Capture Point
A model-based estimate of where a robot can step to bring its motion to a balanced stop.
The Tiny Spot That Saves a Robot From Eating Floor
Ever tripped, taken two weird panic steps, and somehow stayed upright like a budget ninja?
Congrats. You’ve used the same basic idea as a humanoid robot’s capture point.
The simple version
A capture point is the spot a robot has to step to if it wants to stop moving and not fall over.
That’s it. It’s the robot’s “put your foot there or prepare to meet the tiles” location.
Think of it like this. If your body is tipping forward, you can’t just politely decide to stop. You have to plant your foot in the right place to catch yourself. Too short, you still fall. Too far, you do a weird lunge and look like you’re fighting an invisible goose.
Robots have the same problem. Except they do math instead of panic.
Why this is a thing
Walking is basically controlled falling with better PR.
Every step, a humanoid robot is a little off-balance on purpose. To stop safely, it needs to know:
- where its body is heading
- how fast it’s moving
- where to place a foot so everything settles down instead of face-planting
That target spot is the capture point.
If the robot can step there, it can “capture” its motion and come to rest. If it can’t reach that spot in time, or the spot is blocked, things get messy fast. Like carrying soup in socks fast.
Why people care
This matters because balance is the whole game for humanoids.
A robot that knows its capture point can recover from shoves, bad terrain, slippery floors, and its own occasional bad decisions. Which is pretty useful if you want robots in warehouses, homes, hospitals, or anywhere humans leave chaos lying around.
So, capture point sounds fancy. But really it means this. The exact place where one smart step turns “disaster” into “nailed it.”