NVIDIA Halos for Robotics Full Stack Safety System Sets the Platform for Large Scale Autonomous Machine Deployment and the Physical AI Era Through Taiwanese Hardware Collaborations
NVIDIA launched NVIDIA Halos for Robotics full stack safety system, providing an end to end safety solution to enable autonomous machines to evolve from concept validation to large scale commercial deployment. This safety release is setting the platform for the physical AI era as envisioned by CEO Jensen Huang at the CES show. To build an industrial ecosystem, NVIDIA has also collaborated with top Taiwanese hardware maker companies including Advantech, Inventec and NexCOBOT, 100 percent owned by Nexcom.
The official NVIDIA launch story tells us that maturing these robotics area is just like the growth of generative language models. To grow these machines safely, it needs a unified structure fastening oil, operating systems, sensor outputs and accreditation logics. The recently released Halos system draws on 18600 engineering years worth of functional safety data originally built up over the course of NVIDIA's autonomous self driving vehicle programs.
The safety components work within hardware, software, and validation levels. Hardware: The system leverages the NVIDIA IGX Thor and the Holoscan Sensor Bridge for industrial grade AI processing in conjunction with sensor synchronization. Software: Based on Halos OS and Halos Core, the system uses external cameras and AI agent frameworks to expand its visual perception, allowing the robot to perceive movements in the environment and modify its trajectory accordingly to avoid potential hazards.
The collaboration network divides tasks into two parts: embedded systems design and physical industrial application. Advantech and NexCOBOT serve as embedded systems partners by developing embedded systems toward a secure computing platform based on NVIDIA IGX architecture. Inventec will be the industrial applications partner by deploying secure autonomous systems in factories and logistics warehouses.
In this world, NexCOBOT has designed its own safety controller, the RCB400 T20 05. NexCOBOT combined NVIDIA IGX T5000 platform with its own real time operating system (NexRTOS) and EtherCAT based motion control (NexECM) to enable a shop floor manager to co locate AI inference, physical motion control and functional safety on 1 single computing platform, which is a reduction in hardware complexity. Advantech is also increasing its industrial edge computing portfolio for new secured robotic needs.
Market analysts said the advance of Halos marks the completion of the chipmaker's robotics workflow. NVIDIA unwraps the complete robotics workflow from the GR00T foundation model, Cosmos world simulation model, Isaac Sim virtual training environment and the Jetson Thor compute engine to the Halos safety verification system.
