CN-122008765-A - Optimal tracking control method for multi-mode automobile suspension system

CN122008765ACN 122008765 ACN122008765 ACN 122008765ACN-122008765-A

Abstract

The invention discloses an optimal tracking control method of a multi-mode automobile suspension system, and belongs to the technical field of intelligent control of vehicles. Aiming at the problems of dynamic jump, strong uncertainty and multi-actuator cooperative control of an automobile suspension system in various driving modes, a robust tracking control framework based on interval two-type fuzzy Markov jump system modeling and combining distributed minimum maximum game and integral reinforcement learning is provided. Firstly, the unknown nonlinear dynamics, external road surface excitation and parameter random jump existing in a suspension system are considered, the system is modeled into a section two-type fuzzy Markov jump system, and secondly, a distributed control architecture is constructed aiming at the actual constraint that the dynamic part of the system is unknown and only output information can be obtained. The invention can ensure the tracking performance, the closed-loop stability and the multi-actuator cooperative efficiency of the suspension system under the environment with random switching of system modes and strong dynamic uncertainty, and remarkably improves the smoothness, the safety and the robustness of vehicle running.

Inventors

LI JIE
QI WENHAI
SUN YUXIN

Assignees

山东外国语职业技术大学
曲阜师范大学

Dates

Publication Date: 20260512
Application Date: 20260402

Claims (8)

1. An optimal tracking control method of a multi-mode automobile suspension system is characterized by comprising the following steps: S1, establishing an interval two-type model Markov jump system model, and considering an automobile suspension system and an automobile suspension model: ; Wherein the method comprises the steps of For the displacement of the vehicle body, For the displacement of the tyre, the tyre is, For the displacement of the road surface, Is used as a pitch angle of the light beam, Is a control input; Defining a state vector The model may be converted into a state space form: ; the system matrix is as follows: ; ; System parameter following Markov process The transition probability matrix for the transition, markov, is: ; s2, constructing a tracking error system and a multi-player game performance index: Defining a reference track: ; constructing a composite state And compound output Defining a tracking error: ; Wherein, the ; Definition player (actuator) The cost function of (2) is: ; Wherein the method comprises the steps of Representing divide decision maker A set of control inputs for other participants; S3, designing a distributed minimum and maximum game control framework: Constructing a cost function containing virtual opponent input: ; Wherein the method comprises the steps of For the virtual opponent input, In order to perturb the suppression level parameter, And A weight matrix which is symmetrically and positively determined; The control target is a solution minimum maximum strategy: ; S4, solving a section two-type model differential game Riccati equation: ; the optimal control law and virtual opponent policies are respectively: ; Wherein the method comprises the steps of Is an optimal gain matrix; S5, designing an offline parallel distributed integral reinforcement learning algorithm: when the system dynamics are fully known, the solution is updated using the following iterations ; ; Wherein: ; s6, designing an online parallel distributed integral reinforcement learning algorithm: When the dynamic part of the system is unknown or model uncertainty exists, an online data-driven integral reinforcement learning method is adopted, the input and output data of the system are acquired in real time, and the control strategy is updated iteratively, so that the system matrix is not required to be known in advance , And The method specifically comprises the following steps: S6.1, constructing a data driving learning equation: S6.2, defining a data acquisition matrix and vectorization representation; s6.3, constructing a linear equation set and solving; s6.4, online iterative updating strategy; S6.5, ensuring convergence and stability.
2. The method for optimal tracking control of a multi-modal vehicle suspension system according to claim 1, wherein in step S1, a fuzzy model is used as follows: fuzzy rule If (1) Is that , Is that ,..., Is that Then: ; Wherein, the As a system state vector of the system, Is the first The control inputs of the individual actuators are provided, For the output vector of the system, Is a precondition variable; Is a section type II fuzzy set; is the total number of fuzzy rules and, Is the number of the precondition variables, To be in a finite set The transition probability of the continuous time Markov process with the value of the intermediate value is as follows: ; Wherein, the And ; By fuzzy mixing, the following overall fuzzy system is derived: ; Wherein: 。
3. The optimal tracking control method for a multi-modal vehicle suspension system according to claim 1, wherein in step S2, the reference track system is designed to: ; the initial state is set as 。
4. The optimal tracking control method for a multi-modal vehicle suspension system according to claim 1, wherein in step S6.1, based on a system dynamic equation: ; defining the relation between the state increment and the cost function to obtain a data-driven learning equation:
5. The method according to claim 4, wherein in step S6.2, for facilitating on-line solution, the above equation is converted into a linear parameterized form, and the following data matrix and vector are defined: State differential matrix: ; State autocorrelation integration matrix: ; Control-state cross integration matrix: ; Virtual control-state cross integration matrix: 。
6. The optimal tracking control method for a multi-modal vehicle suspension system according to claim 5, wherein in step S6.3, the matrix is substituted into a learning equation to obtain a linear equation set: ; Wherein the method comprises the steps of Is composed of And A coefficient matrix is formed; to be solved for parameters, include And ; Is a right vector, composed of The equivalent items are formed; When the data volume satisfies When the total number of parameters is equal, solving by a least square method: 。
7. the optimal tracking control method for a multi-modal vehicle suspension system as claimed in claim 6, wherein in step S6.4, the solution is based on Extracting updated matrix And for the control law calculation at the next moment, repeating the steps S6.1 to S6.3 until the parameters converge.
8. The optimal tracking control method for a multi-modal vehicle suspension system according to claim 7, wherein in step S6.5, the policy sequence generated by the online algorithm is proved by Lyapunov analysis Convergence to an optimal solution in an iterative process And the closed loop system keeps the actual preset time stable in a limited way.

Description

Optimal tracking control method for multi-mode automobile suspension system Technical Field The invention relates to the technical field of intelligent control of vehicles, in particular to an optimal tracking control method of a multi-mode automobile suspension system. Background Automotive suspension systems are critical components that affect ride comfort, operational stability, and safety of a vehicle. Traditional passive and semi-active suspensions are limited by fixed mechanical characteristics or limited adjustment capabilities, and it is difficult to always maintain optimal performance under variable driving conditions and road surface excitation. Active suspensions output control forces in real time through force actuators, providing the possibility to achieve performance jumps, but their core challenge is how to design an effective control strategy to cope with multiple uncertainties in the system itself and in the environment. During actual vehicle travel, suspension system dynamics may jump due to load changes, component aging, or mode switching (e.g., comfort/motion mode), such behavior being suitable for stochastic modeling with Markov jump systems. Meanwhile, the strong nonlinearity, unmodeled dynamics and external random excitation of the system form deep uncertainty, and the conventional model system has limitations in dealing with the uncertainty. The interval two-degree-of-freedom modeling system introduces additional degrees of freedom through membership functions, providing a more powerful framework for characterizing and packaging such uncertainties. However, the deep fusion of the interval two-type fuzzy logic and the Markov jump system is insufficient in research for constructing a high-fidelity suspension model and designing a controller according to the high-fidelity suspension model, and particularly, a theoretical framework and a control algorithm are remarkably blank under a complex scene that the dynamic parameter part of the system is unknown and a plurality of actuators are required to work cooperatively. At the control architecture level, modern automotive suspensions are usually equipped with a plurality of independent actuators, and the realization of cooperative control thereof can be abstracted into a multi-person decision problem. The conventional optimal control method generally assumes that the system model is precisely known and optimized for a single control objective, and is difficult to directly apply to the above-mentioned game scene with multi-source uncertainty and multi-control input. The differential game theory provides a natural framework for the differential game theory, and Nash equilibrium solution can ensure collective optimality of all participant strategies under the premise of considering individual benefit conflicts or coordination. However, the existing suspension control research based on the game theory still depends on an accurate and centralized system model, and often requires full-state feedback which is difficult to directly acquire, so that the deployment and application of the suspension control research in an actual vehicle-mounted embedded system are limited. In order to reduce the dependence on an accurate model, reinforcement learning, particularly an integral reinforcement learning method, has been widely focused in the field of adaptive optimal control due to the characteristics of "model independence" and online learning. The method can learn the optimal strategy on line through the input and output data of the system, and does not need an explicit mathematical model of the internal dynamic state of the system. However, the prior study of reinforcement learning applied to suspension control has obvious limitations that firstly, a plurality of coupling models are concentrated on a single actuator system or are highly simplified, the problem of distributed learning and decision-making of a plurality of actuators in a game relation cannot be deeply discussed, secondly, the multi-actuator system is not combined with a section two-type Markov jump system model capable of accurately describing multi-mode jump and strong uncertainty of the system, thirdly, most algorithms depend on full-state feedback, and an effective solution is lacking for the ubiquitous output feedback constraint situation. In summary, the prior art is difficult to solve four core challenges of multi-mode random jump, deep fuzzy uncertainty, multi-actuator game optimization and output feedback constraint faced by an active suspension system of an automobile. Therefore, the development of the distributed, model-independent and robust optimal tracking control method can be realized in the complex environment, and the method has urgent theoretical requirements and great engineering application value for improving the overall performance and adaptability of the intelligent vehicle chassis system. Disclosure of Invention Aiming at an automobile suspension system with multi-