RatBot
An open robotics platform, from hardware to autonomous multi‑robot systems.
An open-source platform for education and research, spanning electronics, embedded systems, ROS 2, autonomous navigation, multi-robot coordination, and mission planning.
RatBot was born at LISV from a simple need: a fleet of identical robots, fully under our control, for research and teaching. Electronics, real-time firmware and ROS 2 are designed here, documented and measured.
Capabilities
RatBot in action
What the robot does on its own, shown rather than described.
01
Autonomous exploration
Dropped into an unknown place, the robot picks where to go next by itself: it heads for the edges of what it has not seen yet and maps them with its lidar, room after room.
In simulation, on the lab floor: 95% of the plan covered (median of eight runs).
Exploring the lab floor in simulation, on one run: 97% of the plan in 49 simulated minutes (sped up 60 times). 02
Navigation
On a saved map, the robot is given a destination: it plans its path, follows it and steers around the obstacles it meets (Nav2).
Why
Why another robot?
In the lab, no commercial robot met all four of our needs at once.
A fleet, not a robot
Multi-robot research needs several identical machines that can be repaired and reproduced, not a single showcase robot.
Nothing closed
Commercial platforms often stop at the ROS interface, with a black box underneath. Here every layer opens, from the board to the software.
Know where the numbers come from
An experiment is only worth what its measurements are worth: odometry, clock and safety are designed, checked and documented.
A thread for teaching
Electronics, real time, ROS 2, navigation: the same robot can run through a whole course, from one end to the other.
The robot
What we built
A small differential robot, built around an electronics board designed in the lab.
Inside
- A differential base driven by two Dynamixel AX-18A servomotors
- A custom STM32H750 board: traction, safety and sensors
- An RPLidar A1 laser scanner and an inertial measurement unit
- A front 8 × 8 time-of-flight distance sensor, for low obstacles
- Real-time firmware with FreeRTOS and micro-ROS
- A Raspberry Pi 4 running ROS 2 Humble
- An OLED screen, an RGB LED, a buzzer and buttons
What it does today
- Lidar mapping (SLAM), one map per robot
- Autonomous navigation (Nav2) with obstacle avoidance
- Autonomous exploration of an unknown area
- Odometry computed on the board, fused with the gyroscope (EKF)
- Inertial safety: stops when tipped over, blocked, slipping or hit
- Emergency stop on the robot, fleet stop from the gamepad
- One gamepad for up to four robots, or the keyboard
- Fleet commands: battery, clock, diagnostics, which robots answer
- Calibration tools: odometry (UMBmark), gyroscope, lidar mounting
- Gazebo simulation of up to four robots, with the same commands
- Up to 19 robots on one network, each in its own namespace
Build your own
Anyone can build it
RatBot is not a product to buy: it is a robot to build, and everything needed to make it work properly is provided.
Open files
Electronics (KiCad), chassis parts ready to laser-cut, firmware and ROS 2 software, all published: nothing is held back.
A step-by-step bring-up
From a blank board to a robot on the network, one checklist, each step explained and checked.
Calibration tools included
The parameters platforms rarely give — the real radius of each wheel, the effective wheelbase, the gyroscope scale — are measured on your own robot by the tools provided.
Architecture
Three machines, one role each
Each level does what it does best: real time on the board, ROS 2 on the robot, heavy computation on the PC.
The PC
Mapping, navigation, RViz, simulation, and driving with the gamepad.
The Raspberry Pi, on the robot
ROS 2: lidar, sensor fusion, command chain, status signals. It starts by itself when the robot is switched on.
The robot board (STM32H750)
FreeRTOS and micro-ROS: servo traction, odometry, safety, clock and logbook, linked to the Raspberry Pi by a 3 Mbaud serial link.
Measurements
Measured, not assumed
Every setting of the robot comes from a measurement, recorded in a lab notebook. Some results on the first robot:
- −34 %Systematic odometry error after UMBmark calibration: from 86.7 to 57.6 mm on a 1.3 m square.
- 2.3 cmArrival error after a square with odometry fused with the calibrated gyroscope (4.2 cm with the wheels alone).
- 0.8°Difference between the turn seen by the lidar and the turn counted by the robot, over 90°.
- 1.007Gyroscope scale, measured then corrected: it counted 1.007 turns for one real turn.
Gallery
In pictures
The first RatBot 
Two robots in simulation, each with its screen 
The robot board, designed in the lab 
The board, reverse side
Build, understand and extend RatBot
Start with the user guide, dive into the electronics, firmware and ROS 2 internals, or learn step by step with the labs: for those, the board alone is enough, no robot needed.