Physiologically Inspired Robot Control: A Challenge to Bernstein’s Degrees-of-Freedom Problem

摘要

This paper explores what are the fundamentals of physiologically inspired robot control through investigating two interesting but difficult problems of robot control: 1) robotic hand-writing and 2) human-like multi-joints reaching with surplus DOFs (Degrees-of-Freedom). Noteworthy characteristics in control of these systems are a) redundancy in DOF, b) dexterity in execution of a target task, and c) indispensability of sensory motor coordination, though d) their dynamics are nonlinear and governed by geometric constraints in the former system. Due to these inherent characteristics, both the conventional control-theoretic approaches originated from the linear system theory and Lyapunov’s direct method are likely to fail. By gaining physical insights into these nonlinear dynamics with redundancy in DOFs and physical descriptions of target tasks, this article claims that a symbiosis of “robotics” and “neuro-physiology” is quite effective in finding a simpler sensory-motor coordination principle in each case that leads to accomplishment of each imposed task regardless of annoying characteristics of its motion dynamics with nonlinearity and DOF-redundancy.