Adaptive neuro-predictive control of robot manipulators in work space

Abstract

This paper proposes an adaptive Nonlinear Model Predictive Controller (NMPC) for hybrid position/velocity control of robot manipulators. Robot dynamics have generally uncertainties, including parameters variations, unknown nonlinearities of the robot, payload variations, and torque disturbances form the environment. The cost function of the NMPC is defined in such a way that by adjusting its weighting parameters, the end-effector of the robot tracks a predefined geometry path in Cartesian space with a constant velocity. Moreover, in order to cope with uncertainties in the robot model neural networks with Levenberg-Marquardt training algorithm will be used to estimate adaptively the model of the robot. The closed-loop stability is demonstrated using Lyapunov theory. Simulation results of the proposed control method applied to a 3-DOF manipulator actuated by DC servomotors show satisfactory results for trajectory tracking in work space of the robot.