Do Intermediate Gaits Matter When Rapidly Accelerating?

Abstract

Transient locomotion is still poorly understood in terms of planning and implementation on robotic platforms, with most research concentrated on steady-state motion. In this letter, we investigate optimal rapid acceleration (positive and negative) maneuvers of a planar numerical quadruped and biped robot. The question we ask is whether legged robots should transition through discrete, intermediate gaits (walking to trot to bound) or plan a direct transition to the top-speed gait. We present numerical evidence supporting the energetic optimality of transitioning to a desired gait without intermediate gait transitions. Trajectories were generated from rest to steady state and vice versa. Two cost functions (cost of transport and a heat-based cost function) were analyzed and compared to observations made in nature. A full 30-m trajectory was generated and compared to the acceleration and deceleration results, which further supported transitioning directly to the desired gait. All the trajectories were observed to follow a sliding mass template model which, in future, can be used as a heuristic to plan these transient maneuvers.