A Motion Control Method For Agent Based On Dyna-Q Algorithm

Motion control, i.e. the path planning, in the unknown environment of the agent is a general interest for scholars and how to increase the efficiency of motion control is the most important point. For that, this paper puts forward a motion control method based on Dyna-Q algorithm. The motion control steps with reinforcement learning algorithm is designed and for increase the learning efficiency, a Dyna-Q framework is utilized. The experiment results show that, the method this paper proposed has more fast convergence efficiency and is more suitable for the unknown and dynamic environment to the compared methods.