NVIDIA Unveils Comprehensive Safety System for Humanoid Robots
The company's new Halos for Robotics aims to establish a standardized safety architecture for physical AI operating alongside humans.
NVIDIA has introduced Halos for Robotics, a new system designed to be the industry's first full-stack safety solution for robots and physical artificial intelligence intended to operate in proximity to people. This system aims to provide a unified safety architecture, integrating AI compute, safety software, sensor data, and validation tools to enhance the safe deployment of robots in industrial environments.
The development addresses the increasing need for safety as robots move into dynamic workplaces such as factories, warehouses, and logistics operations. These environments often involve human workers, moving equipment, and other robots, creating complex scenarios where robots must reliably sense their surroundings, make rapid decisions, and act safely.
NVIDIA Halos for Robotics draws on extensive engineering experience from the development of autonomous vehicle safety systems. It connects crucial layers of robotic safety, including AI compute through NVIDIA IGX Thor and sensor data via NVIDIA Holoscan Sensor Bridge. The system also includes NVIDIA Halos OS, which provides the software stack for safety functions, and the Halos Outside-In Safety Blueprint, which uses external cameras and AI to help robots adapt their behavior in real-time.
Agility Robotics is the first company to integrate NVIDIA Halos for Robotics into its Digit humanoid robot, which is designed for industrial tasks in logistics, manufacturing, and warehousing. Customers for Agility's robot include Amazon, GXO, Schaeffler, and Toyota Motor Manufacturing Canada. Agility is working with NVIDIA to embed NVIDIA IGX Thor and Halos Core into Digit's safety systems, aiming to meet standards such as IEC 61508, ISO 13849, and ISO/IEC TR 5469 before final certification.
"Safety has to be built into the robot and validated across the entire system," said Peggy Johnson, CEO of Agility. "Partnering with NVIDIA to implement and optimize the Halos for Robotics system extends our leadership in responsible automation, which is a non-negotiable requirement for bringing humanoids safely into industrial workflows."
The NVIDIA Halos AI Systems Inspection Lab, part of the Halos for Robotics offering, is described as the world's first ANSI National Accreditation Board-accredited program for physical AI functional and AI safety. This lab assists partners in preparing their Halos integrations for third-party certification from various industry bodies.
NVIDIA has also established an ecosystem of partners across software, embedded systems, sensors, silicon, industrial applications, and certification. This network includes companies like Acontis, Amazon FreeRTOS, QNX, Advantech, Infineon, NXP, SICK, STMicroelectronics, Texas Instruments, and FORT Robotics, contributing to the development and validation of safe robotic systems.
NVIDIA Halos Core for NVIDIA IGX is currently available in early access for registered developers, supporting Linux and Linux plus QNX configurations. The open-source NVIDIA Halos Outside-In Safety Blueprint is also in early access on GitHub. The system is designed for developers, robotics companies, and industrial operations looking to implement physical AI in real-world work environments.