Action chunk

A robot moving through a crowd needs more than a single “turn now” command. A short plan of several movements can better describe its intent: slow down, avoid the obstacle, and return to the route.

An action chunk is a sequence of future actions generated as a whole. Each position concerns the next time step. The number of actions in the plan need not equal the number performed before observing the surroundings again.

In the studied PDPO, §3, five commands of 0.25 s describe 1.25 s of motion. The robot executes only the first, however, reads the new state, and plans again. The remaining steps help give the current motion a coherent direction.

The plan can become outdated if a person suddenly changes course. Receding-horizon control describes shifting and refreshing it. An action chunk specifies the plan's form, rather than promising execution of every listed action.