Robeon
Turnkey factory robotics: a humanoid, a warehouse platform and a manipulator with in-house electronics and simulation-trained AI.
Robeon is full-cycle robotics: not a single robot but turnkey automation of a factory operation. In-house: AI, computer vision and electronics — the RCB-1 real-time controller on CAN. Actuators are off-the-shelf; everything else is our own design.
Robots are trained in simulation: MuJoCo and reinforcement learning, 4,096 parallel worlds and 300M steps in about 20 minutes. All products are at the digital-prototype stage: CAD, simulation and trained motion policies; next up is the actuator test bench and the first hardware assemblies.
- MuJoCo
- RL · PPO
- Python
- YOLO
- Jetson Orin NX
- STM32H743
- FreeRTOS
- CAN FD
- KiCad
- CAD · STEP
- .NET 8
- Humanoid robot
- Warehouse platform
- RA-6 manipulator
- Computer vision
- Suntint tinting controller
- CAN bus electronics
How it works
- 01
Analyse the factory operation and find what pays off to automate.
- 02
Design the robot in parametric CAD and build a digital twin.
- 03
Train motion and vision in simulation across thousands of parallel worlds.
- 04
Deploy policies to the RCB-1 controller and the onboard Jetson computer.
Features
165–175 cm tall, 28 joints, up to 25 kg carried close to the body. Built for paint factories: moves cans and buckets.
Two versions: Mover transporter and Handler with an arm and a 1.8 m gripper lift, Livox lidar.
6 axes, 900 mm reach, 5 kg payload, 48 V, 7075 aluminium and printed covers.
STM32H743, CAN FD, Ethernet, hardware E-stop under 50 ms, 1 kHz control loop.
An 8-class YOLO detector trained on synthetic data.
.NET 8 software for a paint-tinting station with a 16-channel simulator.
Let'sbuildsomething strong
Software, automation, installations, technical production for events — drop me a line.