Receding-horizon control

You plan several steps, but the environment changes faster than you can execute the whole plan. A robot sees that a pedestrian suddenly turned. Completing the old sequence could be dangerous.

Receding-horizon control creates a plan for several future steps, executes its beginning, measures again, and forms a new plan. The horizon, the future range covered by the plan, shifts as action proceeds.

In the example, the plan has five commands, but the robot executes one. A new observation lets it replace the remaining commands with a fresher sequence. The PDPO paper, §3, describes this organization.

We combine a coherent short-term intent with a response to new information. The cost is repeated planning and dependence on measurements. The mechanism itself does not guarantee collision avoidance: incorrect predictions or a poorly specified objective can still produce a bad plan.