Design of reinforce learning control algorithm and verified in inverted pendulum

Abstract

The reinforce leaning control algorithm is studied in this paper. Two algorithms are designed using one-stage inverted pendulum as an object. One is Q-learning control algorithm, the other is also this algorithm but its parameters are optimized by BP neural network. Algorithms are verified in one-stage inverted pendulum swing up control. Experimental results show that the algorithm with BP NN optimized is more effective and stable, and this algorithm can be used in real time control.