Dynamic Stability

Balance where the center of mass can briefly leave the support region, recovered through continued motion and the next footstep.

Legged robots can be “stable” in two different senses. Static stability requires the center of mass's ground projection to stay inside the support polygon (the convex region enclosed by the feet on the ground) at all times, so the robot wouldn't fall even if it froze at any instant — a hexapod that shuffles slowly on three legs at a time, alternating, walks this way. Dynamic stability relaxes that requirement: the center of mass can run outside the support region briefly, as long as the next footstep and adjustments to the ground forces keep the motion from diverging into a fall — this is how humans walk and run. Marc Raibert's hopping one-legged robots in the 1980s proved that a robot could stay upright through continuous hopping alone. Common ways to assess dynamic stability include the zero moment point (ZMP — the point where the foot's reaction force produces no horizontal moment, which must stay inside the support region) and the capture point (a footstep location that would bring the robot to a stop); modern humanoid locomotion controllers trained with reinforcement learning often learn dynamic balance directly in simulation instead.

ExampleWhen a humanoid robot walks briskly, during single-support phase its center of mass has already tipped forward past the supporting foot; the next foot lands just in time to catch the body — that's dynamic stability. If it instead shifted its center of mass directly above the support foot before every step, that would be a static-stability gait, and much slower.

Also called

Dynamic Balance

Related

Static Stability · Support Polygon · Zero Moment Point · Capture Point · Inverted Pendulum Model (IPM) · Balance Control

Sources

Legged robot - Wikipedia
Zero moment point - Wikipedia
Underactuated Robotics (MIT) - Humanoid Robots