safeOS: A self-healing AI solution for real-time safety monitoring using a Meta-OS paradigm

safeOS has been selected under the aerOS 2nd Open Call. It addresses the challenge of ensuring safety compliance in dynamic, high-risk construction environments, where manual checks, with limited use of Computer Vision (CV) systems, lack real-time monitoring and cannot adapt to the evolving site conditions.

safeOS provides a real-time monitoring portal through AI-powered Computer Vision (CV) to detect safety violations, such as not wearing Personal Protective Equipment (PPE) like helmets and vests, and entering forbidden areas across construction sites. Through a real-time analytics dashboard, safeOS can offer immediate feedback to safety managers, processing of live video streams, and instant alerts for violations, enhancing real-time decision-making and enabling rapid responses to potential hazards.

A hybrid edge–cloud approach is adopted, offloading resource-intensive tasks to High-Performance Computing (HPC) infrastructure, while maintaining low-latency decision-making at the edge. The solution integrates aerOS edge-to-cloud technologies, leading to a self-healing solution that facilitates automated orchestration and resource allocation, with the ability to adapt to dynamic conditions of sites, including cameras losing connection, blocked views, and evolving site layouts. safeOS incorporates aerOS-enabled edge devices for local, on-site processing, achieving efficient real-time data analysis from multiple cameras. By demonstrating the system in real-world worker safety environments, safeOS validates its interoperability with diverse edge devices while ensuring compliance with EU regulations such as GDPR.