dimensionalOS / dimensionalOS/dimos
M20 3D nav testing, bug reports and validation
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 4.5k
- Forks
- 808
- Avg merge
- 3d 5h
- Merged PRs (30d)
- 233
Description
BUGS:
OPEN QUESTION: Latency issue when operating rerun
a) hardware side?
-- less lickely
b) software side? which side, deeprobotic | dimos
-- most likely
--- poor integration
--- zenoh ?
c) compute issue?
--less lickely
Testing
Blueprints
Two blueprints we have:
- (A) m20-nav-3d
location: dimos/dimos/robot/deeprobotics/m20/blueprints/basic.py (Ivan made)
- (B) m20-simple-nav (we tried this in the office, it worked)
location: dimos/dimos/robot/deeprobotics/m20/nav/m20_simple_nav.py
Try to find a safe location. Better to have laptop and
use free hangar next to the office, if not ensure safety.
pre-tests
- Try both blueprints with:
dimos run <A | B>
2. Use simple testing
- a) check camera, lidar info present
- b) check QWEASD keyboard teleop
- c) check cliking on the map and response
- Make logs for both
- describe failures if any
- if m20-nav-3d works same as m20-simple-nav, use m20-nav-3d
- if not, use m20-simple-nav (we know it works)
main testing:
- Every testing must be recorded into memory
- Use like 10 cliks for each recoding session
- Do not do stairs for now when clicking. Only flat
A) Manuall
- find a clear space where you can actually click on the map with safety
- no obstacles around
- Click on the map and test how it moves
- The second guy always with the remote to stop the m20 when needed
B) Programmed
- ask an agent to write you a new blueprint based on the current you use
- ask to ingect synthetic paths instead of receiving a click
- MLSPlanner needed to be removed
- ask the agent to produce 4 straight lines and 4 curves
- you need to give agent the environment and some specs:
-- write four lines 4 metre each, fwd <-> bkw
-- write a circle (a different test)
The expected outcome is:
- you run:
dimos run m20-test-lines(some blueprint) - the dog comes up and runs what the agent prepared
- Always check safety, remote check, and if something wrong, Ctrl+c.
Basically the goal is to test these bluprints
- m20-nav-3d (should work, if not you may pass logs to agent)
- m20-simple-nav (we know it works)
Objective
- Collect evidence of how the robot moves for each working bluprint
- Inspect recordings later, check for:
- how well m20 tracks the path
- how well tha path is produced (like, should be straight, but is curved)
- how well the map is produced
Make a China Office recording
- Two people:
- one behind the laptop or computer, recording
- the other with remote teleoperating the dog
SAFETY
- Try to make a simple recording first, check if all is good
and the recording goes well
- Maintain safety when operating the m20
The dog is large; plan your route in advance.
- DO NOT CLICK ON THE MAP, use plain teleop and record
The planner may give you unexpected route, m20 may be hard to control
Synced from DIM-1084 by ivan
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the two named blueprints: dimos/dimos/robot/deeprobotics/m20/blueprints/basic.py and dimos/dimos/robot/deeprobotics/m20/nav/m20_simple_nav.py. Run both with dimos run, perform the listed camera, lidar, teleop, and navigation checks in a safe location, and record logs and sessions. Done means collecting evidence for each blueprint and inspecting path tracking, path shape, and map quality.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- robotics, testing
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100