facebookresearch / facebookresearch/fairo

droidlet slam improvements checklist

Open
#730 0 comments 0 reactions 2 assignees Claimed by @soumith View on GitHub
Dominant language
Jupyter Notebook
Stars
929
Forks
123
PR merge metrics
No merged PRs in 30d

Description

## Type of Issue

Adding a laundry list of improvements for slam on droidlet

- [x] Modularize SLAM more to build the right abstractions - for example how the distance matrix in fmm planner is accessed should be opaque to slam. This has been a source of a bunch of indexing bugs (because (x,y) co-ordinates need to be accessed as (y,x))
- [ ] We need SLAM in memory for better correctness of things. For example, checking if an object in memory is in line of sight of the agent is not possible without knowing the obstacles (which is missing from memory)
- [ ] Add tests for fmm planner, slam.py
- [ ] Handle exceptions in Slam gracefully - currently it crashes the agent.
- [ ] How can we speed up the time of getting a slam exploration outcome (currently requires launching habitat + letting the agent run for minutes + visualizing the frames dumped)
- [ ] What are some better terminating conditions - currently it bails out when robot reaches a loc which is marked as an obstacle
- [ ] Make mover level APIs collision aware
- [x] Improve slam visualization - integrate world_view branch for 3D visualization into the dashboard
- [ ] Move from print statements to logging

Select the type of issue:
- [ ] Bug report (to report a bug)
- [ ] Feature request (to request an additional feature)
- [x] Tracker (I am just using this as a tracker)
- [ ] Refactor request
- [ ] Documentation Ask

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.