Neural Networks for Redundant Robot Manipulators Control with Obstacles Avoidance

Abstract

In this paper, neural networks of MLP type are used to control constrained redundant robot manipulators with obstacles. The proposed controller is determined using extended Cartesian space to minimise the joint displacements and to avoid obstacles. The neural networks have been used to approximate separately, the functions of the dynamic model of the robot manipulator expressed in the Cartesian space. The adaptation laws weights of each neural network, are obtained via stability study in Lyapunov sense of the system in closed loop. The performances of the proposed control approach are tested on a 3-degree of freedom robot manipulators involving in the vertical space.