On the Connectivity of Motion Spaces for Biologically-Inspired Legged Robots

This paper discusses recent analyses of the shape of motion spaces for legged robots in quasistatic equilibrium at a given set of contacts. Two robots were examined: HRP2, a biologically inspired humanoid designed by AIST Japan, and ATHLETE, a six-legged, non-biologically inspired robot designed by NASA. Probabilistic roadmap techniques are used to compute the approximate connectivity of several hundred motion spaces. Results suggest that motion spaces for HRP2 are connected, while those of ATHLETE are frequently disconnected. This signifies that when planning long-range locomotion, HRP-2 only needs to reason about its footsteps, while ATHLETE must reason about its pose as well. This makes planning much more difficult for ATHLETE, a claim that is supported by experience both in human teleoperation and in motion planning algorithms. If similar findings hold for other biological mechanisms, then it can be argued that evolution has selected for mechanisms that minimize the cognitive load needed for locomotion, which could form the foundation of a strong argument for biologically-inspired design.