Emergence of coordinated gait patterns via physically coupled active rotators.

Abstract

In contrast to the central pattern generator hypothesis, which posits that neural networks generate rhythmic motor patterns without sensory feedback, recent robotics studies have demonstrated that independent oscillating agents with load-dependent feedback can organize coordinated gaits in quadrupedal robots. In this study, we develop minimal mathematical models to describe how such coordination emerges from physical interactions through the trunk and environment. By employing active rotators as limb controllers, we demonstrate their capacity to generate distinct gait patterns, including the trot, pace, and bound. We can also predict gait transitions with walking speed. These models provide insight into why different animals with specific physiques have limited gait patterns and offer suggestions for designing quadrupedal robots.