Shaun Kirby

Simulation-based reinforcement learning advances Physical AI by enabling rapid policy learning for complex tasks. NVIDIA Isaac Lab offers a reinforcement learning environment to train and test policies in realistic virtual environments using physics-based simulation.

Physical AI is changing robotics by enabling systems to interact intelligently with the world, but development requires extensive compute, simulation, and deployment resources.

The future of robotics lies in physical AI—robots capable of autonomously performing intricate tasks in the real world. To meet this grand challenge, accelerating the development of AI-powered robots is essential.

Simulation-based reinforcement learning (RL) has benefited the field of robotics, helping manufacturers automate complex tasks. With generative AI, robotic simulation has accelerated the development and training of models.

Join this session to learn how Panasonic Energy of North America (PENA) implemented an Industrial Data Fabric solution to improve the efficiency of their battery manufacturing using AWS Professional Services and Palantir Foundry.

Learn how to train robots quickly and easily with containers in the cloud. Parallelize training for multiple robots on multiple multi-GPU nodes to expedite results.

Training robots requires high-fidelity simulation, a multitude of scenes with domain randomization, and numerous simulation runs to develop robots that perform well in the real world. Tools like NVIDIA Isaac Sim, NVIDIA Omniverse Replicator, and NVIDIA TAO Toolkit provide realistic simulation environments but take time and effort to set up.