Nonprehensile manipulation moves objects without grasping them — pushing, sliding, toppling, or throwing — extending what a robot can do beyond what its gripper can hold, and letting it handle objects too big, heavy, or awkward to grip.
Nonprehensile manipulation is moving something without picking it up — pushing a box across the floor, sliding a card off a table, nudging a part into place. The robot uses contact, not a grip.
Not everything needs to be picked up. Slide a heavy box across the floor, nudge a part into a corner, tip a bottle upright, flick a card off a table — all are ways to move objects without grasping them. That's nonprehensile manipulation.
What it is
"Prehensile" means grasping; nonprehensile manipulation controls an object's motion through contact alone — pushing, sliding, pivoting, toppling, rolling, even throwing — without a grip that fully constrains it. The object is free to move, and the robot shapes that motion through carefully applied contact forces and friction.
Move it without holding it
Instead of enclosing the object in a grasp, the robot uses controlled contact and friction to drive it where it needs to go.
Why a robot would choose it
Objects too big/heavy to grip. Push a crate that no gripper could lift.
No good grasp exists. A flat object flush on a table, or one wedged in clutter, may be un-graspable but pushable into the open first.
Speed and efficiency. Sweeping several objects, or nudging a part the last centimeter, can beat a full pick-place-repeat cycle.
Extending the workspace. Toppling a tall object or throwing it can place it beyond the arm's normal reach.
Often nonprehensile moves enable a grasp — pushing an object away from a wall so the gripper can get around it.
Why it's hard
It hinges on friction and contact mechanics, which are notoriously hard to model precisely — the same push can slide or tip an object depending on subtle surface conditions and where the force is applied relative to the center of mass. The object is underactuated (fewer controls than its degrees of freedom), so guiding it precisely is a genuine control challenge. This makes it a natural fit for reinforcement learning and contact-rich planning.
Where you'll see it
Pushing-based rearrangement in clutter, "pre-grasp" manipulation to set up a pick, sweeping and sorting, and research on throwing (a robot tossing objects into bins faster than it could place them). It's a growing part of general dexterous manipulation.
Why it matters
Nonprehensile manipulation expands a robot's repertoire beyond the limits of its gripper — letting it handle objects it can't hold and act more like a resourceful human who pushes, slides, and nudges. It's key to robust manipulation in messy, real environments.
