IMOCO4.E
"Intelligent Motion Control under Industry 4.E"
Status: Concluded, 2022
Tech Stack: C++, Python, PyTorch
The broad IMOCO4.E challenge was to bridge the gap between the latest research results and best industrial practices in digital twins, AI, and advanced mechatronic motion control systems. IMOCO4.E strove to create solid and reliable knowledge for optimizing machines and production lines throughout their entire lifecycle. Software and hardware building blocks (BBs), distributed from edge to cloud and featuring standardized interfaces, were developed to deliver a complete IMOCO4.E reference framework. These building blocks embedded the latest advances from the academic community and could be further enhanced with new research results in the future. The project delivered a flexible, scalable, future-proof, and fully functional product architecture to be exploited in industry for high-performance motion control applications, with several overlaps in the health, mobility, and supply chain management domains.
Funding
H2020 ESCEL Joint Undertaking under grant agreement No.101007311
Contribution
In the IMOCO4.E project, the robotics Group continued to research and develop intelligent industrial robot systems. The first challenge was to demonstrate the ability of an adaptable robot to operate on real production lines, where it had to handle objects of different types and sizes from a random pile and place them into appropriately sized sockets. The second challenge was to enable the robot to be quickly retrained by anyone to work with previously unseen objects. To achieve this, the Robotics Group explored reinforcement learning approaches, AI training in simulation environments, and methods for transferring learned control policies from simulators to real-world environments (sim2real).