Neural Impedance Control of Uncertain Robotic Systems with Prescribed Performance

Abstract

This paper proposes a neural impedance control method to limit the uncertain robotic systems tracking error within a predetermined range, and improve the interaction safety and compliance. A position-based impedance controller is proposed to enhance the compliance of human-computer interaction. In the controller design process, the time-varying energy function is taken as the given error boundary, and the unsymmetrical potential barrier Lyapunov functions (BLF) are chosen to solve the constrained problem, so that it can finally make the robotic system have a good dynamical performance and steady-state performance. Besides, it is helpful to solve the uncertainty of robot system dynamics by combining with radial basis function neural network (RBFNN). The controller has better adaptive performance to improve the safety and compliance of the uncertain robotic systems when it is in physical contact with the rigid environment. According to the effect of the controller, the tracking error can not exceed the preset boundary and quickly converge. Finally, the effectiveness and practicability of the controller are verified by simulations.