NexCOBOT : Part of the NVIDIA Halos for Robotics Safety Ecosystem
Part of NVIDIA's Robotics Safety Ecosystem
NEXCOBOT is part of the NVIDIA Halos for Robotics ecosystem, the industry's first full-stack safety system for physical AI. Built around the NVIDIA IGX T5000 system-on-module, the MARS400 T20-05 combines Blackwell-class AI compute with NEXCOBOT's own functional-safety architecture, giving humanoid and quadruped robotics developers a safety-designed foundation for machines that sense, reason, and move in dynamic environments.
Helping Accelerate AI Safety Certification
NEXCOBOT is also a member of the NVIDIA Halos AI Systems Inspection Lab, the world's first ANAB-accredited inspection body for functional and AI safety in physical AI, and among the first companies to join. Working alongside NVIDIA, NEXCOBOT helps streamline the path customers take toward AI safety certification, drawing on standards developed specifically for AI-based systems — including ISO/IEC TR 5469:2024 and the forthcoming ISO/IEC TS 22440-1 — to help customers build AI platforms aligned with functional-safety requirements.
A Layered Safety Architecture, Isolated by Design
- Layer 1 – Hardware:
The NVIDIA IGX T5000 SOM contributes a Functional Safety Island (FSI) alongside its CPU and GPU, sharing the primary carrier board. An external Safety MCU (EBK-210) sits on a separate carrier board — physically distinct hardware from the AI compute path.
- Layer 2 – OS and BSP:
The FSI and the Safety MCU each run their own dedicated BSP and real-time OS, independent of the AI/motion-control software. On the AI side, an NVIDIA Hypervisor partitions QNX — running in its own VM — from IGX OS, keeping safety-relevant and general-purpose workloads separated even though they share the same CPU and GPU.
- Layer 3 – Communication:
Safety-relevant traffic moves over a dedicated Edge Safety Link running an FSoE (Fail Safe over EtherCAT) master protocol stack, separate from the EtherCAT master and Ethernet used for motion control and AI data.
- Layer 4 – Applications:
Safety applications run on the FSI and Safety MCU stacks; motion control and AI applications run on the IGX OS side — sharing the same platform without sharing the same failure domain.
This isolation — carried from silicon, to software, to network, to application — is what allows a single MARS400 T20-05 to handle both AI inference and safety-relevant control without a separate safety controller bolted onto the system.
Key Specifications
- Based on NVIDIA® IGX™ T5000
- Designed to SIL2 / PLd principles
- Halos OS compatible
- NexCOBOT SW Packages: NexECM (EtherCAT Master), NexRTOS (Real-Time Linux), FSoE Master
- Rich I/O interfaces
For more information, visit NVIDIA Halos for Robotics / NVIDIA Halos AI Systems Inspection Lab
