Artificial neural network based local motion planning of a wheeled mobile robot

In this paper we present an artificial neural network based motion and path planning system of a wheeled mobile robot navigating among stationary and moving obstacles. The neural network is aware of its distance sensor readings and its relative position from the target. The neural network is used in this system as a controller, and it is trained using a previously proposed extension of the backpropagation through time algorithm, which uses potential fields for obstacle avoidance. The operability of this method is presented in a series of simulation results.