Skip to main content

📝 Map-Based Localization and High-Definition (HD) Maps

Description​

< What is it? >​

Map-based localization estimates a vehicle's pose by matching live sensor observations to a prebuilt map. High-definition (HD) maps provide the detailed, structured road information that makes this possible, often at much higher precision than consumer navigation maps.

  • Example: a vehicle aligns its current LiDAR point cloud with a stored 3D point-cloud map while using GNSS and IMU data as an initial guess. The resulting pose helps it identify its lane and position along the route.

Key points​

< What an HD map contains >​

Map layerTypical contentMain use
Geometric map3D point clouds, curbs, poles, road boundariesLiDAR or camera localization
Lane-level vector mapLane centerlines, boundaries, intersections, stop linesRouting, planning, and lane association
Semantic landmarksSigns, traffic-light locations, crosswalks, and road rulesLocalization and driving behavior

The map usually stores stable, surveyed structure. Live perception still needs to handle temporary objects, changing traffic lights, pedestrians, and road work.

< How localization works >​

prebuilt HD map + GNSS/IMU prior + live LiDAR or camera observations
                              ↓
                     scan, feature, or landmark matching
                              ↓
                  vehicle pose and uncertainty estimate
                              ↓
                        planning and control

The prior narrows the search area, while the map match corrects drift. Common approaches include LiDAR scan matching against a point-cloud map and visual landmark matching against a camera or semantic map.

< Mapping and map maintenance >​

SLAM, surveying, and repeated fleet data collection can create or refine the map. A map is not permanent: construction, new lane markings, road closures, and changed landmarks require quality checks and updates.

< Benefits and limitations >​

  • Benefit: map matching can provide precise, repeatable localization when GNSS is degraded.
  • Dependency: it requires a compatible, accurate, and current map for the operating area.
  • Safety concern: stale or mismatched maps can cause a localization failure, so systems monitor match quality and retain independent sensor evidence.

Comparison​

< Map-based localization vs GNSS and SLAM >​

Map-based localizationGNSS-based positioningSLAM
Main source of global correctionPrebuilt map featuresSatellite signalsMap constraints and revisited places
Map requiredYesNoBuilt or updated during operation
Strong situationMapped roads with reliable visual or LiDAR featuresOpen sky with good satellite visibilityMapping fleets or unknown environments
Main limitationMap freshness and sensor-map mismatchTunnels, urban canyons, and multipathDrift, compute, and long-term map consistency

Reference​