Neural network based adaptive impedance control of constrained robots

To achieve the desired dynamic impedance, traditional impedance control requires an exact dynamic modeling of the robot and the environment. Though recently some robust control and adaptive control schemes were incorporated in the impedance control for uncertain constrained robot systems, most of them still require some exact information of the modelling such as the regressor matrix and nominal values of the dynamic terms. In this paper, a model free neural network based adaptive impedance control scheme is developed. The controller does not require dynamic modelling of the system, and the weights of the neural network are tuned directly with the impedance tracking errors. It guarantees that the desired impedance is achieved asymptotically and both the position errors and force errors are bounded. Simulation results are provided to verify the effectiveness of the scheme.