Network inversion based neural controller for robot manipulations

This paper proposes an indirect adaptive control scheme using the concept of network inversion. The neural model of the robot manipulator was obtained by training a radial basis function network from the input-output data generated from the plant. A query based learning algorithm has been proposed to improve the model prediction which uses an extended Kalman filtering based network inversion technique. A control scheme is designed incorporating the network inversion technique. The controller ensures Lyapunov stability of the dynamic system. The proposed control scheme is implemented on a two-link manipulator through simulation. Simulation results indicate that the control scheme is robust and stable and corresponding trajectory tracking is accurate.