Overcoming the Limitations of Traditional Inside-Out Safety Architecture

In traditional manufacturing setups, cobot safety architectures operate on a proximity-and-contact basis. These "inside-out" systems utilize integrated onboard sensors to trigger an emergency stop upon physical impact. However, this approach creates operational bottlenecks. With a restricted field of view, the robotic system possesses zero peripheral awareness, meaning it cannot anticipate human movement or optimize its cycle times based on proximity.

Enter the Outside-In Safety Agent

At COMPUTEX, Inventec demonstrated how to transition from traditional, contact-reliant safety to real-time situational awareness with safety agents. In a smart factory quality inspection showcase, we utilized a cobot arm for Automated Inspection Operations (AIO) integrated with enhanced spatial intelligence. Rather than relying solely on the cobot's internal impact sensors, the solution incorporates the open source  NVIDIA Halos Outside-In Safety Blueprint. This framework, part of NVIDIA Halos for Robotics functional safety system, leverages external sensor infrastructure to extend the robot's perception far beyond its own physical chassis, creating a comprehensive overview of the entire workspace.

How It Works: The Atlas Edge AI Server in Action

While cobot arm retains its native reactive safety—halting upon physical contact—the true innovation lies in the broader workspace awareness. This is achieved by utilizing external monitoring cameras to establish a tiered safety environment. When the system detects a breach of the outer "Warning Zone," it signals the robotic arm to decelerate, allowing operations to continue at a reduced speed. Should an individual progress into the inner "Critical Zone," the system triggers an immediate halt before any physical interaction can occur.

This sophisticated safety layer is powered by the Inventec Atlas Edge AI Server, powered by NVIDIA IGX Thor™ in conjunction with specialized software. The Atlas provides the industrial-grade compute performance necessary to process high-fidelity camera streams in real-time, while the software manages the visual analytics and executes the cobot’s dynamic responses.

A Framework for Scalable Manufacturing

This integration represents a foundational step toward true, dynamic human-robot collaboration on the factory floor. By deploying outside-in safety agents, manufacturers can eliminate traditional sensor blind spots, significantly minimize unnecessary downtime caused by false stops, and optimize throughput without compromising workforce safety. Inventec is committed to developing the high-performance hardware ecosystem required to make this level of physical AI safety both certifiable and globally scalable.

Driving Industry Collaboration

Inventec is looking to partner with industry leaders to co-develop and deploy enterprise-grade smart factory solutions leveraging the Inventec Atlas Edge AI Server, powered by NVIDIA IGX Thor™ and the NVIDIA Halos ecosystem. We invite technology partners and industrial manufacturers to connect with us to discuss how we can scale these capabilities across your production operations.