Research on Control of Move-in-mud Robot Based on Q Learning

In order to improve the behavior self-control ability of move-in-mud robot in unknown environment, this paper proposes a behavior control algorithm based on radial basis function neural network Q learning.The algorithm enhances the interaction between robot and environment and improves self-learning ability through using the enhanced Q learning method.By employing the radial basis function neural network to approximate the state space and Q function, the learning system has good generalization ability and effectively solves the dimension disaster problem of the state space under complex and continuous environment.Simulation experiment results show that this method not only can make move-in-mud robot have strong motion control ability, but also improve the ability of robot to adapt to the environment.