Adaptive Admittance Control for Optimized Robot-Environment Interaction Without Restrictive Initial Conditions

This paper investigates an adaptive optimal admittance control scheme for robot manipulators interacting with unknown environment. To resolve the optimized interaction performance considering tracking error and interaction force, an impedance adaptation approach is developed without the initial stabilizing policy. Based on the gradient-based updating method, the online solution can exponentially converge to the optimal impedance gain without prior knowledge of environment dynamics. A nonlinear mapping method is integrated into the admittance control, transforming the constrained system into an equivalent system without state constraints. By eliminating feasibility conditions, the tracking controller can achieve the full-state asymmetric time-varying constraints under a broad range of initial conditions. Through the Lyapunov analysis, it is proven that the closed-loop signals are bounded. Finally, simulation and experiment results demonstrate the effectiveness of the proposed methods.