Open-source robot vacuum you build yourself.
Clean well · Raspberry Pi · ROS2 · Home Assistant · 2D LiDAR · 3D printed · ESP32 · Arduino
OOMWOO is an open-source home robot vacuum you can build yourself, made for the Raspberry Pi, ROS2, Home Assistant, and 3D-printing communities. It uses an affordable 2D LiDAR to map your home and navigate on its own. Local, no cloud required for regular functionality, no vendor lock-in. Follow the community newsletter.
Reference design images - this is approximately how the finished design will look:
- Affordable, fully open hardware, software and firmware
- Home appliance product quality - not a throwaway build
- Easy to build, with step-by-step zero-to-hero instructions
- 2D LiDAR mapping and autonomous navigation (ROS2 / Nav2)
- Native Home Assistant integration for local control
- 3D-printable, documented, and hackable chassis
- Buildable from parts you source yourself
- Local, no cloud required for regular functionality
- Optional extra functionality when connected cloud
- Apps on top of ROS2 to customize vacuum operation
- Stretch goal: App store
- Stretch goal: LeRobot integration, OpenClaw
v0 target: bare-bones build:
- 3D-printed chassis
- ROS2 Gazebo sim
- LiDAR with manual SLAM
- ROS2 on Raspberry Pi 5 AND/OR ESP32 running micro-ROS with ROS2 on local PC - decision TBD
Open Source Deliverables:
- Software development environment, robot description package and tutorials (ROS2)
- Placeholder real vacuum cleaner and tutorials (temporary while OOMWOO is being designed)
- Bill of materials (BoM) (in progress)
- 3D-printable files
- Firmware
- Motor drivers and sensors I/O PCB
- Build, setup, bringup and troubleshooting instructions
- Demo video(s)
Would you like to contribute? See CONTRIBUTING for the full guide.
OOMWOO is organized to built by the community, massively in parallel. The vacuum and its software are subdivided into modules, see list below.
A volunteer picks whatever module she wants, works on that module whenever she wants, submits her contribution as a PR under contributions/module-name/.
Multiple developers are welcome to work on the same module. The best solution for each module surfaces for over time, with the project master having the last call.
- Pick a contribution from the list below.
- Let us know you're working on it and your progress.
- Check ARCHITECTURE.md for the system design and interfaces.
Follow us building in public:
- Community newsletter.
- Discord
- YouTube: build-in-public channel
- X: @0OMWO0
Every module below is actionable now — build it against the Gazebo simulation (oomwoo-one) or a real placeholder robot (a Proscenic M6 Pro connected to ROS2), until OOMWOO hardware is ready. Pick one, tell us in Discussions, and open a PR.
Planned and on-hold modules (mechanical design, later-phase software) live in the RFC backlog.
- OOMWOO ROS2 and Ubuntu installation source code
- OOMWOO ROS2 URDF package and config source code
- remakeai reference vacuum teardown — a consumer LiDAR vacuum with a basic dock and stationary mop.
- AlieksieievYurii/vacuum-cleaner — a DIY 3D-printed robot vacuum (Raspberry Pi Zero W, gyroscope-based, Fusion 360, Android control app, no dock)
- kaiaai/LDS, kaiaai/lds2d — open-source 2D LiDAR libraries (C++, Python) supporting 23+ LiDAR models
- remakeai/vacuum_ros2_bridge — ROS2 bridge for a 3irobotix CRL-200-based vacuum (Proscenic), full ROS2 control
- Valetudo — cloud-free firmware replacement for commercial vacuums (local app-level control, not ROS2)
- Dennis Giese / robotinfo.dev — teardowns and rootability of commercial robot vacuums.
- codetiger/VacuumTiger - 3irobotix CRL-200-based vacuum low-level control reverse engineered
- Build a ROS2/LiDAR robot crash course - watch this if you have no robotics experience
- Open Mower - open-source outdoor lawn mower
We reviewed the 2025–2026 consumer robot vacuum landscape (global + China-sourceable brands, all price tiers) to decide which solutions to copy and which to skip. Key takeaways for the build:
- Suction is a sourcing problem, not an engineering one. Real-world cleaning does not track advertised suction (Pa); ~$500 mid-tier models beat flagships. A moderate sealed sourced motor + a good brush + tight airflow sealing matches flagships — no custom impeller needed.
- "Never gets stuck" needs camera + AI sensor fusion, not LiDAR alone — LiDAR is blind below its ~10 cm turret (cables, socks). v1 leans on the bumper for low obstacles; vision-based avoidance is a later / experimental goal, not an MVP promise.
- Anti-tangle brush: a tapered rubber roller resists hair-wrap best (a top user complaint) and is easy to 3D-print.
- Mop: a 3D-printed dual-spinning mop is competitive; the self-washing roller mop's edge is overstated and hard to replicate — skip it for now.
Well-loved models worth studying: Eufy Omni S2 (obstacle avoidance), Narwal Flow (roller mop), Ecovacs Deebot T90 Pro Omni (~$499 all-rounder), Dreame X40 Ultra (dual-spinning mop). Dreame is also the most Valetudo-rootable brand — the safest donor to study. (Per-model rankings are directional, from single-run reviewer tests.)
The project name "OOMWOO" is a rotational ambigram - it reads the same flipped 180°, like the robot itself, roaming your floor in every direction.
The project is sponsored by makerspet.com and remake.ai. We are reusing their open-source solutions.
- If you'd rather skip the parts hunt, a kit (motors, PCB, brushes, gaskets, LiDAR) will be available at makerspet.com, from the same maker behind this project. The kit is a convenience, never a requirement. Everything here stays open.
- When we get to apps, remake.ai will be providing its robot apps platform and app store. Using the app store will be entirely optional. The vacuum will always support cloud-free, local operation for regular functionality out-of-the-box.
Code is released under the Apache License 2.0.
Hardware design files, once added, to be released under an open hardware license (TBD).