global planner vs local planner

Getting a robot from where it is to where it wants to be is usually split into two jobs done by two planners. The global planner is the big-picture route-finder: given a map of the whole area, it works out a complete path from start to goal before the robot even moves — the equivalent of opening a map app and seeing the full highlighted route across the city. The local planner is the moment-to-moment driver: as the robot actually rolls along that route, it watches the small patch of world right in front of it and steers second by second to follow the plan while dodging whatever turns up. The two run as a team — one decides the overall direction, the other handles the immediate steering.

They are split this way because each job has a different shape. The global planner can afford to think slowly and broadly: it considers the entire map, so it will not send the robot down a corridor that turns out to be a dead end — but it usually plans against a map drawn earlier and cannot know about a box someone just dropped in the hallway. The local planner is the opposite — it thinks only a few seconds and a few meters ahead, but it reacts fast to live sensor readings, swerving around a person who steps into the path or a chair that was not on the map. Coarse and far-sighted versus reactive and near-sighted: that division of labor is the whole point.

A handy way to picture it is a road trip. The global planner is the route you chose before leaving — the sequence of highways that gets you across the country. The local planner is you behind the wheel: staying in your lane, braking for the car ahead, easing around the pothole, all while keeping the big route in mind. If a road is unexpectedly closed, the local planner cannot fix that on its own; it signals back, and the global planner re-plans a fresh route around the blockage.