What is SLAM?
SLAM solves a chicken-and-egg problem: to know where you are you need a map, and to build a map you need to know where you are. A robot doing SLAM does both at once, incrementally constructing a map of its surroundings from sensor data while continuously estimating its own pose within that partial map.
It is what lets a device move confidently through a space it has never seen before, without external positioning like GPS.
Key takeaways
- SLAM builds a map and localizes within it at the same time.
- It works in unknown environments without external positioning.
- It fuses sensors like cameras, LiDAR, and IMUs over time.
How it works
As the device moves, it extracts features or geometry from sensors such as cameras or LiDAR, matches them against what it has already mapped, and updates both the map and its estimated pose. Odometry and inertial measurements provide short-term motion estimates, while loop closure, recognizing a previously visited place, corrects accumulated drift. The result is a consistent map and trajectory.
Why it matters
SLAM is the backbone of autonomous navigation for robots, drones, and vehicles operating where GPS is unavailable or imprecise. Because it depends on fusing multiple noisy sensor streams over time, it sits at the intersection of perception and the multimodal sensor data that physical AI systems rely on.
Frequently asked questions
Why is SLAM called simultaneous?
Because the map and the device's location are estimated together, each informing the other, rather than one being known in advance.
What sensors does SLAM use?
Commonly cameras, LiDAR, and inertial measurement units, often fused together, with the exact mix depending on the platform and environment.
Related terms