Real-Time Mobile Crowd Sensing Model for Remote Detection of Flying UAVs

Abstract

Conformance monitoring is crucial for ensuring the safe operation of uncrewed aerial vehicles (UAVs). It aims to alert relevant parties to any deviations from authorized flight plans. The European Union Aviation Safety Agency (EASA) mandates conformance monitoring and suggests integrating it into UAV traffic management (UTM) systems. Although traditional monitoring systems, such as wireless sensor networks, can serve this purpose, the associated expenses for installation and maintenance make them impractical for large-scale implementations. To tackle this challenge, we propose a monitoring system that capitalizes on UAV remote identification (RID) technology and mobile crowd sensing. RID is becoming a global regulatory requirement. It enables ground observers to identify drones using standard mobile devices. Our solution collects RID data by gathering reports from these observers to assess UAV operations in real-time. A significant component of this approach is determining the optimal number of reports that should be considered to allow reliable and quick evaluation. While processing more reports can enhance the evaluation accuracy, it increases the computational demands and may compromise the system’s real-time performance. Moreover, in a reward-based system, processing more reports incurs higher costs. Therefore, our system uses a mechanism that adjusts the number of received and processed reports based on airspace conditions and the crowd density in the area of interest. In addition, our approach incorporates signal quality metrics - path loss, shadowing, and received signal strength (RSS) - along with distance considerations in the activation process of ground observers for RID message forwarding. This comprehensive consideration of both signal integrity and spatial proximity significantly improves detection and monitoring precision of the UAVs. We validated the proposed model through Monte Carlo simulations. Results indicate that, compared to a naive model, our approach outperforms in terms of received RIDs, paid rewards, and real-time performance.