Directional traversability maps improve off road vehicle navigation
DGT-Map: Directional Global Traversability Mapping Utilizing Multi-Task Learning for Heterogeneous Vehicles
Robotics
Summary
Off-road vehicles can travel differently depending on the direction and type of vehicle, but most maps treat all directions the same. The authors present DGT-MAP, a system that learns how easy or hard it is for different vehicles to cross terrain depending on direction, using cameras and movement data. It trains on multiple types of vehicles to share common terrain understanding while keeping vehicle-specific details. When tested in simulation, it helps vehicles plan routes more successfully on tricky landscapes.
What this means in practice
- •For autonomous vehicle developers: Generate direction-aware traversability maps that help autonomous vehicles pick better routes on uneven terrain with vehicle-specific details.
- •For robotics simulation engineers: Integrate directionally conditioned traversability costs for various robot types to improve simulation-based navigation testing.
Authors
Jaskrit Singh, Kashif K. Noori, Jing Xiao, Constantinos Chamzas
Abstract
Off-road traversability is direction-dependent and vehicle specific, yet most global maps assign a single isotropic cost to each location. Existing learned estimators are also commonly trained independently for each vehicle; this preserves vehicle-specific behavior but prevents vehicles from sharing common terrain representations. DGT-MAP addresses both limitations through a self-supervised framework that learns global, directional, and vehicle-conditioned traversability costmaps from RGB-D observations and locomotion signals. A shared multi-task backbone learns common terrain features across training vehicles while vehicle-specific prediction heads preserve platform-dependent responses. At inference, DGT-MAP produces a heading-indexed costmap that can be used by a direction-aware planner. We evaluate DGT-MAP in simulation by integrating it into a Hybrid A* navigation stack and measuring downstream task success on challenging terrains, including slopes that are traversable downhill but not uphill and a ridge obstacle that is traversable by some vehicles, but not by others. Across evaluated tasks, DGT-MAP achieves the highest or tied-highest navigation success rate when compared against geometric, binary, and learned direction-agnostic baselines.