Adaptive Neural Network Position/Force Hybrid Control for Constrained Reconfigurable Manipulators

An adaptive neural network position/force hybrid control scheme is proposed for reconfigurable manipulators in the presence of environmental constraint. For the development of the control law, a reduced-state dynamic model of the system is deduced firstly based on the relationship between the reconfigurable manipulators and the environmental constraints. According to the reduced dynamics, a neural network system is introduced to approximate the overall dynamics of the manipulator system by using adaptive algorithm. The effect of the approximation error, which may even make the system unstable, is removed by employing an adaptive sliding mode compensation item. In this controller, a PI-type force feedback control manifold is adopted to guarantee the force tracking along with the tracking of position and velocity. All those measures lead to a simpler controller design and easier accomplishment. The simulation results of two different configurations of two degrees of freedom reconfigurable manipulators are presented to show the effectiveness of the proposed control scheme.