Mobile robot navigation using neural Q-learning

Continuous Q-learning algorithm has been widely used in robotic domains for its simplicity and well-developed theory. In this paper mobile robot navigation using neural Q-learning is processed. Firstly, according to our developed mobile robot CASIA-I and its working environment, an approach is proposed, used to determine the reward/penalty function of Q-learning. Secondly, after analysis of the continuous Q-learning algorithm based on the multi-layer feedforward neural network, a method for computing the weights of the hidden and output layers is given, and mobile robot navigation using neural Q-learning is implemented. At last, experimental results are included to show that the action policy obtained through Q-learning can make the mobile robot reach the destination without obstacle collision.