Learning Traversability-Aware Global Planners for Long Horizon Off-Road Navigation

2026-07-26

AI Summary

This study addresses the challenge of safe and efficient long-horizon path planning in large-scale off-road environments, where onboard sensors suffer from limited perception range. The authors propose a learning framework that leverages multi-source geospatial data—including satellite imagery, airborne LiDAR, and vector maps—to estimate traversability directly from sparse human driving trajectories as supervision. By integrating self-supervised geometric priors, the method constructs a continuous global traversability map, overcoming the limitations of conventional approaches that rely solely on local sensing. In real-world evaluations, the system generates paths reaching 96.34% of the length of human-driven routes and reduces human intervention by approximately 85% compared to purely local planning baselines.

Abstract

Autonomous navigation across large off-road environments remains a challenging problem. Onboard sensors perceive only the immediate surroundings, yet safe and efficient routes depend on terrain features that extend well beyond the sensor horizon. Geo-spatial data sources such as satellite imagery, aerial LiDAR, and vector maps can close this gap, but learning traversability from them is difficult: dense labels are unavailable at scale, and existing methods rely on short-range sensing. We propose an efficient formulation that learns a continuous traversability map from overhead data, supervised directly by human-driven GPS trajectories and shaped by self-supervised geometric priors from LiDAR. Alongside the model, we release a public dataset of 299 scenes spanning $\sim\!1{,}244\,\mathrm{km}^{2}$ of diverse terrain, paired with $1{,}130\,\mathrm{km}$ of human driving. In field trials on a Clearpath Warthog across seven routes at two sites, our method achieves trajectories within $3.66\%$ of human path length and reduces operator interventions by $\sim\!85\%$ compared to local-planner-only autonomy.

Problem

• off-road navigation
• traversability learning
• long horizon planning
• geospatial data
• autonomous navigation

Innovation

• traversability learning
• off-road navigation
• overhead data
• self-supervised geometric priors
• GPS trajectory supervision