Cadence, a major supplier of software used to design advanced computing chips, is working with Nvidia to integrate physics engines that model how real-world materials behave with Nvidia AI models designed to train robots in computer simulations.
The Partnership
According to Tech-Economic Times, Cadence and Nvidia are collaborating to integrate Cadence physics engines—software that predicts how real-world materials interact—with Nvidia AI models designed to train robots inside computer simulations. The integration combines physics-based modeling with AI training: the physics engine supplies environment behavior, while AI models learn robot control policies within that simulated environment.
Robotics training depends on how accurately simulation represents the physical world. The integration aims to merge a physics engine’s material interaction predictions with AI training runs that occur in simulation.
Why Physics Engines and AI Models Work Together
A key challenge in robotics is that simulation is only useful for training if it captures the interactions robots will face in the real world. Cadence’s physics engines are designed to predict how real-world materials interact, allowing the simulation environment to incorporate material behavior rather than relying solely on simplified assumptions.
Nvidia’s AI models are designed to train robots inside computer simulations, meaning the AI training loop occurs in a controlled simulated setting where robots can be evaluated repeatedly.
The integration suggests a workflow where:
(1) the physics engine estimates material interactions in the simulated world, and (2) the AI model uses those simulated outcomes to learn robot behavior. This represents an end-to-end approach combining physics and AI rather than standalone efforts.
Implications for Robotics Simulation
The integration of physics engines directly into the simulation environment used by AI models could reduce the gap between simulated training conditions and real-world material behavior. Potential implications include:
• More physically grounded training: Physics engines modeling material interactions during training could result in simulated experience that reflects physical behavior more closely than simulations without such predictions.
• Consistent environment and learning: The integration suggests that the environment model (physics) and the learning system (AI models) are being treated as linked components, which could improve consistency between what the AI learns and the simulated dynamics it experiences.
Cadence’s Role in the AI Stack
Cadence is one of the major suppliers of software used in designing advanced computing chips. This positions Cadence’s expertise in the hardware design and simulation ecosystem. In this partnership, Cadence contributes physics engines that predict material interactions, suggesting a trend in which companies with roots in computing design and simulation bring their modeling capabilities into AI training workflows for robotics.
Nvidia provides AI models for training robots in simulation. Together, the partnership highlights a division of labor: physics and environment modeling on one side, and AI training systems on the other.
Source: Tech-Economic Times