Reproduce “Simulated Imagination to Real-World Decisions: World Models for Robot 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
Long Quang, Akash Guna, Athul Radhakrishnan, Deepanshu Mody, Bradley Woosley, Kiana Bronder, Benjamin Dossett, Carlos Nieto and 1 more
The authors integrate a simulation-trained world model into a navigation stack, where it predicts latent futures from images and odometry and evaluates paths toward GPS waypoints. In a controlled site mirroring the simulator, its reference-trajectory error was lower than A*’s (8.45 versus 11.79 meters); in the field, it was not (16.55 versus 16.32 meters), and it produced paths less reliably.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.