Embodied AI Glossary中文

Backdrivability

反驱性Common

Whether an external force applied at a joint's output can easily push the whole drivetrain and motor backward.

Backdrivability describes whether a force applied at a joint's output end can easily push back through the drivetrain and turn the motor in reverse. A joint with good backdrivability yields when a person pushes or bumps it, and the motor's current reflects the size of the external force, which is useful for force control, collision detection, and kinesthetic (hand-guided) teaching. A joint with poor backdrivability — for example, one with a large gear ratio, a worm gear, or a self-locking screw — can't be pushed back at all; it feels “stiffer” and is more prone to damage from impacts. Backdrivability is mainly determined by gear ratio, transmission efficiency, and friction: a larger gear ratio means more reflected inertia and friction at the output end, making the joint harder to backdrive. The quasi-direct-drive joints common in quadruped and humanoid robots deliberately use a smaller gear ratio to get good backdrivability.

ExampleWhen a quasi-direct-drive leg lands and takes an impact, the ground can push it back slightly, cushioning the hit; a joint with a high gear ratio instead transmits the impact straight into the gears.

Related
Quasi-Direct Drive · Reflected Inertia · Gear Ratio · Proprioceptive Actuator · Kinesthetic Teaching · Compliance
Sources
Wikipedia: Backdrivability

See it in the full glossary →