Feedforward control of a mobile robot using a neural network

Abstract

Many papers about mobile robot control deal only with the kinematic model of the robot. Since it is the kinematics and not the dynamics that introduces the nonholonomicity into the system, this approach provides a simple model while preserving the most interesting portion of the system. However, for large robots, or robots moving at high speeds, the dynamics of the body can not be neglected. This paper proposes a controller for a nonholonomic mobile robot developed by combining a kinematic controller based on nonsmooth discontinuous transformation and a dynamic controller based on a neural network (computed torque like). The overall system stability is proved by Lyapunov theory.