Reproduce “P$^{3}$Nav: End-to-End Perception, Prediction and Planning for Vision-and-Language 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

Tianfu Li, Wenbo Chen, Haoxuan Xu, Xinhu Zheng, Haoang Li

P³Nav integrates object and map perception, candidate waypoint prediction, future-scene prediction, and planning in a shared pipeline. It uses panoramic observations to build a bird’s-eye-view representation and predicts semantic features around candidate future locations. On RxR-CE validation unseen, it reaches 58.01 SR, versus 57.10 for G3D-LF. Its end-to-end design scores 64.29 nDTW versus 63.59 modular, on that split.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.