Floor-Surface Handling & Edge Cleaning (ROS2 package)

Clean better by adapting to the floor. This package adds wall following (to clean tight against walls), carpet-edge following, recognizing carpet vs hardwood, and lifting / lowering the mop accordingly (never wet a carpet). Because the physical robot isn't built yet, this is a Gazebo simulation — you model the surfaces and the surface sensor in Gazebo; it is later re-validated on hardware in the live-robot-bringup RFC.

Status — ready to start work. No need to wait for OOMWOO hardware — develop it in the Gazebo sim (urdf-gazebo-sim) or on the real placeholder Proscenic M6 Pro. Say so in the discussions so we can coordinate.

Important References

clean-and-map RFC — coverage cleaning that this refines at edges and surface transitions.
urdf-gazebo-sim RFC — robot URDF; this package likely needs a surface sensor and a mop lift/lower actuator modeled.
ROS2 software interfaces — shared topic/action/service contract for simulation-first modules.
OOMWOO ROS2 development — build OOMWOO ROS2 Docker image(s) with your packages.
Project discussions
Discord server

Note: this package may need new sim sensors/actuators (surface sensor, mop actuator). Add them inside your submission for now and propose folding the stable ones back into urdf-gazebo-sim later — please don't rewrite that RFC.

Request for Contribution - Instructions

wall following
- detect walls from the LiDAR and drive a controlled offset along them so edges get cleaned, not just the room interior
- integrate with clean-and-map coverage so edges and interior are both covered, without re-cleaning everything
- post in Project Discussions to let everyone know you're working on it, and post your progress
carpet-edge following
- detect carpet boundaries and follow the edge to clean along them
surface recognition (carpet vs hardwood)
- model a plausible sensor in sim (wheel current / IMU vibration / a dedicated downward sensor) and publish the surface type
mop lift / lower
- actuate the mop up on carpet, down on hardwood; model the actuator in Gazebo; never wet a carpet
- coordinate with cleaning-jobs for mop-related job constraints
test it well
- build worlds with mixed carpet + hardwood regions and walls
- verify edge coverage, correct surface classification, and correct mop state at every transition
regression tests (headless, CI-friendly)
- surface-classification accuracy
- edge-coverage percentage (walls / carpet edges actually cleaned)
- zero carpet-wetting events (mop never down on carpet)
submit a PR (pull request) to contributions/floor-care/<your-github-username>/
- link to ROS2 package(s) and any sim sensor/actuator + world additions
- instructions, documentation - how to install, run, configure, troubleshoot, test results
- videos of wall/edge following and mop lift/lower at surface transitions
- announce your submission in Project Discussions
iterate with review
TBD, expect the RFC to evolve

Acceptance criteria

Objective, measurable. Examples:

Wall / carpet-edge following cleans tight along edges and integrates with coverage (high edge-coverage %)
Surface recognition classifies carpet vs hardwood accurately
Mop lift/lower tracks surface type correctly with zero carpet-wetting events
Regression tests pass (classification accuracy, edge coverage, no carpet-wetting), runnable headless in CI
Works in at least one mixed carpet + hardwood world
Documented and reliably reproducible by someone else
TBD, expect criteria to evolve

The maintainer selects among compliant candidates using these criteria. Multiple attempts are welcome and useful even if not selected — modules are swappable, and a non-selected design is still a valid learning exercise and a fallback.