Reinforcement Learning for Cart Pole Inverted Pendulum System

Recently, reinforcement learning considered to be the chosen method to solve many problems. One of the challenging problems is controlling dynamic behaviour systems. This paper used policy gradient to balance cart pole inverted pendulum. The purpose of this paper is to balance the pole upright with the movement of the cart. The paper employed two main policy gradient-based algorithms. The results show that PG using baseline has faster episodes than reinforce PG in the training process, reinforce PG algorithm got higher accumulative reward value than PG using baseline.

Automated summary

This paper successfully tackled the challenge of controlling dynamic behavior systems using policy gradient methods. The experiments demonstrated that PG with baseline had faster convergence, while REINFORCE PG achieved higher cumulative rewards.