Adaptive locomotion to periodic perturbation. Adaptation mechanism with coupling of oscillator and link dynamics

Abstract

Quadrupeds can acquire new gait patterns with respect to environmental changes. Yanagihara et al. have demonstrated this adaptability by experiments on a decerebrate cat. These experiments indicate that quadrupeds gradually adapt to their environment by repeating locomotion in a steady environment, and that the acquired gait pattern is persistently memorized after the locomotion. Our research aims at formulating a mathematical model of these cats' behavior and constructing a quadrupedal walking robot to realize such adaptive behavior. To date, we have proposed a mathematical description of adaptation at the level of gait pattern generation using neural oscillators. In this paper, we extend it to take into account limb dynamics. We study how to design the interaction of the oscillator and limb dynamics.