A robust position/force learning controller of manipulators via nonlinear H∞ control and neural networks

Abstract

A new robust learning controller for simultaneous position and force control of uncertain constrained manipulators is presented. Using models of the manipulator dynamics and environmental constraint, a task-space reduced-order position dynamics and an algebraic description for the interacting force between the manipulator and its environment are constructed. Based on this treatment, the robust nonlinear Hinfinity control approach and direct adaptive neural network (NN) technique are then integrated together. The role of NN devices is to adaptively learn those manipulators' structured/unstructured uncertain dynamics as well as the uncertainties with environmental modelling. Then, the effects on tracking performance attributable to the approximation errors of NN devices are attenuated to a prescribed level by the embedded nonlinear Hinfinity control. Whenever the adopted NN devices have the potential to effectively approximate those nonlinear mappings which are to be learned, then this new control scheme can be ultimately less conservative than its counterpart Hinfinity only position/force tracking control scheme. This is shown analytically in the form of theorem. Finally, a simulation study for a constrained two-link planar manipulator is given. Simulation results indicate that the proposed adaptive Hinfinity NN position/force tracking controller performs better in both force and position tracking tasks than its counterpart Hinfinity only position/force tracking control scheme.