基于遗传算法优化的LQR路径跟踪控制  

LQR Path Tracking Control Based on Genetic Algorithm Optimization

作　　者：王文博 赵书尚[1] 李阁强[1] 李子璋 WANG Wenbo;ZHAO Shushang;LI Geqiang;LI Zizhang(School of Electrical and Mechanical Engineering,He’nan University of Science and Technology,He’nan Luoyang 471000,China)

机构地区：[1]河南科技大学机电工程学院,河南洛阳471000

出　　处：《机械设计与制造》2025年第3期325-329,337,共6页Machinery Design & Manufacture

基　　金：山东省重点研发计划(国际科技合作专项)资助项目(2019GHZ013)。

摘　　要：针对转向机器人路径跟踪问题,设计了基于遗传算法优化的线性二次型最优控制器(LQR)。首先建立车辆二自由度“自行车”动力学模型,得到车辆横向误差模型。然后构建横向误差目标函数,设计线性二次型最优状态调节器(LQR),得到控制车辆的前轮转角。进一步地,在控制环中加入前馈控制以消除稳态误差。在此基础上,采用遗传算法迭代优化权重矩阵Q、R来改进LQR控制器。经过Simulink与Carsim联合仿真表明,采用遗传算法优化后的LQR算法相较于未优化的LQR,横向误差减小了53%,并且大大节省了找寻最优权重矩阵的时间。A linear quadratic optimal controller(LQR)based on genetic algorithm optimization is designed for the steering robot path tracking problem.Firstly,a two-degree of freedom"bicycle"dynamics model of the vehicle is established to obtain the lateral error model of the vehicle.Then the lateral error objective function is constructed and the linear quadratic optimal state regulator(LQR)is designed to obtain the front wheel rotation angle of the control vehicle.Further,feedforward control is added to the control loop to eliminate the steady-state error.Based on this,a genetic algorithm is used to iteratively optimize the weight matrices Q and R to improve the LQR controller.Simulink and Carsim simulations show that the lateral error of the optimized LQR algorithm is reduced by 53%compared with the unoptimized LQR,and the time required to find the optimal weight matrix is greatly reduced.

关 键 词：横向运动控制 路径跟踪 遗传算法(GA) 线性二次型最优控制(LQR) 

分 类 号：TH16[机械工程—机械制造及自动化] TP273[自动化与计算机技术—检测技术与自动化装置] TP242[自动化与计算机技术—控制科学与工程]