Insight | Let the physics do the work: the idea behind graceful legged robots
Russ Tedrake, Robert Playter and Marc Raibert all trace agile robots back to the same principle from the MIT Leg Lab era, which is to work with a body's natural dynamics instead of fighting them. It matters because it explains why balance and movement, not raw computing power, were the first breakthroughs in legged machines.
Graceful legged robots came from a change of attitude toward physics more than from faster computers. The move is to stop using motors to cancel what a body naturally does, and instead add a little energy and intent to motion that gravity, springs and momentum already supply. Three roboticists describe it.
Russ Tedrake, an MIT professor and vice president of robotics research at the Toyota Research Institute, points to passive dynamic walkers: machines that walk down a small ramp after a push, with no controllers and no batteries. He explains them with a bicycle wheel that has lost its rim. Each spoke that hits the ground loses a little energy and each step down the slope gains a little, and the two balance into a steady motion that corrects itself. Conventional control, as he describes it, pushes back against the world and imposes motion chosen because the mathematics is simple. His alternative: "You can just let physics do most of the work." His favorite proof is a dead fish tied behind a cylinder in a water tunnel that starts to swim upstream.
Robert Playter, chief executive of Boston Dynamics, recognized the idea as a former gymnast: a skilled athlete puts the body where momentum will carry it. Early walking machines tried never to tip over. The insight of dynamic stability is the opposite: "let the tipping happen. Let yourself fall, but then catch yourself with that next foot." He says the result looks natural to people and is easier to control, because the physics solves part of the problem.
Marc Raibert, who founded Boston Dynamics, rejected a slow six-legged robot that always kept three feet on the ground. He built a pogo-stick hopper that worked around 1982 with a three-part rule: regulate the bounce by adding or removing energy, decide where to place the foot, and keep the body upright. He also credits good hardware, and control that predicts instead of looking backward.
They agree on the principle and differ in emphasis. Tedrake stresses how little energy is needed and counts MIT's Atlas work among the inefficient approaches, while Raibert still loves powerful hydraulics and bold movement. Playter says he saw the natural walking he expected from Atlas about 2022; Raibert says its walking is still not quite as graceful as a human's. Raibert notes the most impressive performances still come from traditional control, and Tedrake warns that learning methods make it easy to get things working without thinking critically.
The idea has limits. A passive walker only goes downhill, so flat ground needs added energy. Heavy humanoid bodies are much harder to balance, contact with the world is not handled by the principle, and later gains leaned on prediction and faster computers.
What to do with it: ask what a system does with the motors off. Start with the simplest machine that contains the core problem. Design the controller to add energy and intent, not to cancel natural motion. For balance, let the body tip and solve for where the next foot goes. Break control into a few plain rules, predict a second or two ahead, expect to break hardware, and test more than feels necessary.
Tedrake applies the idea to barefoot running; anyone changing how they run should get advice from a qualified professional first.
