Key Takeaway
Built on NVIDIA IGX Thor and the NVIDIA Halos Outside-In Safety Blueprint, the solution will be demonstrated at the International Robot Safety Conference 2026 in Detroit, November 4
Detroit, Michigan, October 30, 2026: Neurealm, a Physical AI to Agentic AI services company, today announced that its Outside-In Safety solution for industrial robotics now runs on the Advantech MIC-735-IT, an industrial-grade edge AI platform built on NVIDIA IGX Thor. Using the NVIDIA Halos Outside-In Safety Blueprint and the IGX Thor Safety Extension Package, the solution is designed to generate a real-time hardware emergency stop (e-stop) for robotic systems operating in collaborative environments. Neurealm will demonstrate the solution with Advantech at the International Robot Safety Conference (IRSC) 2026, hosted by the Association for Advancing Automation (A3) at Huntington Place, Detroit, from November 4.
Extending robot safety beyond onboard sensors
Robots on the factory floor have traditionally relied on onboard, or inside-out, sensors to manage their immediate safety envelope. Outside-in safety adds a second layer: external infrastructure cameras monitor the shared workspace from multiple vantage points and establish virtual safety zones around the robot. As more robots work alongside people, this wider view helps address blind spots that onboard sensors alone may miss.
How the solution works
The NVIDIA Halos pipeline within the IGX Thor integrates cameras that continuously monitor the workspace to ensure collaborative safety. Software running on the Central Complex (CCPLEX) detects potentially unsafe events in the workspace and generates corresponding triggers, which are routed to the real-time Functional Safety Island (FSI). The safety software running on the FSI aggregates these triggers while simultaneously monitoring the correct operation of the CCPLEX itself, before generating hardware emergency stop (e-stop) signals to bring industrial machinery to a safe state.
"Vision AI has given manufacturers a far richer picture of what is happening around a robot. The next step is making sure that insight can trigger a dependable safety response," said Sanjay Jayakumar, President, Semiconductor, Embedded & Automotive Business, Neurealm. "By bringing our Outside-In Safety solution onto the Advantech MIC-735-IT, built on NVIDIA IGX Thor and NVIDIA Halos, we are moving this work closer to how safety needs to be deployed in real industrial environments."
“Bringing Physical AI into real-world robotics requires functional safety to be addressed at the system level. Advantech’s MIC-735-IT, powered by NVIDIA IGX Thor and designed to support Safety MCU integration, provides a robust hardware foundation for outside-in safety architectures. Combined with Advantech’s design and manufacturing services and system integration expertise, we help customers develop application-specific safety architectures and support alignment with applicable standards and derived requirements, including IEC 61508 and ISO 13849. By combining high-performance AI computing with functional-safety-oriented system design, Advantech helps enable safer human-robot collaboration in real-world environments,” said Logan Wang, Senior Product Manager of Applied Server Group, Advantech.
Building on a year of Outside-In Safety milestones
Neurealm first demonstrated its Outside-In Safety work at NVIDIA GTC 2026 in March, followed by Hannover Messe in April and Automate 2026 in June. Neurealm is an ecosystem partner of NVIDIA Halos for Robotics and has joined the NVIDIA Halos Inspection Lab, reflecting its commitment to functional safety and AI integrity.
See it at IRSC 2026
Visitors can see a live demonstration of the Outside-In Safety solution running on the Advantech MIC-735-IT at the event.
-50/MIC-735_Main20260219162932.png)
