pick-and-place
Pick-and-place is the most basic and most common job a robot arm does: take hold of an object in one spot and move it to a target spot, then release it there. Pick it up here, put it down there. Nearly every factory and warehouse robot is, at heart, doing pick-and-place all day long — loading parts onto a machine, packing items into a box, sorting parcels, stacking products onto a pallet.
Though it sounds trivial, a single cycle quietly chains together several skills. The robot must first find the object and figure out its position and orientation; then plan a grasp — where to put the gripper and how to come at it — and a collision-free path to reach it; then close the gripper and confirm it actually caught the object; then carry it over and set it down accurately at the destination, often in a specific pose; and finally let go and retreat, ready for the next one. Each link in that chain is its own small problem, and any one of them failing means a dropped or misplaced part.
Pick-and-place is the workhorse against which much of manipulation is measured, and the gap between easy and hard versions is huge. Picking identical rigid blocks from fixed, known positions and dropping them into fixed slots is straightforward and was automated long ago. Picking jumbled, unfamiliar, or fragile items out of a cluttered bin — figuring out a good grasp on the fly, every object different — is far harder, and is exactly the cutting edge that modern perception and learning are pushing on.
On a packaging line, an arm picks a chocolate from a moving belt, swings over to the open box, and sets it gently into an empty pocket — then repeats sixty times a minute.
One pick, one move, one place — repeated endlessly.
The hard version of pick-and-place from a messy, mixed container is so common it has its own name: bin picking.