控制与信息技术Issue(5)：40-45,6.DOI:10.13889/j.issn.2096-5427.2025.05.005
基于迭代学习的多维泰勒网离散自适应最优控制
Discrete Adaptive Optimal Control Based Multi-Dimensional Taylor Networks Within Iterative Learning Framework
摘要
Abstract
General nonlinear discrete systems are characterized by high nonlinearity,model uncertainties,and unknown dynamics.Traditional control methods typically rely on a simplified or linearized model to represent the actual system,which leads to inherent errors.To address this limitation,this paper proposes an approximate optimal iterative dynamic programming method based on multi-dimensional Taylor networks(MTN)for general nonlinear discrete systems.Designed for online implementation throughout the control processes,this method eliminates the need for offline training steps.Within an actor-Critic framework,the approach incorporates three MTNs:a utility MTN for determining performance indicators without relying on internal dynamic information of systems;a Critic MTN for approximating performance functions;and an execution MTN for adjusting the control strategy online within the dynamic programming framework.The entire system operates as a double closed-loop control structure.The outer loop performs tracking control through main feedback signals,while the inner loop further enhances dynamic performance through auxiliary feedback signals.This article fully utilizes the structural characteristics of these MTNs,significantly reducing computational complexity of the controller and greatly improving the dynamic response speed of the iterative adaptive programming algorithm.Simulation experiments were carried out by tracking step signals and sinusoidal signal within a hydraulic servo system.The simulation results show that the multi-dimensional Taylor network discrete adaptive optimal controller based on iterative learning has good tracking performance and dynamic responsiveness,verifying the effectiveness and practicability of the method proposed in this paper.关键词
多维泰勒网/自适应控制/最优控制/动态规划/非线性系统Key words
multi-dimensional Taylor network/adaptive control/optimal control/dynamic programming/nonlinear system分类
信息技术与安全科学引用本文复制引用
张超,孙启鸣,邱亚琴..基于迭代学习的多维泰勒网离散自适应最优控制[J].控制与信息技术,2025,(5):40-45,6.基金项目
河南省高等学校青年骨干教师培养计划(2023GGJS182) （2023GGJS182）
河南省高等教育教学改革研究与实践项目(2024SJGLX0557) （2024SJGLX0557）
河南省本科高校课程思政样板课程(教学团队)"电气控制与PLC(A)"(教高[2023]431号) （教学团队）
河南省本科高校智慧教学专项研究项目(教高[2023]334号) （教高[2023]334号）
河南省科技攻关项目(252102211073) （252102211073）
江苏省高校面上项目(24KJB470016,21KJB630011) （24KJB470016,21KJB630011）
河南省线缆结构与材料重点实验室开放基金(CAMIM202505) （CAMIM202505）
