A COLLISION-FREE ALGORITHM OF A POINT-MASS ROBOT USING NEURAL NETWORKS

Abstract

In this paper an artificial neural network approach is proposed to solve the findpath problem of a mobile point-mass robot which is required to move safely to a prescribed target in a priori known workspace. A novel velocity algorithm is designed to ensure that the robot moves towards its goal at all times and remains there once reached. Our method in the construction of a collision-free path for the robot amongst a randomized number of arbitrarily configured obstacles is based on learning via the multilayer perceptron. Two neural networks are used to determine the direction and angle of turn of the robot so that it moves in the free space of the workspace while avoiding the obstacles in its path. The training data for the neural network are obtained using computer simulations where the initial path is traced by the user. Final- ly, simulations using point-mass robot and an anchored 2-link (RP) manipulator highlight the stabilizing controls of the neural network ap- proach.