Let the physics do the work: the idea behind graceful legged robots
October 5, 2026 · 12 min read
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.
The rules, numbers and steps worth keeping in front of you.
- Core idea: add a little energy and intent to motion that gravity, springs and momentum already supply.
- Tedrake: mechanics is what happens with the motors off; control is commands that add new energy.
- Tedrake: passive dynamic walkers go down a ramp with no controllers and no batteries.
- Tedrake: conventional control erases the dynamics of the world; go with the flow instead.
- Playter: let the robot tip, then catch it with the next foot.
- Playter: physics solves part of the problem, so control gets easier and more efficient.
- Raibert's three-part hopper rule: regulate bounce height, place the foot, keep the body upright.
- Raibert: start with the simplest machine, a pogo stick, before a gallop.
- Raibert: natural movement also needs good hardware and predictive control.
- Limit (Tedrake): a passive walker only goes downhill; flat ground needs added energy.
- Limit (Playter): a humanoid's heavy upper body and legs make balance much harder.
- Limit (Tedrake): making and breaking contact is discontinuous and hard to plan for.
- Playter and Raibert: expect to break hardware; keep spare parts and a repair crew.
- Tedrake: learn the fundamentals rigorously before reaching for learning methods.
This is the short form. The full insight has every expert's reasoning, their own words and the sources.