DGT-Map: Directional Global Traversability Mapping Utilizing Multi-Task Learning for Heterogeneous Vehicles

2026-09-24

AI Summary

This study addresses the challenge that off-road traversability is inherently direction-dependent and vehicle-specific, whereas existing global maps typically model isotropic costs and are trained independently, hindering cross-platform terrain representation sharing. To overcome this, we propose a self-supervised multi-task learning framework that fuses RGB-D observations with motion signals to generate directional, vehicle-conditioned cost maps. By employing a shared backbone network coupled with vehicle-specific prediction heads, the architecture effectively decouples universal terrain feature transfer from personalized behavior modeling, and is further integrated into a Hybrid A* navigation stack. Evaluated across complex simulated terrains, the proposed approach achieves optimal or co-optimal navigation success rates compared to geometric, binary, and non-directional baselines.

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.

Problem

• off-road traversability
• directional cost mapping
• vehicle-specific navigation
• multi-task learning
• heterogeneous vehicles

Innovation

• Directional Traversability Mapping
• Multi-Task Learning
• Self-Supervised Learning
• Vehicle-Conditioned Costmap
• Off-Road Navigation