New Method Uses Coastlines to Navigate GPS-Denied Boats

A team led by researcher Derek Benham has published a preprint on arXiv describing two complementary approaches for localizing autonomous surface vessels (ASVs) in GPS-denied environments by exploiting the geometry of coastlines and water surfaces.

The first framework uses LiDAR to observe the water surface directly, estimating the vessel's roll, pitch, and heave (vertical motion). It then recovers global position and heading by registering shoreline observations against a satellite-derived coastline map.

The second framework is purely vision-based. It detects the shoreline and horizon through semantic segmentation of monocular camera images, then uses the resulting coastal scene geometry to infer distance to shore. These shoreline observations are accumulated into short-duration local submaps, matched against the same satellite-derived coastline map, and combined within a hierarchical factor graph.

According to the paper, as reported by arXiv, both pipelines were tested across three real-world coastal datasets. The LiDAR-based method consistently improved trajectory accuracy compared to standard baselines, while the camera-only architecture kept long-term drift bounded despite relying on a single sensor. The authors also report that modern zero-shot foundation models proved reliable at extracting shoreline features across a variety of coastal environments, without needing environment-specific training.

The findings suggest coastlines could serve as a dependable, globally referenced landmark for maritime robots operating where GPS signals are jammed, spoofed, or simply unavailable — a persistent challenge for autonomous boats used in surveying, patrol, and environmental monitoring. The 22-page paper, which includes 13 figures and 7 tables, was posted to arXiv's Robotics and Computer Vision categories on August 21, 2026.