Adaptive Cartesian space control of robotic manipulators: A concurrent learning based approach

Abstract

This work introduces a concurrent learning-based adaptive control design for end-effector tracking and the corresponding stability analysis for robotic manipulators. The presented controller is developed directly in Cartesian space, thereby removing the necessity for inverse kinematics calculations at the position level. The designed adaptive controller ensures global exponential tracking of the end-effector in Cartesian space. Moreover, the developed controller assures globally exponential convergence of uncertain dynamical parameters to their actual values without demanding persistence excitation conditions via a combination of a standard gradient-based adaptation with a novel integral concurrent learning component. The developed integral concurrent learning part operates both real-time output data and the most informative historical data gathered by employing the singular-value maximization algorithm (SVMA) to reduce the size of memory allocation. Lyapunov-based arguments are applied to ensure the exponential stability of the closed-loop system. Numerical studies are performed to depict the feasibility and performance of the proposed design.