Embodied AI Glossary中文

Slip

打滑Common

Relative sliding between contact surfaces, usually because the tangential force needed exceeds the friction limit.

Slip means two touching objects have relative motion at their contact point; the opposite state, where the contact point stays relatively still, is called sticking. Under the common Coulomb friction model, the tangential friction force f_t must satisfy f_t ≤ μf_n (μ is the friction coefficient, f_n is the normal force); the moment the tangential force actually needed exceeds this limit, the contact switches from sticking to sliding, at which point friction equals exactly μf_n and points opposite the direction of motion — geometrically, the contact force has left the friction cone. In manipulation, a grasp that slips can drop or spin an object out of position, so the fix is either more grip force or slip detection with a tactile sensor; in legged robots, a support foot that slips is a common cause of falling, so reinforcement-learning training often penalizes horizontal velocity at the contact foot. Deliberately controlled sliding is also a skill in its own right — pushing an object across a table, or letting it rotate between the fingers.

ExampleA two-finger gripper holds a 0.5 kg cup vertically, with its roughly 4.9 N of weight split between the two contact points' friction. If μ = 0.5, each finger needs at least about 4.9 N of normal force (2×0.5×4.9≈4.9 N), or the cup slides downward. Humanoid-Gym's foot_slip reward term penalizes a foot's horizontal velocity whenever it's in contact with the ground.

Also called
Sliding, Slippage, Foot Slip
Related
Coulomb Friction · Friction Cone · Friction Coefficient (Coefficient of Friction) · Slip Detection · Force Closure · Stance Phase / Swing Phase
Sources
Modern Robotics (Lynch & Park), Ch.12.2 Contact Forces and Friction
Humanoid-Gym: humanoid_env.py (_reward_foot_slip)

See it in the full glossary →