QWE AI Academy Key takeaway:Oomwoo, an open-source robot vacuum you build yourself, just landed in mid-June 2026 – and yes, it’s trending. But the honest truth buried under the hype: there’s no bill of materials yet, no build guide, and the chassis hasn’t been physically assembled. Do not try to build one this week. Instead, run the Gazebo simulation locally in about 15 minutes and (if you have skills) claim a module. That’s the only real move today.
Wait, what is oomwoo?
Oomwoo is a Maker’s Pet project by developer Ilia O., released under the Apache 2.0 license. It combines a Raspberry Pi 5, ROS 2 with the Nav2 stack, a cheap 2D LiDAR, an ESP32 running micro-ROS, and a 3D-printed chassis – controlled natively from Home Assistant, with no cloud. Coverage spread quickly across tech outlets after the first commit went public.
The hook isn’t just “open source vacuum.” The reference design uses only LiDAR and bumper sensors – no camera pointed at your living room. Given the last two years of vacuum-hijack stories, that’s the actual selling point.
Build now vs. wait: pick one
The community is splitting into two camps. Here’s the honest tradeoff:
| Path | Method A: Build hardware now | Method B: Run the sim + contribute |
|---|---|---|
| What exists today | Nothing physical – status is RFC/design stage, first BOM targeted mid-July 2026 | Software dev environment is ready; sim runs in ~15 minutes |
| Cost | Unknown – creator targets $100-$200 in parts + a Pi 5 4GB (kit pricing not yet published) | Zero (you already have a computer) |
| Risk | You’ll buy parts that may not match the final BOM | None – worst case, you learn ROS 2 |
| Payoff | Bragging rights once hardware lands | Your PR could ship in v0 (deadline August 31, 2026 per community tracking) |
Method B wins for anyone reading this in July 2026. There is literally no way to build the hardware – the parts list doesn’t exist. Method A is the trap: people are already asking about cost estimates on the launch blog, and the creator’s answer is basically “give me a week.”
The winning path: run oomwoo in simulation
This works today on Linux (or WSL2 on Windows). You need ROS 2, Gazebo, and about 15 minutes per the official launch post. When the sim spins up, you get the URDF robot model moving through a virtual room – it’s not glamorous, but it’s the only version of oomwoo that actually exists right now.
Step 1 – Clone the repo
git clone https://github.com/makerspet/oomwoo.git
cd oomwoo
cat docs/design-document.mdRead the design doc before touching anything. It tells you which modules are green (ready), amber (blocked on parents), blue (needs real hardware), and grey (on hold). Green is where you can help immediately.
Step 2 – Install ROS 2 and Gazebo
Follow the official Gazebo + ROS 2 installation guide. The oomwoo v0 target uses a Gazebo sim with a URDF model – this is what the urdf-gazebo-sim module produces, and it’s marked ready in the repo’s module graph.
Step 3 – Pick a module and open a PR
Contributions go under contributions/module-name/ as pull requests. Multiple people can work on the same module in parallel – the maintainer picks the best solution over time. Green-lit modules as of the last README update: urdf-gazebo-sim, source-3d-models, and part-specs.
Pro tip: Don’t start with the vacuum-fan or dust-bin modules even though they look approachable. They’re marked “on hold” in the module dependency graph – your PR will sit unreviewed. Start with
part-specsif you own a caliper and want a low-friction way in.
Edge cases the news articles skipped
Every write-up so far has covered the same three security stories and the same Valetudo comparison. Here’s what they missed.
- The Pi 3B+ won’t work. LiDAR mapping needs 4GB RAM minimum for the first release, per the creator’s own reply in the launch thread. If you already have a spare Pi 3B+ sitting around thinking you’ll re-use it – you can’t. The Pi 5 4GB isn’t a suggestion, it’s a floor.
- LiDAR is blind low. A 2D LiDAR is physically mounted about 10 cm up on its turret. Anything shorter than that – phone charging cables, socks, a cat toy – is invisible to the sensor. The v0 design leans on the bumper to bounce off these. Vision-based avoidance is an experimental later goal, explicitly not part of MVP. If your floor is a cable maze, you’ll be picking oomwoo up a lot.
- Retrofitting an existing vacuum breaks docking. A community question asked whether you could scrap the PCB of an existing robot and drop in a Pi + LiDAR + I/O board. Creator’s answer: yes, it’s on the roadmap, but the original PCB usually contains the docking IR sensors. Rip it out and your robot can’t find its charger anymore.
- The price argument is contested. On Hacker News, commenters pointed out you can buy a LiDAR vacuum new for £70-80 (as of mid-2026) or scavenge one used for parts. The $100-$200 build cost only wins against $500-$600 flagship models, not the whole market.
- Kit pricing is unknown. The convenience kit (motors, PCB, brushes, gaskets, LiDAR) is announced but not priced. Anyone budgeting off “$100-$200 in parts” is guessing until the BOM ships – and the kit price, once published, may shift that total significantly.
Why the module system actually matters
Here’s the structural bet Oomwoo is making: most open-source hardware projects stall because everything depends on one person. Split the vacuum into modules with an explicit dependency graph, and that bottleneck breaks. Contribute to urdf-gazebo-sim today – no hardware needed – and you directly unblock clean-and-map and recovery-safety downstream. That’s not just organizational tidiness; it’s how design-stage work gets to a shipped v0 by August 31.
It’s also a low-key filter: if you contribute to urdf-gazebo-sim, you don’t need any hardware, and your work unlocks clean-and-map and recovery-safety downstream. That’s how a design-stage project becomes a shipped product.
Should you actually care about this yet?
Only if you’re a ROS 2 or robotics tinkerer. If you want a vacuum that works today, buy one. If you want to root an existing one and go cloud-free, the pre-existing path is Valetudo – maintained since 2018, but it requires rooting vendor firmware, voids the warranty, and can’t be undone on supported models.
Oomwoo’s differentiator is starting from zero, which sounds worse until you consider that Valetudo’s install steps involve disassembling a working appliance. Neither is casual-user territory.
Frequently asked questions
Can I build oomwoo this weekend?
No. The bill of materials doesn’t exist yet – targeted for mid-July 2026 – and no prototype has been physically assembled. Run the simulation instead.
Is the $100-$200 budget realistic once the BOM drops?
It’s the creator’s target, not a confirmed cost. The reasoning is in the repo’s market analysis: suction is a sourcing problem, not an engineering one – a moderate sealed motor plus a good brush plus tight airflow sealing matches flagship performance, no custom impeller needed. That said, the Pi 5 4GB adds to that figure (check current pricing), and the convenience kit has no published price yet. Real total cost is genuinely unknown until the BOM and kit pricing land together.
What’s the difference between contributing to a green module and an amber one?
Green modules (urdf-gazebo-sim, source-3d-models, part-specs) are unblocked – their dependencies are done and your PR can be reviewed and merged now. Amber modules are waiting on green ones to finish first. Submit an amber PR today and it may sit unreviewed until the parent module lands, so factor that into where you spend time.
Do this next: open the oomwoo design document, find the module graph, pick one green module, and leave a comment on the repo saying you’re working on it. That’s how you claim it. Everything else – parts, printing, assembly – waits for July.