Elif Özaksu
Corporate
- Thread Author
- #1
A groundbreaking development has occurred in the world of industrial automation! NVIDIA has introduced Halos, the industry's first full-stack safety platform designed for robotic systems. This innovative system aims to accelerate the safe deployment of physical AI and humanoid robots by combining artificial intelligence (AI) computing power, software, and certification tools.
─────────────────────────
💡 Halos: A Holistic Approach to Robotic Safety
NVIDIA Halos for Robotics offers a full-stack, open safety architecture that integrates AI computing and sensor connectivity. This solution has been specifically developed for autonomous machines and humanoid platforms operating in dynamic environments such as factories, warehouses, and logistics. Halos will play a critical role in these areas where human-robot interaction requires continuous spatial awareness and rapid decision-making.
─────────────────────────
⚙️ Transferring Autonomous Vehicle Safety to Industry
Leveraging over 18,600 engineering years of autonomous vehicle safety development experience, NVIDIA provides a unified framework for building, validating, and deploying robotic systems. This platform eliminates the need for developers to design separate safety systems for computing, sensors, software, and certification. Deepu Talla, Vice President of Robotics and Edge AI at NVIDIA, states that this will enable safer robots to be integrated into industrial operations.
─────────────────────────
⚡ Hardware and Software Architecture for Real-Time Computing
The system utilizes the NVIDIA IGX Thor platform and the Holoscan Sensor Bridge to provide industrial-grade AI processing capabilities and secure sensor integration. This hardware, with its built-in safety mechanisms and advanced error management, can meet stringent functional safety standards such as IEC 61500 and ISO 13849.
The software layer, Halos OS, includes Halos Core, which manages safety-critical operating functions. It also features an application framework called Halos Outside-In Safety Blueprint, which connects external facility cameras and AI agents to dynamically control robot behavior based on real-time environmental data.
─────────────────────────
🤝 Ecosystem Integration and Third-Party Certification
Humanoid robotics developer Agility is integrating the IGX Thor compute module and Halos Core software into its Digit robot's proprietary human detection system. This integration supports safe operation in manufacturing and warehouse logistics for customers such as Amazon, GXO, Schaeffler, and Toyota Motor Manufacturing Canada. Agility CEO Peggy Johnson emphasizes that the Halos system establishes a requirement for the inclusion of humanoid robots in industrial workflows by validating safety across the entire stack.
To facilitate regulatory compliance, the NVIDIA Halos AI Systems Audit Lab operates as a program accredited by the ANSI National Accreditation Board. This lab evaluates robotic AI systems against recognized safety requirements, preparing them for certification by independent organizations such as TÜV Rheinland, UL Solutions, TÜV SÜD, exida, SGS, and CertX.
The broad Halos ecosystem includes software support from Acontis, FreeRTOS, and QNX, along with embedded systems from Advantech and NexCobot. Sensor and silicon integration is provided by Infineon, NXP Semiconductors, STMicroelectronics, and Texas Instruments. Industrial applications utilizing the Outside-In Safety Blueprint are being developed by FORT Robotics, Inventec, KION Group, Lyte AI, and Neurealm.
─────────────────────────
🌐 Accessing the Halos Architecture
NVIDIA Halos Core for NVIDIA IGX is available in early access for registered developers using Linux and Linux plus QNX OS for Safety 8.0 configurations. Additionally, the open-source NVIDIA Halos Outside-In Safety Blueprint is also accessible in early access via GitHub.
─────────────────────────
📊 The Halos Difference Against Traditional Approaches
The traditional industrial robotic safety market typically relies on separate functional safety controllers and programmable logic controllers (PLCs) from manufacturers like SICK, Pilz, and Rockwell Automation. These "inside-out" safety mechanisms often use embedded sensors and impose limited operational constraints, requiring system integrators to design and validate complex, custom safety architectures for each deployment.
In contrast, the NVIDIA Halos for Robotics system offers a unified, full-stack platform that combines high-performance AI processing with comprehensive, pre-validated safety software through the IGX Thor platform. By including the Outside-In Safety Blueprint, the NVIDIA architecture dynamically modulates robot efficiency using external infrastructure sensors and visual AI agents, providing a broader mechanism for situational awareness compared to standard localized safety scanners. The platform also facilitates compliance with emerging AI safety standards like ISO/IEC TR 5469, distinguishing itself from hardware-centric localized safety components.
NVIDIA Halos is redefining robotic safety, taking a significant step towards shaping the future of industrial automation. This innovative platform paves the way for robots to work more safely and efficiently alongside humans.
