基于在线附加Q学习的伺服电机速度最优跟踪控制方法

Optimal Tracking Control of Servo Motor Speed Based on Online Supplementary Q-Learning

该文将在线Q学习方法与附加控制思想相结合,讨论了其在伺服系统中电机速度最优跟踪控制问题上的应用。首先在线性二次型跟踪器问题的框架下对待求解问题进行了定义;然后给出了在线附加Q学习迭代式地进行策略评价、策略改善的具体算法。仿真测试中,首先为电机速度跟踪问题设计了传统的PI控制器,然后将基于该文思路所设计的附加控制器与其并联,组成新的速度控制器。仿真结果表明,附加控制器显著改善了电机速度跟踪的动态响应特性,并且具备在被控系统参数发生改变时自动调优的自适应能力。非线性系统在特定条件下可进行局部线性化时,也可用该方法来得到更优的控制性能。

This paper combined online Q-learning with supplementary control and discussed its application to the optimal tracking control problem of servo motor speed. Firstly, the problem to be solved was defined in the framework of linear quadratic tracking. Then, the iterative algorithm of policy evaluation and policy improvement for online supplementary Q-learning was given. In the simulation test, for the motor speed tracking problem in servo system, the traditional PI controller was firstly designed, and then the supplementary controller proposed in this paper was connected to it in parallel to form a new speed controller. The simulation results showed that the additional controller significantly improves the dynamic response characteristics of motor speed tracking, and has the adaptive ability to automatically adjust when parameters of the controlled system changes. When the nonlinear system can be locally linearized under certain conditions, the proposed method can also be applied to obtain better control performance.