基于Q-学习算法的SAS与EPS协调控制仿真研究  

Research and Simulation for Coordination Control of SAS and EPS Based on Q-learning Algorithm

作　　者：简文国[1] 谭光兴[1] 李珊[1] 叶鑫鹏[1] 

机构地区：[1]广西工学院电气与信息工程学院,广西柳州545006

出　　处：《计算机测量与控制》2012年第12期3253-3256,共4页Computer Measurement &Control

基　　金：广西自然科学基金(2010GXNSFA013126)资助

摘　　要：为提高汽车的行驶平顺性和转向稳定性,用Matlab/simulink平台建立了SAS(半主动悬架)与EPS(电动助力转向)的集成模型,并与Carsim整车动力学模型结合,建立联合仿真模型;在此基础上,提出了一种基于Q-学习的协调控制方法,在给定转向工况范围情况下,通过Q-学习算法获得的Q值,将其用于控制器设计,实现对SAS与EPS两个子系统进行协调控制;通过比对仿真结果中的车身横摆角速度和车身侧倾角等性能参数表明:采用Q-学习协调控制方法比常规的集成控制方法有效地降低了车身横摆角速度和车身侧倾角,更好地优化两者之间的匹配关系,使汽车行驶的平顺性和转向的稳定性得到了有效的改善,从而提高了整车的安全性能。In order to improve the vehicle ride performance and steering stability, based on Matlab/Simulink a co--simulink platform of SAS （Semi--Active Suspension） and EPS （Electric Power Steering） intergrated model combined with Carsim vehicle dynamics model was built. On this basis, a coordination control method based on Q--learning was presented, in a given range of operating conditions a controller was designed use Q value obtained through the study for coordination control of the two subsystems, achieved coordination control simulation of system; By comparing the results of the yaw angular velocity and the roll angle show that using the Q--learning coordination control methods can effectively reduce the yaw angular velocity and the roll angle better than conventional integrated control method , the match between these two parameters, the vehicle ride performance and steering stability is effectively improved, and the vehicle safety performance is also improved.

关 键 词：Q-学习 半主动悬架 电动助力转向 协调控制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]