Deep Reinforcement Learning - based Rotary Inverted Pendulum

Abstract

Artificial Intelligence systems have become a very important term in many engineering fields. Intelligent robot systems are an example of Intelligence systems (IS). To develop an IS, we need Intelligent control methods specially when the dimension of state space gets very huge and it is impossible to cover all states in traditional programming techniques with feasible time. This project aims to develop an intelligent algorithm based on deep reinforcement learning that can learn from the environment to make the rotary inverted pendulum stable in an upward position. The system is designed to be realized on a Raspberry Pi microcontroller. TensorFlow was used for implementing the deep reinforcement learning algorithm. The simulation result showed that the system was able to learn to balance the pendulum in an upward position that starts from a random state in a proper range