Physical Ant
A quadruped ant-style robot with eight Dynamixel actuators that runs learned locomotion policies directly on hardware, controlled by a single computer over USB.
- Quadruped
- Open source
- Works with AI
- 3D printable
Suggested by Robotheca, created by @Openmind-Research-Institute (opens in a new tab), from github.com/Openmind-Research-Institute/open-ant (opens in a new tab). Pinned to commit a3c5a38 (opens in a new tab).
- Hardware license
- MIT (opens in a new tab)
- Software license
- MIT (opens in a new tab)
- Commercial use
- Allowed
Some folders have their own license: embodied_ant_env/apriltags/ (BSD-2-Clause).
This repository contains executable files that Robotheca doesn't check. Only run them if you trust the creator.
Photo and 3D model
3D preview converted from github.com/Openmind-Research-Institute/open-ant at a3c5a38, under MIT. View the original files from GitHub (59.1 MB) (opens in a new tab)
What it can do
Written by Robotheca from the project's documents.
- Walks as a quadruped ant with hip and knee joints, hip range about +/- 45 degrees and knee range about +/- 70 degrees.
- Learns locomotion with reinforcement learning agents (SAC and SARSA) run from a laptop, for simulation or hardware.
- Runs a MuJoCo simulation of the ant for training and testing before hardware.
- Uses an overhead camera tracking fiducial markers to compute position, heading and rewards for locomotion tasks.
- Estimates body orientation with an IMU flight controller; signals include joint position and velocity, body angular rate, and commanded velocity.
Works with AI
Found by Robotheca in the project's files at commit a3c5a38, its robot.yaml and Robotheca's notes on its documents. Never guessed; Robotheca doesn't run the code.
“Learns locomotion with reinforcement learning agents (SAC and SARSA) run from a laptop, for simulation or hardware.”
Robotheca's notes: What it can do
Bill of materials
20 parts20 of 20 parts shown
≈ € amounts use the ECB reference rate of 15 Oct 2026, for comparison only.
1 of 20 parts have no price in the source, so there is no parts total. Read from Readme.md at the pinned commit.
Build guide
The creator's steps, each with its instructions.
- 1Configure Dynamixel motor IDs and set baudrate to 1Mbaud before assembly
- 2Assemble the robot using the 3D printed parts and mounting screws
- 3Mount the IMU autopilot and USB hub
- 4Install Python dependencies in a virtual environment
- 5Create the robot config file with the April tag ID and set IMU port, camera ID a
- 6Run the simulation or the SAC/SARSA agents on hardware
The guide appears here as you scroll.
Guide by @Openmind-Research-Institute (Readme.md), under MIT.
More guides
Wiring and electronics
Notes written by Robotheca from the project's documents.
Main parts
- Four Dynamixel XM430-W350-T and four Dynamixel XC430-W240-T servo actuators, eight joints in total.
- Dynamixel U2D2 USB-to-Dynamixel interface with power hub board and 12V power supply.
- Kakute H7 Mini / TBS Lucid Freestyle mini flight controller used as an IMU.
- Two Logitech Brio 101 cameras: one on-board and one top-down overhead camera for tracking.
- 12V 8A AC to DC power adapter; the robot runs from wall power with no battery.
Connections
- Each Dynamixel servo is daisy-chained on a single bus and connected to the computer through the U2D2 over USB.
- The U2D2 is powered from the 12V adapter through the power hub board; power connects to the control board's screw terminals instead of the barrel connector to avoid dropouts.
- The IMU flight controller connects to the computer over USB-A to USB-C; the U2D2 uses a short USB-A to micro-USB cable.
- A USB hub joins the peripherals and a USB-A extension cable serves as the tether to the robot.
- No soldering is required; the robot is connected by cables to AC power and to an external computer running the agent.
Cables and connectors
From the parts list.
- Cable Matters Ultra Mini USB Hub × 1
- Short USB-A to USB-C Cable × 1
- Short USB-A to micro-USB cable × 1
- USB-A extension cable × 1