Reproduce “WayTrace: A Map-to-Language-to-Trace Hierarchy for Embodied Urban Navigation” 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
Manthan Patel, Jonas Frey, Pascal Roth, Marco Hutter, Cesar Cadena
WayTrace turns web-map directions into motion by using a vision-language model to select route commands, a second model to predict camera-image paths, and a controller to follow them. In simulated Zurich routes, it completed 63.3% of runs with 7 m radial-RMS GNSS error, versus 23.3% for a point-goal baseline using the same controller. This benchmark does not establish traffic safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.