Next robot →

Zooids

Stanford Shape Lab / Inria (USA / France) · 2016 · All Stanford University robots

Not for saleNever sold. The UIST 2016 paper puts the cost at about $50 per robot as built in the lab, possibly down to $20 if mass manufactured. The projector, radio receivers and charging station cost extra. Hardware, firmware and software are on GitHub under CC BY-SA 4.0 (October 2026 check, last updated May 2023). Year 2016 is the UIST paper.
Not for sale (research)Via SSH / web / bridgeCoin-sizedTabletop displaySwarm run by one computerOpen hardware
Not for saleShipping: Not for sale. Never sold. The circuit boards, 3D-printed parts, charging station, firmware and software are published on GitHub for labs that want to build their own, and the tracking setup needs a TI DLP LightCrafter projector.
Remote control: A Mac runs ZooidManager (a prebuilt macOS app is in the repository, or it builds from source in Xcode or Visual Studio) and talks to the robots through USB radio receivers. Client programs in openFrameworks or Processing connect to ZooidManager over a network socket and send target positions.

Overview

Coin-sized wheeled robots built at Stanford's Shape Lab with Inria to act as a moving tabletop display, where each robot is a physical pixel that can also be grabbed and moved by hand. The robots do not talk to each other at all, since a central computer is in charge of the whole group. Each robot decodes its own position from gray-code patterns flashed by a 3,000 Hz projector above the table and reports it back by radio. The computer runs the application, plans collision-free paths for every robot (HRVO) and sends each one its next waypoint 60 times a second, with each USB radio receiver serving 10 robots. The catch is that it only works inside the area lit by a special high-speed DLP projector, the robots must turn before moving since the drive is not holonomic, and every robot has to be built from the published files.

Specifications

CategoryWheeled rovers · Research mobile manipulators & platforms
Released2016
CountryUSA, France
AvailabilityNot for sale (research)
ProgrammabilityOpen source (hardware + software)
Built-in autonomyNone / scripted · Moves only when controlled, or plays preset motions
LLM supportPossible via SDK · No LLM features. Applications are openFrameworks (C++) or Processing programs on a computer, which could in principle call an LLM to choose where the robots go.
MuJoCo simulationNone found · Nothing found.
NVIDIA Isaac simulationNone found · Nothing found.
Gazebo simulationNone found · Nothing found. The ZooidManager software has its own simulation layer that plans paths before sending goals to the real robots, but no Gazebo, ARGoS or Webots model was found.
Best forProfessionals & labs
Vision sensorsNo vision sensors
ActuatorsDC / gear motors
AudioNo audio
Compute typeArduino / ESP32 / microcontroller
Remote controlVia SSH / web / bridge
Size26 mm diameter, 21 mm tall, about 12 g
DriveTwo micro gear motors (FA-GM6-3V-25) placed off-axis to save space, differential drive with caster wheels
ComputeSTM32F051C8 (48 MHz Arm Cortex-M0)
VisionNone (two photodiodes on top read patterns from an overhead projector to find the robot's position)
LidarNone
SensorsTwo photodiodes for projector-based position tracking (about 73 updates per second, about 1.1 mm resolution over a 1 × 0.63 m area), capacitive touch sensing around the body, RGB LED, nRF24L01+ 2.4 GHz radio
RuntimeAbout 1 hour of continuous moving on a 100 mAh LiPo, longer in normal use
SpeedAbout 74 cm/s maximum, about 44 cm/s in typical use
SDKOpen-source ZooidManager (openFrameworks, C++) with a prebuilt macOS app, client APIs for openFrameworks and Processing, robot and receiver firmware for the STM32, example apps (shapes, keyframes, scatterplot, stop motion)

Ratings & reviews

Similar robots

Image credits