Reinforcement Learning For Walking Robot

Reinforcement learning namely Deep Q Learning neural network is used to make a robot walk in a controlled environment. The main objective of this paper is to make a robot learn to walk by building a model in hardware with six forms of movement. The basic alignment of robot at any moment depends on the direction of movement of stepper motor of a leg joint by test and trial method without any preordained model. The robot is made to freely walk which is stopped without falling down and is able to move freely over a transportable platform. Hardware robot communicates to the learning model which is run on the local machine through a USB port connection.