TechPrufer
Sign inGet started

Blogs›NVIDIA›Full-Stack Functional Safety System for Robotics

Full-Stack Functional Safety System for Robotics

Full-Stack Functional Safety System for Robotics

2
posts
2026

This feature thread tracks the evolution of full-stack functional safety systems for robotics, building upon NVIDIA's extensive experience in autonomous vehicle safety. It encompasses the integration of specialized hardware like NVIDIA IGX Thor with a comprehensive safety OS (Halos OS) to provide standards-compliant, extensible safety solutions for industrial robots, humanoids, and AMRs. Key advancements include the development of safety islands, high diagnostic coverage, and mechanisms for free-running simulation environments. This post introduces NVIDIA Alpamayo and its AlpaGym framework for closed-loop reinforcement learning post-training of autonomous vehicle models, enabling policies to learn from the consequences of their own actions in simulation, thereby bridging the gap between open-loop training and closed-loop deployment.

2026

Inside NVIDIA Halos for Robotics: A Full-Stack Functional Safety System for Physical AI | NVIDIA Technical Blog

6/22/2026

This post introduces NVIDIA Halos for Robotics, a new full-stack functional safety system for physical AI. It details the extension of NVIDIA's autonomous vehicle safety stack to robotics, leveraging the NVIDIA IGX Thor platform and Halos OS. The post highlights the hardware safety features of IGX Thor, including its IEC 61508 SIL 3 capable Safety Island, high diagnostic coverage, and support for diversity and redundancy. It also introduces the NVIDIA Holoscan Sensor Bridge for extending the safety chain to sensors and actuators, and the Halos AI Systems Inspection Lab for certification.

How to Post-Train Autonomous Vehicle Models in Closed-Loop with NVIDIA Alpamayo | NVIDIA Technical Blog

6/1/2026

This post introduces NVIDIA Alpamayo and its AlpaGym framework for closed-loop reinforcement learning post-training of autonomous vehicle models. It details the installation and configuration of AlpaGym, the definition of closed-loop rewards, and the process of launching closed-loop training. The framework connects AlpaSim simulator rollouts to the policy training loop, allowing models to learn from their own actions in simulation, which is crucial for addressing compounding errors in autonomous driving. The post also outlines how to export post-trained checkpoints for downstream use.