Conventional approaches to vision-and-language navigation (VLN) are trained end-to-end but struggle to perform well in freely traversable environments. Inspired by the robotics community, we propose a modular approach to VLN using topological maps. Given a natural language instruction and topological map, our approach leverages attention mechanisms to predict a navigation plan in the map. The plan is then executed with low-level actions (e.g. forward, rotate) using a robust controller. Experiments show that our method outperforms previous end-to-end approaches, generates interpretable navigation plans, and exhibits intelligent behaviors such as backtracking.
@article{arxiv.2012.05292,
title = {Topological Planning with Transformers for Vision-and-Language Navigation},
author = {Kevin Chen and Junshen K. Chen and Jo Chuang and Marynel Vázquez and Silvio Savarese},
journal= {arXiv preprint arXiv:2012.05292},
year = {2020}
}