AI-DFS

AI-DroneFireSense (AI-DFS) is a safety-critical AI vision application that integrates drone-based video streaming with real-time smoke and fire detection over advanced 5G infrastructure, validated within the 6G-PATH testbeds. The project addresses the growing threat of wildfires by delivering low-latency, high-reliability monitoring capable of supporting rapid intervention and reducing disaster response times.

Drones capture high-resolution video streams and transmit them via RTSP to two alternative processing environments:

• A 5G edge server for ultra-low-latency inference, and
• A5G cloud server for centralized, scalable processing.

In both scenarios, AI models (e.g., lightweight CNN/YOLO-based smoke and fire detectors) perform real-time inference, overlay bounding boxes, and forward annotated streams to a Command Center dashboard. Network KPIs—including latency, jitter, throughput, packet loss, and frame drops—are continuously monitored. In parallel, computational metrics such as inference latency distribution, resource utilization, and energy efficiency per inference are evaluated.

The system is benchmarked against an existing 4G baseline implementation (TRL 6), enabling a direct quantitative comparison between 4G and 5G performance under single-stream and multi-stream stress conditions. Additional experiments include concurrent execution of multiple AI models to assess edge and cloud scalability and resilience.

The expected outcome is a validated operational prototype demonstrating the advantages of 5G for real-time AI video workloads. Beyond wildfire detection, the modular architecture supports extensibility to Smart Cities and Farming applications, positioning AI-DroneFireSense as a scalable blueprint for beyond-5G, AI-driven public safety and environmental resilience services.