Comau and NVIDIA train industrial robots in a virtual environment before they reach the shop floor. What is a digital twin and why does it matter?
On 26 October 2026, Italian robot maker Comau announced a collaboration with NVIDIA that makes robot simulation a core stage of every project. The idea is simple: the robot learns, tests and makes its mistakes in a virtual factory, and arrives on the real shop floor already prepared.
What the Partnership Includes
- NVIDIA Jetson – an embedded computing module that will power Comau's next generation of controllers.
- NVIDIA Isaac Sim – a physically accurate simulation environment that will host Comau's entire portfolio of industrial, collaborative and mobile robots.
- NVIDIA Isaac Lab – training AI control policies, for example using reinforcement learning for palletizing and depalletizing.
- Comau's MI.RA vision system is trained with AI models.
- Ready-made 3D models (SimReady OpenUSD) of the robots are published openly on Hugging Face for other developers to use.
"Physical AI brings a new level of intelligence and adaptability to industrial robotics," says Paolo Guglielmini, NVIDIA Vice President for EMEA.
Why a Digital Twin Saves Time and Money
The most expensive stage of robotization is often not the robot itself but stopping the line during commissioning. A digital twin lets you check reach, cycle time and collision risk before a single bolt is installed.
- Shorter downtime when launching a new cell.
- Verified throughput and cycle time already at the quotation stage.
- Faster changeover for a new part or batch.
The Trend: A New Task in Minutes, Not Weeks
Comau is not alone. At TechCrunch Disrupt 2026, the startup Industrial Next, founded by former Tesla engineers, showed robotic stations that can be reconfigured in minutes without classic programming – new tasks are defined through simulation and demonstration. The company is also a partner of the BMW Group Technology Office. The direction is clear: robots are becoming flexible and easy to train.
Robot Simulation with Artec Robotics
ROKAE robots are programmed and simulated offline, and for standard tasks the collaborative models can also be taught by hand guiding (drag teaching). At the Artec Robotics training center in Dimitrovgrad we show live how a robotic cell is designed and verified before it reaches you.
Want to see a simulation of your process? Send us a drawing of your part and we will propose a solution.
Read more: The Era of Physical AI and Why Python Is a Good Foundation in Industrial Robotics.