Dynamic Event-Triggered Composite Learning Robot Control With Exponential Tracking

Abstract

Composite learning can guarantee the exponential stability of adaptive robot control under a condition termed interval excitation (IE) that is strictly weaker than the classical condition of persistent excitation. However, existing composite learning robot control (CLRC) must be updated in real-time even if tracking and estimation errors have already converged, which wastes computational resources. This article proposes a dynamic event-triggered CLRC (ET-CLRC) strategy, in which a dynamic ET condition with adaptive capability is designed to flexibly adjust the update frequency of CLRC for effectively utilizing available computational resources. The proposed ET-CLRC is distinguished from existing ET adaptive robot control schemes in two aspects: 1) The designed event is comprised of tracking errors and internal variables without involving the real-time updated information of control and adaptation laws, so the control and adaptation laws are updated only when the change rate of the tracking error is relatively high; 2) the practical exponential convergence of tracking and parameter estimation errors under the IE condition are guaranteed without the Zeno behavior. Simulations and experiments on an industrial robot named Franka Emika Panda with seven degrees of freedom have validated that the proposed ET-CLRC per-forms similarly to the real-time updated CLRC while significantly alleviating the computational burden.