YOU CAN BUILD A ROBOT BOAT SWARM FOR LAB RESEARCH FOR 522 CANADIAN DOLLARS A BOAT - OR 317 EACH WHEN YOU BUILD TEN
A printable autonomous boat for 522 Canadian dollars that lets you run a ten-vessel swarm in a lab tank without GPS.
by Calvin Gregory and Andrew Vardy
RoboticsScience
Built withArduinoRaspberry Pi3D printing
- difficulty
- ●●●●○
- time
- a weekend-plus
- license
- CERN-OHL-1.2 / GPL-3.0
- repo
- repo FINISHED0 stars
●●●●○ · a weekend-plus · CERN-OHL-1.2 / GPL-3.0 · 0 stars · repo FINISHED
WHAT YOU’LL NEED
Jump to section
COMPAREE VERDICT
This is a university robotics lab build published in a peer-reviewed hardware journal, not a weekend boat project. You are building a 23 cm autonomous surface vehicle for swarm research, and the entire positioning strategy depends on an overhead camera tracking AprilTags on each hull. If you do not have a tank, a camera rig, and the patience to tune controllers, this is the wrong project. What it delivers is real value: one boat costs 522 Canadian dollars and the per-unit cost drops to 317 when you build ten, while commercial unmanned surface vessels - 1.3 to 8 m long and made for open water - start around 30,000 US dollars. The hull prints on a hobby printer, the electronics are an off-the-shelf Raspberry Pi Zero W and Arduino Nano, and the host PC kept a stable 17 to 21 Hz pose update for up to ten boats. The one thing most likely to go wrong is the camera setup - you need a camera above your tank with a clear overhead view of every boat, calibrated, with the pose estimation running in real time; the authors recommend an Intel RealSense D435 over the Logitech C920 they started with.
IN THE REPO
GOOD TO KNOW
- —STL files for the hull, lid, keel, electronics bracket, stuffing tube and gearmotor brackets and a display stand are on the Open Science Framework, plus a SolidWorks assembly.
- —Bill of materials lists every component with supplier part numbers and costs in CAD.
- —Arduino and Python code for peripheral and primary controllers is included.
- —The paper explains the design and measured performance; detailed fabrication, assembly and testing instructions are on the OSF project wiki.
- —Hardware is CERN Open Hardware Licence v1.2, software is GPL v3, both permit commercial use with attribution and share-alike.
- —The overhead camera positioning system is required — this is not a standalone GPS boat, it is a platform for indoor tank research.
Parts to buy
9 itemsFrom our check of the build. Exact quantities and part numbers are in the creator’s BOM.
Can I build this?
Build at your own risk. Projects involve tools, electronics and sometimes mains voltage — follow the creator’s safety notes.
More builds like this
All projectsGallery
Start here
Navigation into the creator’s own docs — we don’t rewrite the guide, we route you to the source.
- 1.Read the HardwareX paper (The paper is the manual — it walks through the design decisions, assembly, calibration, and control.)
- 2.Download the files from the Open Science Framework (STLs, code, and the full bill of materials with part numbers are all there.)
- 3.Print the hull and check the fit of the electronics before ordering motors(The bracket is designed for specific motor dimensions — verify clearances with the STLs before committing to a bulk order.)
- 4.Set up the overhead camera rig and test AprilTag detection before building multiple boats(The positioning system is the entire reason this is cheap — if the camera cannot see the tags clearly, the boats are blind.)
KNOWN ISSUES
- The overhead camera is not optional — this is an indoor tank platform, not a GPS boat. If you do not have a way to mount a camera above your water with a clear downward view, the positioning system will not work.
- The 317-dollar figure (Canadian) is per boat when you build ten. One boat costs 522 because fasteners, propellers, shafts and adhesives come in packs you only use part of; electronics are about 200 of that.
- PID tuning is manual and iterative. The paper reports good waypoint following, but you will spend time adjusting gains for your specific motors, hull weight, and water conditions.
- The hull is sized for a specific motor and propeller combination. Substituting different thrusters may require reprinting the motor mounts or redesigning the bracket.
- The boats are designed for a calm indoor tank, not outdoor or rough water.
- You need a tank or pool large enough to run multiple boats without constant wall collisions. The authors used a 3.65 m by 3.65 m indoor tank.
Can I run this outdoors with GPS instead of the overhead camera?
Not without a significant redesign. The entire control architecture depends on the overhead camera streaming pose data back to the boats over WiFi. Adding GPS would mean rewriting the positioning stack, sealing the hull, and paying for GPS modules — which defeats the point of the low-cost design.
How many boats can one camera track at once?
The paper reports a stable 17 to 21 Hz pose update for up to ten boats, measured with an Intel RealSense D435 camera and updated software (the authors recommend it over the Logitech C920 they started with). Beyond that, the limits are camera resolution, tag size and how much of the tank you need to cover.
What happens if the WiFi drops or the camera loses a tag?
The boat loses its position estimate and the control loop fails. The paper does not describe a fallback behaviour — this is a research platform, not a production system.
Can I use a different microcontroller instead of the Raspberry Pi?
Possibly, but the onboard Raspberry Pi Zero W runs the control software and WiFi link while the Arduino Nano handles the motor controller and IMU; pose comes from the host PC. Swapping either board means rewriting code and checking it keeps up with a roughly 20 Hz update rate.
Community builds
No community builds yet — be the first, we feature the best ones.
Discussion1
FROM THE COMPAREE TEAM
The paper reports a stable 17 to 21 Hz pose update for up to ten boats in a 3.65 m square tank. What is the smallest useful swarm you could run in a home setup, and what would you use it for?
Calvin Gregory and Andrew Vardy
Calvin Gregory and Andrew Vardy work at Memorial University of Newfoundland, in the Department of Ocean and Naval Architectural Engineering and the Department of Computer Science respectively. They published the microUSV in HardwareX so labs can test marine swarm algorithms in an indoor tank, where commercial USVs - 1.3 m long and up, from 30,000 US dollars - would barely fit.
Keep building — related projects
Same category or platform, picked by how close they are to this build.
FULL CAD310 printed parts become a real WALL-E you drive from your phone
RoboticsFor advanced builders
NO CAD FILESYou can build a palm-sized robot dog that does backflips
RoboticsFor intermediate builders
FULL CADYou can build a robot that hangs from your ceiling and picks things up off the floor
RoboticsFor expert builders
FULL CADAmazing Hand: A Fully 3D-Printed Robotic Hand
RoboticsFor intermediate builders
FULL CADThese 60 dollars 3D-printed gloves let you actually feel objects in VR
DisplaysFor advanced builders
FULL CADA 300-dollar kit turns a decades-old lab microscope into an automated scanning system
ScienceFor intermediate builders
DISCLAIMER
- Comparee is not the author of the projects featured here. All rights to each project belong to its creator — every page links to the original source, and we never host creators’ files.
- Information is provided without warranty and may become outdated as projects evolve. Prices are indicative bands only — always check the creator’s parts list for current costs.
- Building and operating any project is at your own responsibility. Protective equipment, safe workshop practice and compliance with local regulations are the builder’s responsibility.
CompareeTEAM17d agoedited
Practical notes from our verification: the files are on the Open Science Framework rather than GitHub, so there is no star count or commit history; the project is published and finished, not actively developed. Fabrication, assembly and testing instructions are on the OSF project wiki, and three demo videos (linear path, elliptical path and multi-vessel test) are linked from the paper. The cost table is in Canadian dollars, with parts from suppliers such as BuyaPi, Digikey, Pololu, Amazon and McMaster-Carr, so expect conversion and shipping if you order elsewhere. Each boat runs a Raspberry Pi Zero W and an Arduino Nano on two 9 V batteries, while the pose tracking runs on a host PC with an overhead camera; the authors started with a Logitech C920 but recommend an Intel RealSense D435 for new builds. If you have never set up camera-based tag tracking before, budget time for calibrating that rig before you print the first hull. Correction (4 October 2026): we re-checked this page line by line against the project's own repository, documentation and videos, and fixed errors in earlier versions.