Skip to main content

📝 Localization & Mapping

Description​

< What is it? >​

Localization and mapping let a robot answer two foundational questions: “Where am I?” and “What stable structure is around me?” In autonomous driving, the result is a reliable vehicle pose and a map representation that perception, planning, and control can use.

The area includes Simultaneous Localization and Mapping (SLAM), Sensor Fusion, and Map-Based Localization and High-Definition (HD) Maps.

Key points​

< Core concepts >​

ConceptPurpose in autonomous driving
LocalizationEstimate the vehicle's position, orientation, velocity, and uncertainty
MappingRepresent stable geometry and road semantics, such as lanes, curbs, signs, and landmarks
SLAMEstimate pose while building or updating a map
Sensor fusionCombine complementary, imperfect sensor signals into a more reliable estimate
Map-based localizationMatch live observations to a prebuilt map to refine the vehicle pose

< How the pieces fit together >​

LiDAR, cameras, GNSS, IMU, wheel odometry
                 ↓
          sensor fusion and localization
                 ↓
        vehicle pose with uncertainty
                 ↓
  map matching → perception → planning → control

Mapping fleets may use SLAM to make or refresh maps. During everyday driving on known roads, a vehicle more commonly fuses sensors and localizes itself against a prebuilt map.

< Why uncertainty matters >​

Every sensor can be wrong: GNSS can fail in tunnels, cameras can be affected by glare, and LiDAR can be degraded by weather or occlusion. A robust system tracks confidence as well as its best pose estimate, detects inconsistent measurements, and chooses a safe fallback when localization is unreliable.

Reference​