Koditschek, D.E.: INTEGRATING MECHANICAL AND BIOLOGICAL HYPOTHESES FOR HIGH PERFORMANCE LOCOMOTION.

Recent years have witnessed a payoff to robotics researchers who have taken the effort to incorporate lessons from biology in their designs. High performance locomotion – legged robots that are both stable and maneuverable over highly varied terrain – now appear to be close at hand. The potential for reciprocal payback to biology arises from the formal mathematical definitions and analysis that such models afford. Within this framework, steady state locomotion may be cast in terms of the local stability properties of selected periodic orbits. An attractor is a collection of orbits against which arbitrary small perturbations result in small deviations that eventually settle back to it. This (local) notion of stability is particularly useful because it entails formally equivalent computationally effective conditions on experimental observables that apply to almost any model of biological interest. In contrast, developing empirically grounded and mathematically well founded paradigms for maneuverability, a non-steady phenomenon, remains an open problem. The (global) collection of non-steady deviant motions that settle down to an attractor comprise its basin of attraction. Attractors with large basins and fast transients have offered useful (hybrid) building blocks for maneuverability in robotics. The paradigm of hybrid global stability mechanisms may offer new directions of experimental inquiry and conceptual unity in biology if: i) computationally effective methods can be developed for their analysis and design in the high dimensional nonlinear settings relevant to animal locomotion; and ii) the hypotheses they prescribe can be tied to specific aspects of animal biomechanics and physiology. Supported by DARPA/ONR N00014-98-1-0747.