HD Map

An HD Map is a three-dimensional map that's precisely surveyed and built in advance, before a self-driving car ever travels a road. It captures details down to the centimeter—lane widths, traffic light positions, road boundaries—and the car matches this map against what it sees in real time to figure out exactly where it is and which lane it should be in.

Think of it like a student who has memorized the entire test syllabus. Within that memorized range, they answer quickly and accurately, but they're stuck when they hit an alley that isn't on the map or a road construction zone that suddenly changed. That's why self-driving companies have to send out survey vehicles to build an HD Map of a new city before they can operate there—and this process can slow down how fast a service expands.

Recently, a different approach has emerged: driving by judging situations on the fly, more like a human, without relying on such maps. Instead of memorizing a map, an AI trained through experience makes its own judgments even on roads it has never seen before. The advantage is that expansion into new cities is faster, since no survey process is needed.

Articles mention things like "Waabi uses neither lidar nor pre-surveyed precision HD Maps," using whether or not a company relies on HD Maps as a marker that distinguishes different technical approaches among self-driving companies. Not using an HD Map doesn't mean having no map at all—it means the vehicle drives using real-time perception rather than depending on a pre-built precision map.