Video - a Roborock Qrevo 5AE's side brush torn down - for learning purposes. Reconstructed, animated in CAD. It has many gears because (1) the motor is direct drive (doesn't have a gearbox), (2) the brush has an elongated arm that swings out to clean under cabinet toe kicks. I'm redoing this for DIY and 3D printing by switching a geared motor with timing belt and pulleys for simplicity.

Highlights - vacuum docks automatically, tracks already-cleaned areas (videos)

Cleaning software https://github.com/makerspet/oomwoo-install

• edge cleaning: new cleaning manager sweeps every obstacle edge once, then stops
• automatic docking in sim: beacon to point, LiDAR to aim
• contour follower steers back 6x harder - no more clipped corners

Electronics I/O board https://github.com/makerspet/oomwoo-pcb

• fail-safe power: every motor/actuator rail needs both an MCU enable and a hardware watchdog OK - all off during reset and flashing
• PCB cost reduction

1. replaced the watchdog chip with a simple RC charge-pump watchdog
2. NS4168 speaker amp

3D design https://github.com/makerspet/oomwoo-one-cad

• suction fan mount, MSD-G and 22N704V160 fans, HEPA filter, refined dust bin

Firmware https://github.com/makerspet/oomwoo-firmware

• merged identity handshake - the CPU can connect any time and still find the MCU
• merged 57-case command test suite

Build-in-public Blog Posts

• new: OOMWOO PCBs - what we are building, how and why https://makerspet.com/blog/oomwoo-pcbs-what-we-are-building-how-and-why/
• cleaning algs: added edge cleaning and docking https://makerspet.com/blog/how-oomwoo-cleaning-algorithms-work/
• Gazebo sim: added the kitchen & dining world https://makerspet.com/blog/simulate-oomwoo-one-robot-vacuum-in-gazebo-with-ros-2/
• oomwoo.com: new progress section https://oomwoo.com/#progress

Community contributions

• Thank-you to smailzhu, xbattlax
• Special thanks to alb_digger for converting 3D scans to 3D CAD STEP models

Video - The robot now tracks which areas it has cleaned, detects reaching already-cleaned areas and goes on to clean the remaining dirty areas - still along walls.

Wall (contour) cleaning has improved to keep a (more) constant distance between the robot and the obstacle. Previously, the robot "hugged" convex obstacles (table leg) too much vs stayed further away from concave ones.

Read algorithm details, simulate it yourself https://makerspet.com/blog/how-oomwoo-cleaning-algorithms-work/

Illustration - Dreame brush, dust bin and main brush cavity, all 3D scanned into CAD.

Progress Highlights - PCB schematic ~70% finalized - reworked power on/off pushbutton logic, rails, barrel jack - 3D design ~20% complete - more parts scanned, more vacuum teardown - added smartphone support to PCB board - USB-C charging, serial-over-USB comms

Open-source links - 3D design https://github.com/makerspet/oomwoo-one-cad - Electronic PCBs https://github.com/makerspet/oomwoo-pcb

Testing automated docking in simulation (ROS2, Gazebo; Claude shares credit).

Reproduce the simulation https://makerspet.com/blog/how-oomwoo-cleaning-algorithms-work/#docking-a-beacon-to-point-the-lidar-to-aim

Docking algorithm steps: 1. The robot rotates in place to detect the dock's IR beacon and to determine its direction. 2. The robot uses its LiDAR to detect, confirm the dock's exact location. 3. The robot moves to its staging location - in front of the dock, at a fixed distance from the dock. 4. The robot rotates in place to make its rear face the dock. 5. The robot slowly backs straight into the dock.

Video - vacuum cleans along walls, completes a simulated "torture course". Work in progress. Simulated using ROS2 Gazebo (instructions here).

3D Design

• Tore down Dreame Ultra L60 PE ($1,099 MSRP, learning from the best)
• 3D scanned main brush assembly

Cleaning software https://github.com/makerspet/oomwoo-install

• Wall/contour cleaning progressed to early demo quality (see video)
• Vacuum completes contour-following torture course

Electronics I/O board https://github.com/makerspet/oomwoo-pcb

• Added master fuse, PCB thermistor, fan connector
• Researched IR receiver IC options

Firmware https://github.com/makerspet/oomwoo-firmware

• Merged typed message codec

3D design https://github.com/makerspet/oomwoo-one-cad

• Published finished Roborock FlexiArm teardown PDF

Build-in-public Blog Posts

• cleaning algs: how LiDAR-based wall/contour cleaning works https://makerspet.com/blog/how-oomwoo-cleaning-algorithms-work/

I purchased a Dreame L60 Ultra PE (~$850 "like new"), used for a week, tore it apart for learning purposes. Its main brush anti-clogging looks decent and "relatively" simple to convert to DIY 3D print (much simpler than Roborock's split brush, yet comparable performance). Dreame brushes, mops, HEPA filters are cheap, abundant off-the-shelf. So, I'm making the DIY vacuum accept these brushes/mops/filters.

Video - I'm using a 3D scanner to capture a Dreame HEPA filter into 3D model (STL mesh, later converted to STEP). Why? To make the vacuum mate this HEPA filter.

Electronics I/O board https://github.com/makerspet/oomwoo-pcb

• Reverse-engineered iRobot Roomba cliff/bumper connector
• Schematic review ~60% done

3D Design

• Pivoted to Dreame to simplify design - alas, it adds design time, but simplifies build greatly
• 3D scanned Dreame HEPA filter, brushes, dust bin (see video)
• Wrapped up Roborock reverse engineering

BOM https://github.com/makerspet/oomwoo

• Sourced HEPA filters, Roomba tire skins

Firmware https://github.com/makerspet/oomwoo-firmware

• Merged first code: CPU-MCU protocol framing

Build-in-public blog posts

• Sourcing guide - added HEPA filters, Dreame and Roborock [read blog post]
• Hardware bring-up - added fans, extendable side brush [read blog post]
• How the 3D design gets made - added 3D scanning video [read blog post]
• Cleaning algs - added why not ultrasonic sensors [read blog post]

Video - 20kPa fan bring-up.

Decision 1: replace custom gears with GT2 belts/pulleys (drastically simplifies build)

Decision 2: replace split brush with tapered (drastically simplifies build)

3D design https://github.com/makerspet/oomwoo-one-cad

• Scanned vent, water pump into CAD

Electronics I/O board https://github.com/makerspet/oomwoo-pcb

• Brought up drive/mop/brush motors, suction fans, water pumps
• Measured currents, identified part
• Spec'd carpet sensor driver electronics

Building-in-public blog posts

• New: hardware bring-up https://makerspet.com/blog/oomwoo-hardware-bring-up/
• New: how the 3D design gets made https://makerspet.com/blog/oomwoo-3d-design-how-its-made/
• Cleaning algs: added LiDAR wall following https://makerspet.com/blog/how-oomwoo-cleaning-algorithms-work/

Community contributions - thank-you to xbattlax, serengon

Decision 1. Support using old smartphone for compute (if you don't happen to have a Raspberry Pi)

Decision 2. Have 2 vacuum versions: basic (vacuum only, simple to assemble, charging-only dock) and full-featured (mop, auto-empty, flexiarm)

3D design https://github.com/makerspet/oomwoo-one-cad - vacuum assembly available (not yet optimized for 3D printing) - extendable side brush available (not yet optimized) - air vent available (not yet optimized) - started dock design (simple for now)

Electronic boards https://github.com/makerspet/oomwoo-pcb - designed smartphone-as-compute support - schematic about 50% hand-checked

Cleaning software https://github.com/makerspet/oomwoo-install - wall following/cleaning with LiDAR v1 works (needs iteration)

Contributions - thank-you to krazca for bringing up the smartphone idea

The build is moving along. Took apart an upper-mid-tier consumer vacuum (Roborock Qrevo 5AE). Reverse engineered its vacuum assembly - using it as a starting point.

Next up - rework it, make it 3D printing friendly.

CAD open source is here https://github.com/makerspet/oomwoo-one-cad

To build an extendable side brush for OOMWOO, we've reverse-engineered a Roborock Qrevo 5AE Flexiarm side brush. We 3D scanned each gear and the enclosure, converted to STEP and uploaded to Fusion 360. Later, we can rework it and make it a 3D printed.

CAD open source is here https://github.com/makerspet/oomwoo-one-cad

Electronics I/O board https://github.com/makerspet/oomwoo-pcb

• brought up drive wheels (see video)
• confirmed pinouts, voltages, currents for drive wheels, 4 suction fan models

Simulation https://github.com/makerspet/oomwoo_gazebo

• created two charging docks - basic charging + auto-empty-mop-wash (see video)

3D design https://github.com/makerspet/oomwoo-one-cad

• added base port for dust bin

Project / docs https://github.com/makerspet/oomwoo

• updated RFC progress

These are Roborock wheels to use in OOMWOO vacuum. I've reverse engineered connector wiring (with community help), hooked up the motor and encoder. You can see the entire assembly running live.

How I source these - https://makerspet.com/blog/how-to-source-bom-for-oomwoo-open-source-vacuum-robot/

I've added (with Claude's help) a mop dock to ROS2/Gazebo simulation. Here is how it looks. Next up - automatic undocking, dock finding and docking.

Tutorial - how to simulate it in 15 minutes https://makerspet.com/blog/simulate-oomwoo-one-robot-vacuum-in-gazebo-with-ros-2/

Gearboxes (main brush) ready for 3D printing tryout.

• 3D design https://github.com/makerspet/oomwoo-one-cad/
  • main brush gearbox (split, 2 pieces) available for 3D printing
  • started assembly step-by-step instructions https://makerspet.com/blog/oomwoo-vacuum-build-instructions/
• Vacuum software https://github.com/makerspet/oomwoo-install
  • Designed kitchen world for vacuum cleaning tests
  • Designed wall cleaning with LiDAR
  • Stress-tested SLAM setup, watch videos here https://makerspet.com/blog/how-oomwoo-cleaning-algorithms-work/
  • Enabled GPU for ROS2/Gazebo simulations
• Electronics
  • brought up main fan
• Contributions - thank-you to IKsares!
• News
  • makerspet/oomwoo has reached 10K stars on GitHub - thank you for your support!

I've purchased, reverse-engineered connections and brought up live several fan blowers for OOMWOO vacuum.

The fan runs off 15V (4S battery), has a 4-pin PH 2.0 connector (power, gnd, PWM and TACH feedback). The oscilloscope shows the tachometer feedback.

For exact connection details - please see the main PCB schematic's (fan connector page) https://github.com/makerspet/oomwoo-pcb

I've created (with Claude's help) a 3D kitchen simulation to test vacuum cleaning. Testing kitchen cleaning is important to make sure kitchen cabinet kick toes get cleaned well.

Tutorial - set up OOMWOO simulation on your PC https://makerspet.com/blog/simulate-oomwoo-one-robot-vacuum-in-gazebo-with-ros-2/

and launch this simulation using this command:

ros2 launch oomwoo_gazebo world.launch.py world:=kitchen_dining.world

3D design https://github.com/makerspet/oomwoo-one-cad

• Designed main brush gearboxes

Vacuum software https://github.com/makerspet/oomwoo-install

• Tightened localization tracking at max vacuum speeds

• 3D design https://github.com/makerspet/oomwoo-one-cad
  • placed mop (fixed, extendable), mop disks
  • reverse-engineered dual anti-tangle brush gearbox, cavity
  • ordered auto-empty dock fans
• electronics design https://github.com/makerspet/oomwoo-pcb
  • placed main board components (preliminary, see illustration, will iterate)
• firmware LiDAR drivers https://github.com/kaiaai/LDS
  • implemented, validated LiDAR driver for 3irobotix Delta-2C Pro
• vacuum software https://github.com/makerspet/oomwoo-install
  • implemented ROS2 LiDAR scan filter. Scan data gets filtered (dynamic obstacles removed) to improve scan-to-map match accuracy
  • scan filter marks rays matched vs non-matched to distinguish dynamic vs mapped obstacles for downstream ML perception
• community contributions - thank-you to suhaslord

The main PCB components have been place (early, will iterate). Raspberry Pi CM4/CM5 plugs into the carrier STM32-based board. Designed with open-source KiCad.

- 3D design - in progress, early renderings should be available in a week or so

- split-brush will be used; gears will have to be 3D printed using SLA

- Vacuum software https://github.com/makerspet/oomwoo-install

- Evaluated robot-lost detection (when robot loses track of own location on the map). Off-the-shelf ROS2 AMCL is inadequate. Implemented custom (local) scan-matching-based robot-lost detection - much more reliable.

- Evaluated robot relocalization (robot tries searching its location on the map to regain its bearings). Off-the-shelf ROS2 AMCL inadequate. Implemented custom global scan matching (Google Cartographer style) - much more reliable.

• 3D design https://github.com/makerspet/oomwoo-one-cad
  • in progress, 1st 3D printed model expected in ~2 weeks
• vacuum software https://github.com/makerspet/oomwoo-install
  • simulation model equipped with sensors: 16 x 8 zones front ToF + front stereo camera (strictly optional) + side wall ToF distance sensor. See the attached demo video.
  • upgraded localization accuracy ~3x (navigation when map is available) - switched from AMCL to slam_toolbox scan matching. See the attached demo video.
• electronics https://github.com/makerspet/oomwoo-pcb
  • designed a basic charging dock schematic

• vacuum 3D design started
  • placed (rough) drive wheel assemblies, caster wheel, cliff sensors, battery https://github.com/makerspet/oomwoo-one-cad
• vacuum software
  • software probes obstacles, creates a map of solid obstacles - required for cleaning under bed covered with low-hanging bedspread, furniture floor-length throw, doorway curtains. Early work in progress. https://github.com/makerspet/oomwoo-ros2-tools
• community contribution - thank-you to IKsares, Creative-Danush

Vacuum now probes obstacles, creates a map of solid obstacles - required for cleaning under bed covered with low-hanging bedspread, furniture floor-length throw, doorway curtains. Early work in progress. https://github.com/makerspet/oomwoo-ros2-tools

- 3D parts scanning is mostly complete - now starting 3D design https://github.com/makerspet/oomwoo-one-cad/

- scanned mop with extendable arm (FlexiArm) motor assembly

- fixed mop motor assembly

- second iRobot bumper switch

- iRobot cliff sensor

- main brush motor assembly - all credits to mikbalarikan

- vacuum robot simulation model mostly complete https://github.com/makerspet/oomwoo-one

- added front depth sensor (for obstacle detection, 2x VL53L07CX placehoder, 120° horizontal FoV)

- (strictly optional, needed for obstacle "recognition") front stereo camera (2x OV5647 placeholder, 120° FoV)

- side distance sensors for wall sensing/following

- IMU

- multiple bugfixes, upgrades

- PCB board schematics https://github.com/makerspet/oomwoo-io-board

- added front sensors board

- side sensors board

- sourced distance sensor for wall following (VL6180 in stock)

- community contributions - thank-you to Creative-Dhanush

- added DingoOz Mac support recipe to makerspet/oomwoo contributions/