensional and does not consider the coupling effect of multi-source data," this approach collects three types of core data based on the onboard sensor network of the target commercial vehicle (such as load sensors, inertial measurement units, high-definition cameras, tire pressure monitoring sensors, etc.): load distribution data (such as axle load distribution and cargo center of gravity position, which directly affect the chassis's load-bearing characteristics and dynamic response), dynamic state parameters (such as vehicle speed, acceleration, yaw rate, and tire slip angle, reflecting the vehicle's real-time driving state), and road surface image information (such as road surface type, smoothness, and water and snow accumulation, which determine the road surface adhesion coefficient and driving resistance). Collaborative training is then performed using a federated learning framework. Specifically, federated learning can achieve collaborative feature extraction of multi-source heterogeneous data without leaking the data privacy of each sensor node, avoiding the data silo problem, while improving the generalization ability of features. Ultimately, it extracts a coupled feature vector representing the coupling relationship between "load-dynamics-road surface," providing high-quality input for subsequent state estimation.

[0083] Step S20: Construct the corresponding nonlinear sliding mode observer based on the coupled feature vector, and introduce the load mass change rate and road surface adhesion coefficient as the gain adjustment factors of the nonlinear sliding mode observer to output the corresponding observation equation;

[0084] It should be noted that, secondly, to address the problem of traditional state observers' "linear assumptions and inability to cope with the strong nonlinear dynamic characteristics of commercial vehicles," a nonlinear sliding mode observer is constructed based on the coupled eigenvectors. Specifically, the sliding mode observer possesses strong robustness and can effectively cope with system parameter perturbations and external disturbances. The load mass change rate (such as mass changes caused by cargo loading and unloading and load transfer during driving) and the road adhesion coefficient (such as the adhesion difference between dry / wet / ice roads) are introduced as gain adjustment factors for the observer. These two factors are core variables affecting the chassis dynamic response. Dynamically adjusting the gain allows the observer to adapt to different loads and road conditions, improves the accuracy of state estimation, and ultimately outputs an observation equation that accurately characterizes the relationship between load and road conditions.

[0085] Step S30: Calculate the joint estimation result between the total load mass and the road surface adhesion coefficient through the observation equation, and generate a chassis integrated control strategy adapted to different working conditions by combining fuzzy inference and defuzzification operation.

[0086] It should be noted that, next, the joint estimation results of the total load mass and the road adhesion coefficient are calculated through the observation equation. Specifically, the joint estimation can avoid the one-sidedness of single-state estimation (e.g., estimating only the load will ignore the influence of road adhesion on braking / steering). The chassis integrated control strategy is generated by combining fuzzy inference and defuzzification operation: fuzzy inference is good at handling complex and uncertain nonlinear systems, and can match the corresponding control rules (e.g., increasing the braking pressure threshold of the anti-lock braking system (ABS) and adjusting the yaw moment distribution ratio of the electronic stability program (ESP)) according to the joint estimation results (e.g., "full load + wet road surface" and "no load + dry road surface"). Defuzzification operation transforms the fuzzy control rules into clear control instructions that can be recognized by the actuators, so as to achieve precise adaptation of the control strategy and the working conditions.

[0087] Step S40: The integrated chassis control strategy is sent to each actuator of the chassis. The output response data of the actuator and the actual dynamic state of the vehicle are collected simultaneously to calculate the deviation between the ideal state and the actual state. The integrated chassis control strategy is then adaptively corrected based on the deviation.

[0088] It should be noted that, finally, the chassis integrated control strategy is distributed to each actuator (such as the braking system, steering system, and suspension system). Simultaneously, the output response data of the actuators (such as braking pressure, steering assist torque, and suspension damping force) and the actual dynamic state of the vehicle (such as actual yaw rate and braking distance) are collected. The deviation between the ideal state (observer estimate) and the actual state is calculated. Based on the deviation, the control strategy is adaptively corrected (e.g., when the actual yaw rate is greater than the ideal value, steering assist is reduced or ESP yaw torque is adjusted), forming a closed-loop control of "perception-decision-execution-correction" to ensure that the chassis control performance is always in the optimal state.

[0089] Second Embodiment

[0090] Furthermore, the step of performing collaborative training through a federated learning framework to extract the corresponding coupled feature vectors includes:

[0091] Different working condition labels are added to the load distribution data, the dynamic state parameters, and the road surface image information, and corresponding domain data pools are constructed simultaneously based on the working condition labels.

[0092] By extracting the corresponding local initial features from the domain data pool through preset edge nodes, and simultaneously constructing screening criteria based on chassis dynamics prior constraints, redundant items in the local initial features are removed according to the screening criteria, and corresponding target features are generated simultaneously.

[0093] The target features are input into the dynamic simulation model to verify the representation capability, and the coupled feature vector is generated simultaneously based on the verification results.

[0094] It should be noted that, firstly, to achieve accurate data classification under different working conditions, load distribution data, dynamic state parameters, and road surface image information are labeled with working condition tags (e.g., load tags: unloaded / half-loaded / fully loaded; road surface tags: dry / slippery / snow-covered; driving tags: straight / turning / braking). Based on the working condition tags, a domain-specific data pool is constructed. Specifically, the domain-specific data pool can classify and integrate multi-source data under the same working condition, avoiding feature distortion caused by the mixing of data from different working conditions, and providing an accurate data foundation for subsequent local feature extraction.

[0095] Secondly, based on preset edge nodes (such as vehicle controllers and sensor local processing units), local initial features are extracted from the domain data pool: features such as axle load ratio and center of gravity height are extracted from load distribution data; features such as peak yaw rate and rate of change of acceleration are extracted from dynamic state parameters; and features such as road texture and grayscale gradient are extracted from road image information. At the same time, feature selection criteria are constructed by combining chassis dynamics prior constraints (such as the correlation between axle load ratio and tire lateral stiffness, and the physical model of vehicle speed and braking distance). Redundant items in the local initial features (such as road image color features and instantaneous noise features of load sensors that are not related to chassis control) are removed according to the criteria to generate target features. Specifically, the introduction of prior constraints can ensure that the target features are strongly correlated with the chassis dynamic characteristics, thereby improving the effectiveness and relevance of the features.

[0096] Finally, to verify the ability of the target features to represent the chassis dynamics, the target features are input into a preset chassis dynamics simulation model (such as CarSim or TruckSim simulation model): the vehicle driving state corresponding to different target features is simulated, and the matching degree between the dynamic parameters (such as theoretical yaw rate and theoretical braking deceleration) output by the simulation and the feature input is compared; based on the verification results, target features with strong representation ability (matching degree ≥ 90%) are selected and fused across dimensions to finally generate a coupled feature vector. Specifically, the simulation verification can eliminate invalid features in advance to avoid the accumulation of errors in the subsequent state estimation and strategy generation stages.

[0097] Furthermore, the step of inputting the target features into the dynamic simulation model to verify the representation capability, and simultaneously generating the coupled feature vector based on the verification results, includes:

[0098] The target features are fused across dimensions and simultaneously subjected to noise reduction and enhancement processing through a variational autoencoder to generate the corresponding target feature set.

[0099] The representation errors of the target feature set in simulation and real vehicle scenarios are calculated respectively, and the corresponding error distribution is generated simultaneously. The corresponding comprehensive deviation is calculated based on the error distribution.

[0100] Based on chassis control requirements, dynamic weights are assigned to the comprehensive deviation and the target feature set through fuzzy decision-making, and simultaneously subjected to dimensionality reduction processing to generate the corresponding coupled feature vector.

[0101] It should be noted that, firstly, since the target features come from different sensors, there are issues of data noise and dimensional heterogeneity. Therefore, the target features are fused across dimensions (e.g., multiplying load features with road surface features to represent the "load-road surface" synergistic effect). Simultaneously, noise reduction and enhancement are performed through a variational autoencoder. The variational autoencoder can remove sensor noise (such as image noise from the camera and vibration noise from the load sensor) through the encoding-decoding process, restore the true physical meaning of the features, and generate a target feature set with high signal-to-noise ratio and uniform dimensions. Specifically, noise reduction and enhancement are key prerequisites for improving the accuracy of subsequent state estimation.

[0102] Secondly, to ensure the consistency of the target feature set in simulation and real vehicle scenarios, the representation error of the target feature set in simulation scenarios (such as simulation models under standard operating conditions) and real vehicle scenarios (such as actual road tests) is calculated separately: the simulation representation error is the deviation between the feature input and the simulation dynamics output, and the real vehicle representation error is the deviation between the feature input and the real vehicle dynamics acquisition data; based on the two types of errors, an error distribution (such as normal distribution and skewed distribution) is generated, and the comprehensive deviation is calculated according to the mean and variance of the error distribution. Specifically, the comprehensive deviation quantifies the generalization ability of the target feature set. The smaller the deviation, the stronger the applicability of the feature in simulation and real vehicle scenarios.

[0103] Finally, based on the actual needs of chassis control (such as braking control prioritizing road surface adhesion and load characteristics, and steering control prioritizing yaw rate and vehicle speed characteristics), dynamic weights are assigned to the comprehensive deviation and target feature set through fuzzy decision-making: features with small comprehensive deviations are given high weights, and features of key dimensions of chassis control are given high weights; dimensionality reduction processing is performed simultaneously (such as reducing feature dimensions through principal component analysis), eliminating redundant features with low weights, and finally generating coupled feature vectors that balance accuracy, generalization, and computational efficiency. Specifically, dynamic weight allocation ensures that the feature vectors focus on the core needs of chassis control, while dimensionality reduction reduces the computational power consumption of subsequent observers and control strategies.

[0104] Furthermore, the step of constructing a corresponding nonlinear sliding mode observer based on the coupled feature vector, and introducing the load mass change rate and road surface adhesion coefficient as gain adjustment factors for the nonlinear sliding mode observer to output the corresponding observation equation includes:

[0105] The coupled feature vector is subjected to variational mode decomposition to select key mode components that are compatible with the load mass change rate and the road surface adhesion coefficient, and weighted fusion is performed simultaneously to generate the corresponding fused feature vector.

[0106] Using the fused feature vector as input, a composite observer comprising fast terminal sliding mode and nonlinear integral sliding mode is constructed. Simultaneously, the load mass change rate and the road surface adhesion coefficient are used as linkage gain factors to dynamically adjust the gain of the composite observer.

[0107] The composite observer is iteratively optimized to output the corresponding observation equation.

[0108] It should be noted that, firstly, for the multimodal information contained in the coupled feature vector (such as low-frequency modes of slowly changing load and high-frequency modes of rapidly changing pavement adhesion), a variational mode decomposition algorithm is used to decompose the coupled feature vector into several modal components. Key modal components that are compatible with the load mass change rate and pavement adhesion coefficient (such as low-frequency modes corresponding to load and high-frequency modes corresponding to pavement adhesion) are selected and weighted and fused (low-frequency mode weight 0.6, high-frequency mode weight 0.4, adapting to the characteristics of slow load change and rapid pavement change) to generate a fused feature vector. Specifically, the selection of key modal components can focus on core influencing factors and avoid irrelevant modes from interfering with the observation accuracy.

[0109] Secondly, to improve the observer's response speed and anti-interference capability, a composite observer comprising fast terminal sliding mode and nonlinear integral sliding mode is constructed using fused feature vectors as input. The fast terminal sliding mode has finite-time convergence characteristics and can quickly track the dynamic changes of load and road surface. The nonlinear integral sliding mode can eliminate the steady-state error of traditional sliding mode observers and improve the accuracy of state estimation. At the same time, the load mass change rate and road surface adhesion coefficient are used as linkage gain factors to dynamically adjust the gain of the composite observer. Specifically, when the load mass change rate is large (such as during cargo loading and unloading), the observer gain is increased to accelerate the convergence speed; when the road surface adhesion coefficient is low (such as on a wet and slippery road surface), the gain is increased to improve the observer's anti-interference capability, thus achieving condition-adaptive adjustment of the observer gain.

[0110] Finally, the composite observer is iteratively optimized: multiple sets of dynamic data under load and road conditions are collected through real vehicle testing and used as training samples for the observer; the gradient descent algorithm is used to iteratively adjust the parameters of the observer (such as sliding surface parameters and integral coefficients) to minimize the error between the observer output and the actual measured values ​​of the real vehicle; after multiple rounds of iterative optimization, the state estimation accuracy and robustness of the composite observer meet the preset requirements, and finally outputs an observation equation that can accurately characterize the relationship between the total load mass and the road adhesion coefficient.

[0111] Furthermore, the step of iteratively optimizing the composite observer to output the corresponding observation equation includes:

[0112] The preliminary observation output of the composite observer and the actual measured values ​​of the sensor are collected, and the corresponding observation error sequence is calculated simultaneously.

[0113] Variational mode decomposition is used to separate the corresponding systematic error, disturbance error and random error from the observation error sequence, so as to construct the target mapping relationship between the error components and the observer parameters and the linkage gain factor.

[0114] Based on the target mapping relationship, multi-vehicle data is federated to generate a corresponding extreme working condition dataset. The extreme working condition dataset is then synchronously input into the internal system of the composite observer to output the corresponding observation equation.

[0115] It should be noted that, firstly, in order to locate the source of error of the composite observer, the preliminary observation output of the composite observer (such as the estimated total load mass and road adhesion coefficient) and the actual measured values ​​of the sensors (such as the actual values ​​collected by the high-precision load cell and the road adhesion coefficient tester) are collected, and the difference between the two is calculated to obtain the observation error sequence. Specifically, the observation error sequence is the core basis for optimizing the observer parameters.

[0116] Secondly, the variational mode decomposition algorithm is used to separate the observation error sequence into three types of error components: systematic error (caused by biases in the observer model assumptions, such as linearization approximation error), interference error (caused by external environmental interference, such as crosswinds and road surface unevenness interference), and random error (caused by sensor noise). Based on the characteristics of the three types of error components, a target mapping relationship is constructed between the error components and the observer parameters (such as sliding surface parameters) and the linkage gain factor (such as the load mass change rate weight). Specifically, this mapping relationship can accurately locate the observer defects corresponding to different error components, providing direction for subsequent optimization.

[0117] Finally, to improve the performance of the observer under extreme conditions (such as fully loaded uphill climbing, emergency braking, and steering on slippery roads, where traditional observers are prone to failure), an extreme condition dataset is constructed by integrating extreme condition data from multiple vehicle types (such as light trucks and heavy trucks) through a federated learning framework. The extreme condition dataset is then input into the composite observer, and the observer parameters and linkage gain factors are iteratively adjusted based on the target mapping relationship to reduce the observer's error under extreme conditions to below a preset threshold. The composite observer optimized for extreme conditions has stronger robustness and ultimately outputs an observation equation that can adapt to all operating conditions.

[0118] Furthermore, the step of calculating the joint estimation result between the total load mass and the road surface adhesion coefficient through the observation equation, and generating a chassis integrated control strategy adapted to different working conditions by combining fuzzy inference and defuzzification operations includes:

[0119] The joint estimation results are subjected to nonlinear feature enhancement processing using the kernel principal component analysis algorithm, and the vehicle driving parameters are simultaneously fused to construct the corresponding feature matrix.

[0120] Density peak clustering algorithm is used to identify and calibrate abnormal feature points under extreme conditions, so as to output the corresponding core control feature set;

[0121] The feature matrix and the core control feature set are parsed and processed to generate the corresponding chassis integrated control strategy.

[0122] It should be noted that, firstly, for the joint estimation results of the total load mass and road surface adhesion coefficient output by the observation equation (which exhibit nonlinear correlation characteristics), the kernel principal component analysis algorithm is used for nonlinear feature enhancement processing: kernel principal component analysis can map the nonlinear joint estimation results to a high-dimensional feature space through kernel functions, and extract hidden nonlinear correlation features (such as the threshold change of the influence of road surface adhesion coefficient on braking performance when the load increases); simultaneously, vehicle driving parameters (such as vehicle speed and steering angle) are fused to construct a feature matrix containing "load-road surface-driving state". Specifically, nonlinear feature enhancement can improve the feature matrix's ability to represent complex working conditions.

[0123] Secondly, to avoid interference from abnormal feature points under extreme operating conditions (such as the extremely high adhesion coefficient caused by instantaneous false alarms of sensors, and sudden changes in load values) in the generation of control strategies, a density peak clustering algorithm is used to perform cluster analysis on the feature matrix. Density peak clustering can automatically identify high-density core points (normal operating condition features) and low-density abnormal points (extreme operating condition abnormal features) in the feature matrix. Abnormal feature points are calibrated and corrected (such as using the neighborhood mean to replace abnormal values), invalid abnormal points are removed, and a core control feature set is output. Specifically, the core control feature set includes both typical features of normal operating conditions and calibrated extreme operating condition features, ensuring the full coverage of the control strategy.

[0124] Finally, the feature matrix and core control feature set are analyzed and processed: based on the technical standards for commercial vehicle chassis control (such as GB 7258 "Technical Conditions for Safe Operation of Motor Vehicles") and dynamic constraints (such as braking pressure not exceeding limits and steering assist torque not exceeding standards), the parameters in the feature matrix are mapped to specific control targets (such as braking deceleration targets and yaw rate targets); combined with the operating condition classification of the core control feature set (such as normal operating conditions and extreme operating conditions), corresponding control rules are generated (such as prioritizing economy under normal operating conditions and prioritizing safety under extreme operating conditions), and finally integrated into a comprehensive chassis control strategy.

[0125] Furthermore, the step of parsing the feature matrix and the core control feature set to generate the corresponding chassis integrated control strategy includes:

[0126] The feature matrix and the core control feature set are used as inputs to the graph nodes. The coupling relationship between each feature dimension is mined simultaneously through the graph neural network. Feature weights are dynamically allocated and fused based on the correlation strength to generate the corresponding control feature map.

[0127] Based on the control feature map, a corresponding adaptive neural fuzzy inference system is constructed, the fuzzy rule parameters are optimized, and the corresponding initial control quantity set is output synchronously.

[0128] A sliding mode variable structure algorithm is introduced to perform anti-interference correction on the initial control quantity, and combined with the physical constraint verification of each actuator, to generate the corresponding comprehensive chassis control strategy.

[0129] It should be noted that, firstly, due to the complex coupling relationship between the feature matrix and the parameters of each dimension in the core control feature set (e.g., increased load leads to a decrease in tire lateral stiffness, which in turn affects steering response), both are used as node inputs to the graph neural network: the edges of the graph neural network represent the coupling strength between parameters, and hidden coupling relationships are mined through neighborhood information aggregation (e.g., the correlation chain of load-road adhesion-braking distance); feature weights are dynamically assigned based on the correlation strength (e.g., parameters with high coupling strength are given high weights), and the features are fused to generate a control feature map. Specifically, the control feature map can intuitively present the degree of influence of each parameter on chassis control, providing a clear basis for strategy generation.

[0130] Secondly, an adaptive neuro-fuzzy inference system is constructed based on the control feature map. This system combines the uncertainty handling capability of fuzzy inference with the self-learning capability of neural networks. It can optimize fuzzy rule parameters (such as the center value and width value of the membership function) through training with real vehicle data. According to the working condition classification of the control feature map, it outputs the corresponding initial control quantity set (such as the pressure value of the braking system, the assist torque value of the steering system, and the damping coefficient value of the suspension system). Specifically, the adaptive neuro-fuzzy inference system can realize the autonomous optimization of fuzzy rules, avoiding the limitations of traditional fuzzy control rules that rely on expert experience.

[0131] Finally, to enhance the anti-interference capability of the control strategy, a sliding mode variable structure algorithm is introduced to correct the initial control quantity. The sliding mode variable structure algorithm can effectively suppress the influence of external disturbances (such as crosswinds and road surface unevenness) and parameter perturbations (such as changes in lateral stiffness caused by tire wear) on the control effect. At the same time, the corrected control quantity is verified by combining the physical constraints of each actuator (such as the upper limit of braking pressure and the range of steering assist torque), and control quantities that exceed the constraint range are eliminated. Finally, a comprehensive chassis control strategy that balances accuracy, robustness and safety is generated. Specifically, the physical constraint verification can prevent the control quantity from exceeding the hardware limits of the actuator, preventing damage to the actuator or loss of vehicle control.

[0132] Please see Figure 2 The third embodiment of the present invention provides:

[0133] An adaptive control system for a commercial vehicle chassis, wherein the system includes:

[0134] The acquisition module is used to collect load distribution data, dynamic state parameters and road image information during the vehicle's driving process based on the vehicle's on-board sensor network of the target commercial vehicle. It is then used for collaborative training through a federated learning framework to extract the corresponding coupled feature vectors.

[0135] The construction module is used to construct the corresponding nonlinear sliding mode observer based on the coupled feature vector, and introduces the load mass change rate and road surface adhesion coefficient as the gain adjustment factors of the nonlinear sliding mode observer to output the corresponding observation equation;

[0136] The calculation module is used to calculate the joint estimation result between the total load mass and the road surface adhesion coefficient through the observation equation, and generate a chassis integrated control strategy adapted to different working conditions by combining fuzzy inference and defuzzification operation.

[0137] The correction module is used to distribute the chassis integrated control strategy to each actuator of the chassis, synchronously collect the output response data of the actuator and the actual dynamic state of the vehicle, calculate the deviation value between the ideal state and the actual state, and synchronously complete the adaptive correction of the chassis integrated control strategy based on the deviation value.

[0138] Furthermore, the acquisition module is specifically used for:

[0139] Different working condition labels are added to the load distribution data, the dynamic state parameters, and the road surface image information, and corresponding domain data pools are constructed simultaneously based on the working condition labels.

[0140] By extracting the corresponding local initial features from the domain data pool through preset edge nodes, and simultaneously constructing screening criteria based on chassis dynamics prior constraints, redundant items in the local initial features are removed according to the screening criteria, and corresponding target features are generated simultaneously.

[0141] The target features are input into the dynamic simulation model to verify the representation capability, and the coupled feature vector is generated simultaneously based on the verification results.

[0142] Furthermore, the acquisition module is specifically used for:

[0143] The target features are fused across dimensions and simultaneously subjected to noise reduction and enhancement processing through a variational autoencoder to generate the corresponding target feature set.

[0144] The representation errors of the target feature set in simulation and real vehicle scenarios are calculated respectively, and the corresponding error distribution is generated simultaneously. The corresponding comprehensive deviation is calculated based on the error distribution.

[0145] Based on chassis control requirements, dynamic weights are assigned to the comprehensive deviation and the target feature set through fuzzy decision-making, and simultaneously subjected to dimensionality reduction processing to generate the corresponding coupled feature vector.

[0146] Furthermore, the building module is specifically used for:

[0147] The coupled feature vector is subjected to variational mode decomposition to select key mode components that are compatible with the load mass change rate and the road surface adhesion coefficient, and weighted fusion is performed simultaneously to generate the corresponding fused feature vector.

[0148] Using the fused feature vector as input, a composite observer comprising fast terminal sliding mode and nonlinear integral sliding mode is constructed. Simultaneously, the load mass change rate and the road surface adhesion coefficient are used as linkage gain factors to dynamically adjust the gain of the composite observer.

[0149] The composite observer is iteratively optimized to output the corresponding observation equation.

[0150] Furthermore, the building module is specifically used for:

[0151] The preliminary observation output of the composite observer and the actual measured values ​​of the sensor are collected, and the corresponding observation error sequence is calculated simultaneously.

[0152] Variational mode decomposition is used to separate the corresponding systematic error, disturbance error and random error from the observation error sequence, so as to construct the target mapping relationship between the error components and the observer parameters and the linkage gain factor.

[0153] Based on the target mapping relationship, multi-vehicle data is federated to generate a corresponding extreme working condition dataset. The extreme working condition dataset is then synchronously input into the internal system of the composite observer to output the corresponding observation equation.

[0154] Furthermore, the calculation module is specifically used for:

[0155] The joint estimation results are subjected to nonlinear feature enhancement processing using the kernel principal component analysis algorithm, and the vehicle driving parameters are simultaneously fused to construct the corresponding feature matrix.

[0156] Density peak clustering algorithm is used to identify and calibrate abnormal feature points under extreme conditions, so as to output the corresponding core control feature set;

[0157] The feature matrix and the core control feature set are parsed and processed to generate the corresponding chassis integrated control strategy.

[0158] Furthermore, the calculation module is specifically used for:

[0159] The feature matrix and the core control feature set are used as inputs to the graph nodes. The coupling relationship between each feature dimension is mined simultaneously through the graph neural network. Feature weights are dynamically allocated and fused based on the correlation strength to generate the corresponding control feature map.

[0160] Based on the control feature map, a corresponding adaptive neural fuzzy inference system is constructed, the fuzzy rule parameters are optimized, and the corresponding initial control quantity set is output synchronously.

[0161] A sliding mode variable structure algorithm is introduced to perform anti-interference correction on the initial control quantity, and combined with the physical constraint verification of each actuator, to generate the corresponding comprehensive chassis control strategy.

[0162] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the commercial vehicle chassis adaptive control method as described above.

[0163] The fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the commercial vehicle chassis adaptive control method as described above.

[0164] In summary, the commercial vehicle chassis adaptive control method and system provided in the above embodiments of the present invention can automatically adjust the control strategy of the commercial vehicle chassis, thereby improving control efficiency.

[0165] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0166] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including 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 medium" can be 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.

[0167] More specific examples of computer-readable media (a non-exhaustive list) 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 media can even be paper or other suitable media on which the program can be printed, because 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.

[0168] 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.

[0169] 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.

[0170] 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 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 the present invention should be determined by the appended claims.

Claims

1. A method for adaptive control of a commercial vehicle chassis, characterized in that, The method includes: The on-board sensor network of the target commercial vehicle collects load distribution data, dynamic state parameters and road image information during the vehicle's driving process, and performs collaborative training through a federated learning framework to extract the corresponding coupled feature vectors. A corresponding nonlinear sliding mode observer is constructed based on the coupled feature vector. The load mass change rate and the road surface adhesion coefficient are introduced as gain adjustment factors of the nonlinear sliding mode observer to output the corresponding observation equation. The joint estimation result between the total load mass and the road surface adhesion coefficient is calculated through the observation equation. Combined with fuzzy inference and defuzzification operation, a chassis integrated control strategy adapted to different working conditions is generated. The integrated chassis control strategy is distributed to each actuator of the chassis, and the output response data of the actuator and the actual dynamic state of the vehicle are collected simultaneously to calculate the deviation between the ideal state and the actual state. The integrated chassis control strategy is then adaptively corrected based on the deviation.

2. The adaptive control method for commercial vehicle chassis according to claim 1, characterized in that, The step of performing collaborative training through a federated learning framework to extract the corresponding coupled feature vectors includes: Different working condition labels are added to the load distribution data, the dynamic state parameters, and the road surface image information, and corresponding domain data pools are constructed simultaneously based on the working condition labels. By extracting the corresponding local initial features from the domain data pool through preset edge nodes, and simultaneously constructing screening criteria based on chassis dynamics prior constraints, redundant items in the local initial features are removed according to the screening criteria, and corresponding target features are generated simultaneously. The target features are input into the dynamic simulation model to verify the representation capability, and the coupled feature vector is generated simultaneously based on the verification results.

3. The adaptive control method for commercial vehicle chassis according to claim 2, characterized in that, The step of inputting the target feature into the dynamic simulation model to verify its representation capability, and simultaneously generating the coupled feature vector based on the verification result, includes: The target features are fused across dimensions and simultaneously subjected to noise reduction and enhancement processing through a variational autoencoder to generate the corresponding target feature set. The representation errors of the target feature set in simulation and real vehicle scenarios are calculated respectively, and the corresponding error distribution is generated simultaneously. The corresponding comprehensive deviation is calculated based on the error distribution. Based on chassis control requirements, dynamic weights are assigned to the comprehensive deviation and the target feature set through fuzzy decision-making, and simultaneously subjected to dimensionality reduction processing to generate the corresponding coupled feature vector.

4. The adaptive control method for commercial vehicle chassis according to claim 1, characterized in that, The steps of constructing a corresponding nonlinear sliding mode observer based on the coupled feature vector, and introducing the load mass change rate and road adhesion coefficient as gain adjustment factors for the nonlinear sliding mode observer to output the corresponding observation equation include: The coupled feature vector is subjected to variational mode decomposition to select key mode components that are compatible with the load mass change rate and the road surface adhesion coefficient, and weighted fusion is performed simultaneously to generate the corresponding fused feature vector. Using the fused feature vector as input, a composite observer comprising fast terminal sliding mode and nonlinear integral sliding mode is constructed. Simultaneously, the load mass change rate and the road surface adhesion coefficient are used as linkage gain factors to dynamically adjust the gain of the composite observer. The composite observer is iteratively optimized to output the corresponding observation equation.

5. The adaptive control method for commercial vehicle chassis according to claim 4, characterized in that, The step of iteratively optimizing the composite observer to output the corresponding observation equation includes: The preliminary observation output of the composite observer and the actual measured values ​​of the sensor are collected, and the corresponding observation error sequence is calculated simultaneously. Variational mode decomposition is used to separate the corresponding systematic error, disturbance error and random error from the observation error sequence, so as to construct the target mapping relationship between the error components and the observer parameters and the linkage gain factor. Based on the target mapping relationship, multi-vehicle data is federated to generate a corresponding extreme working condition dataset. The extreme working condition dataset is then synchronously input into the internal system of the composite observer to output the corresponding observation equation.

6. The adaptive control method for commercial vehicle chassis according to claim 1, characterized in that, The steps of calculating the joint estimation result between the total load mass and the road surface adhesion coefficient through the observation equation, and generating a chassis integrated control strategy adapted to different working conditions by combining fuzzy inference and defuzzification operations include: The joint estimation results are subjected to nonlinear feature enhancement processing using the kernel principal component analysis algorithm, and the vehicle driving parameters are simultaneously fused to construct the corresponding feature matrix. Density peak clustering algorithm is used to identify and calibrate abnormal feature points under extreme conditions, so as to output the corresponding core control feature set; The feature matrix and the core control feature set are parsed and processed to generate the corresponding chassis integrated control strategy.

7. The adaptive control method for commercial vehicle chassis according to claim 6, characterized in that, The step of parsing the feature matrix and the core control feature set to generate the chassis integrated control strategy includes: The feature matrix and the core control feature set are used as inputs to the graph nodes. The coupling relationship between each feature dimension is mined simultaneously through the graph neural network. Feature weights are dynamically allocated and fused based on the correlation strength to generate the corresponding control feature map. Based on the control feature map, a corresponding adaptive neural fuzzy inference system is constructed, the fuzzy rule parameters are optimized, and the corresponding initial control quantity set is output synchronously. A sliding mode variable structure algorithm is introduced to perform anti-interference correction on the initial control quantity, and combined with the physical constraint verification of each actuator, to generate the corresponding comprehensive chassis control strategy.

8. An adaptive control system for a commercial vehicle chassis, characterized in that, The system includes: The acquisition module is used to collect load distribution data, dynamic state parameters and road image information during the vehicle's driving process based on the vehicle's on-board sensor network of the target commercial vehicle. It is then used for collaborative training through a federated learning framework to extract the corresponding coupled feature vectors. The construction module is used to construct the corresponding nonlinear sliding mode observer based on the coupled feature vector, and introduces the load mass change rate and road surface adhesion coefficient as the gain adjustment factors of the nonlinear sliding mode observer to output the corresponding observation equation; The calculation module is used to calculate the joint estimation result between the total load mass and the road surface adhesion coefficient through the observation equation, and generate a chassis integrated control strategy adapted to different working conditions by combining fuzzy inference and defuzzification operation. The correction module is used to distribute the chassis integrated control strategy to each actuator of the chassis, synchronously collect the output response data of the actuator and the actual dynamic state of the vehicle, calculate the deviation value between the ideal state and the actual state, and synchronously complete the adaptive correction of the chassis integrated control strategy based on the deviation value.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the commercial vehicle chassis adaptive control method as described in any one of claims 1 to 7.

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

Images