Robust Dynamic Locomotion Through Feedforward-Preflex Interaction

摘要

Abstract Unlike most legged robotic systems built to date, even simple animals have the ability to quickly and robustly traverse through rough terrain and over large obstacles and gaps. Recent evidence from insect physiology research indicates that arthropods achieve this fast robust locomotion largely without relying on sensory feedback or reflex response. Instead, locomotion is the result of the interaction between a basic feedforward motor pattern and the visco-elastic properties of the mechanical system, termed “preflexes.” In this paper, we consider the implications of this control hypothesis for the design of small running robots for uncertain environments. We present working prototypes that show how robust dynamic locomotion can be achieved without the use of sensory feedback. We then discuss modeling approaches for these kinds of systems and present results from simulations of representative models.