NVIDIA has introduced Halos for Robotics, a safety platform designed for robotics and physical artificial intelligence systems that combines AI computing infrastructure, software, sensor connectivity and safety validation tools within a single architecture. The platform extends technologies developed for autonomous vehicle safety to robotics applications, providing a framework intended to support the development, validation and deployment of autonomous systems operating alongside people in industrial environments.
NVIDIA said the system is designed for robots that use AI models, distributed sensors and accelerated computing to perceive, make decisions and perform actions in real-world settings. Agility, a developer of humanoid robots, is the first company to incorporate elements of Halos for Robotics into its safety system. The company plans to use the technology in its Digit humanoid robot, which is deployed in logistics, manufacturing and warehouse environments. Agility’s customers include Amazon, GXO, Schaeffler and Toyota Motor Manufacturing Canada.
According to NVIDIA, Halos for Robotics is built around three primary components. NVIDIA IGX Thor and Holoscan Sensor Bridge provide computing and sensor connectivity capabilities intended for industrial robotics applications. Halos OS serves as the software layer for safety-related operating functions and applications, including Halos Core and the Halos Outside-In Safety Blueprint, which uses external cameras and AI agents to influence robot behavior in industrial environments. The third component, the NVIDIA Halos AI Systems Inspection Lab, is intended to help companies prepare systems for third-party certification.
NVIDIA said the inspection lab is accredited by the ANSI National Accreditation Board for functional and AI safety assessments related to physical AI systems. The company stated that the lab works with certification organizations including TÜV Rheinland, UL Solutions, TÜV SÜD, exida, SGS and CertX.
Laurie E. Locascio, president and chief executive officer of ANSI, said the accreditation confirms the inspection program’s ability to evaluate robotic AI systems against recognized safety requirements and provide a basis for certification activities. Agility plans to integrate NVIDIA IGX Thor and Halos Core into its safe human detection system for Digit. The companies also intend to use the inspection lab to evaluate safety-related software, AI components and cybersecurity protections against standards including IEC 61508, ISO 13849 and ISO/IEC TR 5469 before pursuing third-party certification.
NVIDIA said the broader Halos for Robotics ecosystem includes software providers Acontis, FreeRTOS and QNX; embedded systems companies Advantech and NexCobot; semiconductor suppliers Infineon, NXP Semiconductors, STMicroelectronics and Texas Instruments; and industrial technology companies FORT Robotics, Inventec, KION Group, Lyte AI and Neurealm. TÜV Rheinland is conducting inspections related to certification readiness for NVIDIA IGX Thor, Halos OS and Holoscan Sensor Bridge, while TÜV SÜD has inspected Thor SoC and Halos Core under ISO 26262 processes.
The Halos AI Systems Inspection Lab includes more than 40 organizations spanning manufacturers, certification bodies and safety technology providers. Halos Core for NVIDIA IGX is available through an early-access program for registered developers in Linux and Linux with QNX OS for Safety 8.0 configurations. The open-source Halos Outside-In Safety Blueprint is also available in early access through GitHub.
