An optimization method for the inverted pendulum problem based on deep reinforcement learning

Abstract The inverted pendulum problem is a classical problem. starting at random position keeps moving upwards and aims to reach an upright position. has been solved through some methods based on deep reinforcement learning (DRL) such as Deep Deterministic Policy Gradient (DDPG). However, DDPG also disadvantages. policy not conducive action exploration. Moreover, the Q value needs be estimated reasonably accurately for accurate. Nevertheless, beginning of learning, there certain difference in estimation, parameters learned this time are easy deviate. Therefore, paper combining AdaBound with algorithm proposes optimization method problem, compares performance that four published baselines. experimental results show proposed outperforms above baselines extent.