In the context of the ARISE project, CEA List is leading the development of an advanced interactive simulation framework to support the design and testing of robotic systems and actuators. This simulation environment enables partners to evaluate and optimise their technologies in realistic conditions before physical deployment, ensuring greater efficiency, accuracy, and integration across the platform.

 Why Simulation Matters in Robotics Design

Before real-world deployment, robots must undergo extensive testing in digital environments. Simulations allow engineers to validate complex mechanical behavior, assess system interactions, and optimise designs—all without the cost or risk of physical prototyping.

To create a shared and efficient development environment, CEA List identified Nvidia Omniverse as the most suitable 3D software for the ARISE project. Omniverse offers seamless compatibility with ROS and ROS 2, both widely used frameworks in robotics.

Bringing Physics to Life: XDE Physics Meets Omniverse

CEA List has successfully integrated its powerful physics engine, XDE Physics, into Nvidia Omniverse. XDE excels at simulating multi-body dynamics, intermittent contacts, and complex interactions such as hydraulic behaviors—all of which are essential for validating the advanced robotic systems envisioned in ARISE.

This integration provides partners with a realistic digital twin environment to test robot performance under varying conditions. The integration process is now complete, and CEA List is currently awaiting a patch from Nvidia to resolve final stability issues before full deployment.

 Exploring Neural Networks for Soft Robotics

In parallel, CEA List is pushing the boundaries of real-time simulation by exploring how neural networks can be used to model soft robotic grippers and actuators—components known for their flexibility and complexity.

To support this work, a research paper was recently published presenting the team’s first findings:

Louen Pottier, Anders Thorin, Francisco Chinesta
Latent-Energy-Based NNs: An interpretable neural network architecture for model-order reduction of nonlinear statics in solid mechanics.
 Journal of the Mechanics and Physics of Solids, 2024, Article No. 105953. 10.1016/j.jmps.2024.105953 | HAL: hal-04737657

This innovative approach enables real-time simulation of soft structures while preserving interpretability and accuracy—key factors for the responsive robotic systems ARISE aims to develop.

 What’s Next

With Omniverse and XDE Physics integration nearly complete and research on AI-enhanced simulation progressing steadily, CEA List continues to lay a solid foundation for the design and validation of the next generation of robotic systems. Stay tuned for more updates as we test these tools in real-world use cases across solar farms and hydroponic greenhouses!