Acceleration-Free Recursive Composite Learning Control of High-DoF Robot Manipulators

Abstract

Composite learning robot control (CLRC) is an adaptive control approach that achieves exponential parameter convergence without using a stringent condition termed persistent excitation (PE). For robots with low degrees of freedom (DoFs), a filtered regressor of the robot dynamics needed in CLRC can be calculated analytically without joint accelerations, but this is difficult for high-DoF robots. Under the linear parameterization by the recursive Newton-Euler algorithm, this paper proposes an acceleration-free recursive CLRC (RCLRC) method for high-DoF robots to achieve exponential parameter convergence under a weakened condition termed interval excitation (IE). The proposed method has a low computational cost and avoids undesirable acceleration estimation that seriously affects performance. Simulations and experiments on a 7-DoF robot manipulator have verified the superiority of the proposed RCLRC, where it outperforms its analytical version in both estimation and tracking.