Accelerating Robot Development With AI Agents and NVIDIA Isaac
A NVIDIA presenter describes how developers can use GPT-6 Astra and NVIDIA Isaac tools to build robot simulations, train skills, and move them into real-world deployment.
In this talk
- 0:02
Prompting a robot training world
Robot development is challenging, and creating simulation environments requires time and expertise. The presenter says developers can describe the assets, setting, and task in a prompt, with data, to have Astra build a training world.
- 0:29
Building and tuning the simulation
Astra uses NVIDIA open libraries and models to create the environment. It replays real-world motion and adjusts physics to prepare the simulation for training.
- 0:48
Training and evaluating skills
The workflow uses Isaac Lab and Newton to train and improve robot skills, then Isaac Lab Arena to evaluate them at scale. The presenter says these tools run locally on an NVIDIA DGX Station to support faster iteration.
- 1:09
Moving skills into deployment
The presenter describes training tasks for settings such as homes, hospitals, stores, and factories, then transferring learned skills from simulation to physical robots. AgileX is introduced as a reference platform using NVIDIA Isaac GR00T and Jetson Thor to help developers train and deploy skills.
- 1:49
Broadening access to robot development
The presenter says Astra and NVIDIA Isaac tools let individual developers carry out work that previously required a specialist team. The workflow spans simulation creation, skill training, evaluation, and deployment.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.