terrainrelative

Terrain-relative navigation (TRN) is a navigation approach that determines a vehicle’s position and orientation by aligning real-time sensor observations of the terrain with a pre-existing digital terrain model or map. By anchoring navigation to terrain features rather than relying solely on inertial sensors or external signals, TRN enables operations in GPS-denied environments and can improve landing accuracy on unknown terrain.

TRN systems typically fuse data from sensors such as stereo or monocular cameras, LiDAR, radar, or other imaging modalities with a high-resolution digital elevation model or terrain map. The observed terrain is matched to the map through methods such as image correlation, feature-based matching, optical flow, or probabilistic filtering. The outcome is a real-time pose estimate relative to the terrain, which can be updated during critical phases such as descent or autonomous flight.

Key considerations include the quality and currency of the terrain map, resolution and coverage, and the ability to handle terrain changes or occlusions. TRN often requires preprocessing to generate or update maps and may incorporate safeguards for map drift, sensor occlusion, and computational load. It is designed to operate alongside other navigation ways to provide robust positioning under challenging conditions.

Applications of TRN span space exploration, where vehicles must land safely on unknown planetary surfaces, to autonomous aerial and ground vehicles operating in GNSS-denied or feature-rich environments. Benefits include reduced navigation drift, improved hazard assessment, and enhanced precision during landing or close-approach maneuvers. Limitations involve potential ambiguity in flat or repetitive terrain, reliance on map accuracy, and processing requirements in real time.