odometry

Odometry is how a robot figures out where it has moved by watching its own motion, step by step, and adding up all those tiny steps. It does not get told its position from outside; instead it asks, over and over, "how much did I just turn and how far did I just roll?" and keeps a running total. The word comes from the old practice of measuring a journey by counting the turns of a wheel — like a car's odometer ticking up the miles, except a robot tracks not just distance but the changing direction too, so it can trace out the whole winding path it took.

The way it works is to take small, frequent measurements of incremental motion — a wheel encoder noticing the wheels turned a quarter rotation, a camera noticing the scene slid a little to the left — and stitch them into a guess about the robot's pose, meaning its position and which way it is facing. Each measurement on its own is almost trivial, but chained together they reconstruct the journey. The catch is that every tiny step carries a tiny error, and because odometry adds step onto step, those errors pile up: the estimate slowly slides away from the truth the longer the robot travels, which is why odometry is usually corrected by an outside reference like GPS or a recognized landmark before the drift grows too large.

Odometry matters because it is the cheapest, most always-available sense of self-motion a robot has — it needs no satellites, no map, no external beacons, just attention to its own moving parts. It is the dependable backbone that fills in the gaps between rarer, more trustworthy fixes, giving the robot a smooth, continuous sense of where it is even in places where nothing outside can tell it.