Reproduce “TravSUITE: Traversability via Self-Supervised, Uncertainty-Aware IRL and Terrain Estimation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Samuel Triest, Amirreza Shaban, David D. Fan, Wenshan Wang, Sebastian Scherer
TravSUITE combines a 3D voxel map with a learned bird’s-eye-view map to predict traversability and speed for off-road navigation. It trains map completion from later sensor observations and learns route costs from expert demonstrations. In an offline ablation, inpainting improved the Test 1 trajectory-distance score from 1.7744 to 1.3936; on a winter course, risk-aware planning had zero collisions but lower average speed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.