Intelligent All-Terrain Vehicle Robot with Movable Auxiliary Mass

Abstract

This paper presents a learning control strategy for an all-terrain vehicle robot which consists of two modules: a normal vehicle with wheels or tracks, and a moveable auxiliary mass which is a feature of this vehicle robot. Longitudinal motion of the auxiliary mass can be controlled by a DC motor in order to improve the vehicle mobility. That is, the auxiliary mass can be seen as a rider of motorcycle and utilized to change the center of gravity, the moment of inertia, adaptively corresponding to the environmental. The reinforcement learning is employed for designing a controller with neural networks. It is  demonstrated that the reinforcement learning is useful to get an effective controller under uncertain environment.