A New Augmented L₁ Adaptive Control for Wheel-Legged Robots: Design and Experiments

Abstract

This paper proposes the augmentation of an 1 adaptive controller with a feedback Linear Quadratic Regulator (LQR) to control a wheel-legged biped robot. The performance of linearized model-based controllers, such as LQR, depends on the accurate knowledge of model parameters, a priori information about input and output disturbances, and other unforeseen conditions. We propose a hybrid scheme where an ${\mathcal{L}_1}$ adaptive controller is combined with LQR to compensate for matched uncertainties and other disturbances related to the environment change such as friction conditions of the floor. The proposed control scheme is able to keep the robot stable under model uncertainties and external disturbances through a series of validation scenarios including simulations and real-time experiments.