Adaptive neural network position/force decomposed control for constrained reconfigurable manipulator

This paper is concerned with the position/force control problem for constrained reconfigurable manipulator via dynamic model decomposition. Based on a nonlinear transformation, the dynamic model of the reconfigurable manipulator system is divided into the position subsystem and the force subsystem that are convenient to design the controllers respectively. In the position subsystem, an adaptive neural network is used to approximate the nonlinear term whose upper bound is unknown in different reconfigurable manipulators. And in the force subsystem, based on the computed torque method, another adaptive neural network is employed to compensate the model uncertainties. Then, on the basis of Lyapunov stability theorem, the state errors can converge to zero asymptotically. Finally, the effectiveness of the proposed control scheme is demonstrated by the simulations of two 2-DOF constrained reconfigurable manipulators with different configurations. The superiorities of this scheme lie in that the control strategy can be used in reconfigurable manipulators with different configurations. Meanwhile, it is more convenient to be applied in practice due to the decomposed dynamics. In contrast to the previous works, the force tracking error can asymptotically converge to zero rather than tracking error bounded.