Adaptive neural controller for redundant robot manipulators and collision avoidance with mobile obstacles

The paper presents a new controller approach applied to redundant robot manipulators constrained by mobile obstacles. The proposed controller is constructed in task space by using optimization strategy, in order to achieve a good trajectory tracking of the end effector even if the obstacles are fixed or mobile. The criterion to be optimized is chosen as the sum of the joint displacements energy and the internal penalty functions that take into account the obstacle positions. Any knowledge on the dynamic model is needed, only its structure. All unknown functions in the robot dynamical model, written in extended Cartesian space, are carried out using multilayer perceptron (MLP) neural networks. The adaptation laws of the neural parameters are obtained via closed loop stability analysis of the system (Lyapunov approach). In order to evaluate the proposed controller performance a 3 DOF robot manipulator evolving in a vertical space constrained by a mobile obstacle is used. The obtained results show its effectiveness. (C) 2011 Elsevier B.V. All rights reserved.