基于Q学习加权融合的无模型自适应参数寻优  

Model-Free Adaptive Parameter Optimization Based on Q-Learning Weighted Fusion

作　　者：马振恒 谢丽蓉[1] 叶金鑫 MA Zhen-heng;XIE Li-rong;YE Jin-xin(School of Electrical Engineering,Xinjiang University,Urumqi Xinjiang 830047,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830047

出　　处：《计算机仿真》2024年第6期383-389,共7页Computer Simulation

基　　金：国家自然科学基金(62163034)。

摘　　要：无模型自适应控制算法作为一种数据驱动算法,具有计算量小,无需系统精确模型,易于实现等特点。为使传统无模型自适应控制算法具有更好的控制性能,提出了一种使用Q学习对控制参数进行优化的改进无模型自适应控制方法。在此基础上,采用加权融合的方式对伪偏导数的取值优化,使其具有更好的鲁棒性。然后采用跟踪-微分器对输出数据进行滤波处理,并通过仿真进行上述方法的可行性验证。仿真结果表明,对比传统无模型自适应控制,上述方法更具有良好的控制性能和响应速度。As a data-driven algorithm,the model-free adaptive control algorithm has the characteristics of small computational complexity,low accuracy requirement of the data model,easy implementation and so on.In order to make the traditional model-free adaptive control algorithm have better control performance,an improved model-free adaptive control method using Q-learning to optimize the control parameters is proposed.On this basis,the weighted fusion method is used to optimize the value of pseudo partial derivatives,so that it has better robustness.Then the tracking differentiator is used to filter the output data,and the feasibility of this method is verified by simulation.The simulation results show that this method has better control performance and response speed than the traditional model free adaptive control.

关 键 词：无模型自适应控制 加权融合 学习 数据驱动 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]