5G-based ground risk mitigation for UAVs: A deep reinforcement learning approach

Harchaoui, M. L., Ouahouah, S., Bekkouche, O., Bagaa, M. et Derouiche, A. (2024). 5G-based ground risk mitigation for UAVs: A deep reinforcement learning approach. Dans GLOBECOM 2024 - 2024 IEEE Global Communications Conference DOI 10.1109/GLOBECOM52923.2024.10901435.

PDF Sous embargo jusqu'au 11 mars 2026. Télécharger (433kB)

Résumé

The emergence of the Beyond Visual Line of Sight (BVLOS) operations for Unmanned Aerial Vehicles (UAVs) unlocked a wide range of new applications across various domains, such as urban transportation, package delivery, and aerial surveillance. However, due to the possibility of losing control and collisions, BVLOS operations present several risks to people on the ground. Therefore, it is crucial to minimize safety risks by flying UAVs along paths that traverse less populated areas. Nevertheless, implementing such a solution requires access to real-time data on the population density distribution across UAV operational areas. Consequently, in this paper, we harness the network exposure capabilities of 5G mobile networks, proposing a framework that integrates the UAV Traffic Management (UTM) system with the 5G Core (5GC). The proposed framework can collect real-time information about the density of mobile users in Areas of Interest (AoI), leveraging this data to estimate ground risks and subsequently devise optimized flight paths. Moreover, we propose a Deep Reinforcement Learning (DRL) solution to compute optimized flight paths. The simulation results show the efficiency of our proposed solution to achieve the designed goals in terms of reducing the experienced ground risk and total flight distance.

Type de document:	Document issu d'une conférence ou d'un atelier
Mots-clés libres:	Visualization 5G mobile communication Surveillance Soft sensors Transportation Autonomous aerial vehicles Deep reinforcement learning Real-time systems Safety Vehicle dynamics UAVs 5G Risk mitigation Service-based architecture Flight planning
Date de dépôt:	14 janv. 2026 18:15
Dernière modification:	14 janv. 2026 18:15
Version du document déposé:	Post-print (version corrigée et acceptée)
URI:	https://depot-e.uqtr.ca/id/eprint/12490

Actions (administrateurs uniquement)

Éditer la notice