直接学习控制能够充分利用系统的先前控制经验,直接产生期望控制信号,不需要任何重复的学习过程。本文针对一类具有任意相对度的线性时不变系统,给出了多尺度直接学习控制方法,当目标轨迹发生变化时,即对于数值尺度和时间尺度与先前目标轨迹都不同的新轨迹,运用所给的直接学习控制方法,不需要重新学习,只需一步学习就能获得系统的期望控制信号。仿真结果说明了该方法的有效性。
Direct learning control (DLC) can fully utilize the pre- stored control profiles and it can directly generate the desired control without any repetitive learning process. In this paper, Multi - Scale DLC method is proposed for a class of linear time - in- variant systems with arbitrary relative degree. The desired control can be obtained explicitly without relearning for a new trajectory which is different from all the previous ones in both magnitude scale and time scale. The simulation example shows the effectiveness of this approach.