Algorithmic code pertaining to monocular mapping, by detecting structural vertices in panoramic images and fusing them into a polygonal map. Results include experiments for both a single robot and a 3-robot, centrally coordinated team. This module reproduces core algorithm and results first presented in the following paper:
Gregory J. Stein, Christopher Bradley, Victoria Preston, and Nicholas Roy. “Enabling Topological Planning with Monocular Vision.” In: International Conference on Robotics and Automation (ICRA). 2020. paper, talk (10 min).
@inproceedings{stein2020gapnav,
author = {Gregory J. Stein and Christopher Bradley and Victoria Preston
and Nicholas Roy},
booktitle = {International Conference on Robotics and Automation (ICRA)},
title = {Enabling Topological Planning with Monocular Vision},
year = 2020,
keywords = {topological mapping, topological planning, frontiers,
multi-agent search, exploration, learning},
}Readers are referred to the paper for algorithmic details.
Note: make build (see top-level README) must be successfully run before running the following commands.
The Makefile.mk provides multiple targets for reproducing results.
make vertexnav-dungeonwill generate results from our multi-room “Dungeon” environment for both a single robot and for a centrally-coordinated 3-robot team.
Both targets are run in single-threaded mode by default, however both data generation and evaluation can be run on multiple seeds in parallel. As such, running make vertexnav-dungeon -j3 will run three concurrent instances. As data generation has a smaller VRAM footprint than does evaluation, it is often possible to run more concurrent instances specifically for data generation:
make build
make vertexnav-dungeon-generate-data -j6
make vertexnav-dungeon -j3GNU Make keeps track of progress, so if (for whatever reason) the code is stopped during a run. Rerunning the above commands will resume where it left off, without the need to redo any of the completed data generation, training, or evaluation instances.