The Possible Physics (Mechanics) of Walking

Professor Andy Ruina

Theoretical and Applied Mechanics and Mechanical Engineering Dept.

Cornell University

Robots have motors and people have muscles. What for? To guide motions and to make up for lost mechanical energy. How much guidance is fundamentally needed for repetitive tasks? How much energy needs to be supplied for what losses? One approach to understanding the need for motors and controllers is what can be done without them.

Tad McGeer demonstrated (1988-1993) with simple computational models and with physical devices that uncontrolled human-like walking motions can be achieved with, to put it simply, sticks and hinges that walk downhill. The motions of these toy-like devices are energetically efficient and stable. We have found that, in principle, some of these devices can walk on arbitrarily small slopes and thus approach perfect efficiency and that Robot configurations that have this efficiency are reminiscent of the human design. These models can also limp (period 2), waltz (period 3), and stumble (chaos). One of our devices has the unintuitive feature that it has no stable standing posture, yet can walk stably.

The basic theory is not novel: numerical search for limit cycles and numerical evaluation of their stability. However, two morals seem to be exposed: locomotion efficiency is based on avoidance of impacts, and stability comes from utilizing non-holonomic constraints.

Friday, November 17, 2000

3:30 - 5:00 p.m.

1500 EECS