Suche
Newsletter abonnieren
News
Branchenverzeichnis
Sonderhefte/Specials
Shuttle-Profis
TÜV SÜD and Accenture present the "Physical AI Safety Suite" at the NVIDIA GTC Berlin 2026
The "Physical AI Safety Suite" partially digitizes the manual inspection of equipment and installations, as well as the creation of certification packages, using digital twins and AI-supported simulations. TÜV SÜD estimates that the digital process could speed up commissioning by up to 30 percent. This not only reduces physical dependencies such as installed equipment and on-site inspections; companies can also address safety and certification issues earlier to avoid costly rework during operation.
Connected Assessments on Three Levels
The concept of the “Physical AI Safety Suite” evaluates automation on three interconnected levels, with TÜV SÜD conducting an independent assessment and assurance, while retaining responsibility for certification:
Level 1: Simulation of Industrial Plant Layouts - Where can companies safely use physical AI?
Using the NVIDIA Omniverse libraries, the open-world model platform “Cosmos,” and the open robotics framework “NVIDIA Isaac Sim,” the “Physical AI Safety Suite" virtually assesses routes, free spaces, mixed-use zones, and cooperation areas before robots are deployed or layouts are reconfigured. It then creates a structured evidence package for TÜV SÜD's review, covering requirements for operating areas, standards for hazard and risk mitigation, norms for robot integration, as well as safety aspects of control systems.
Level 2: Integration of Robot Fleets - How Does Physical AI Work Safely?
With the help of "NVIDIA Omniverse," "Cosmos," and "Isaac Sim," the "Physical AI Safety Suite" simulates robot fleets and their behavior to identify operational constraints and potential hazards before deployment, such as the possibility of employees being trapped by AMRs. During operation, the "NVIDIA Halos" safety platform and the "Edge AI Platform NVIDIA IGX Thor" are used to detect potentially dangerous situations that are outside the field of view of a single robot. AI agents developed based on the "NVIDIA Halos Outside-In Safety Blueprint" extend a robot's perception, enabling it to autonomously draw conclusions and take action using cameras installed in the facility. As a result, fleet management systems can respond earlier to improve safety and operational efficiency.
Level 3: Individual Robots - What Is Physical AI Allowed to Do?
Even devices that comply with all applicable product regulations must be tested under operating conditions intended for the specific location, such as load, speed, floor condition, and lighting. With the help of "NVIDIA Isaac Sim" and "Cosmos," the "Physical AI Safety Suite" creates and tests a variety of scenarios relevant to certification that could practically never be covered in tests under real conditions. This allows engineering teams to determine the best placement for safety sensors, for example, long before a physical prototype exists.
Translated automatically from German.
Branchenverzeichnis
