Why a featureless road is hard to place on a map
A self-driving car has to answer two different questions: what is around it, and where is it? Cameras, radar, and lidar—a sensor that measures distance with laser pulses—help detect nearby objects. To estimate its position, the car also compares what its sensors see with a stored map or with features it has observed before.
That comparison works best when a view is distinctive. A sign, a bend, a building, or a clear road edge can help distinguish one stretch from another. On a broad, straight road with few visible features, many locations may look much the same. The car can detect open pavement perfectly well and still have little evidence for exactly where on the route it is.
Small errors can become a location problem
The car can estimate movement from its wheels and other onboard sensors, a process called dead reckoning. But that estimate accumulates error: a slight mismatch in wheel rotation or direction can make the calculated position drift over time. Recognizable landmarks let the system check and correct that estimate. When landmarks are scarce, there are fewer opportunities to make that correction.
Satellite positioning can provide another clue, but it may not be precise or reliable enough on its own for every driving situation. The car therefore combines several imperfect sources rather than relying on one signal.
The result is a subtle tradeoff: a busy street may pose more immediate hazards, yet its signs, corners, and buildings can make location easier to verify. An empty stretch can be physically simple to drive while remaining harder for the car to place precisely.
