MA-QoS ICN: Mobility-aware QoS ICN framework with Independent Q-Learning for UAVs in Disaster Scenarios

초록

Unmanned Aerial Vehicles (UAVs) are critical for maintaining communication in disaster-affected areas, where terrestrial infrastructure is often unavailable and network connectivity becomes highly fragile. To address this challenge, the paper proposes decentralized data forwarding and trajectory alignment in UAV-assisted networks using an information-centric paradigm. Specifically, we propose a Mobility-Aware QoS Information-Centric Networking (MA-QoS ICN) framework, wherein each UAV embeds Quality-of-Service information (QoS info), including priority, deadline, and packet lifetime—as well as UAV Mobility State Information (UMSI), such as velocity, direction, position, and battery level, directly into the ICN packet headers. To model the dynamic and uncertain environment, we formulate the UAV coordination problem as a stochastic game for maximizing long-term content delivery performance, where each UAV acts as a learning agent and makes forwarding and trajectory alignment decisions based on its local state. We then develop a multi-agent reinforcement learning (MARL) framework in which each UAV independently discovers its optimal strategy through local observations and experience, without requiring centralized control. In particular, we adopt an agent-independent learning method based on Independent Q-Learning, leveraging ICN’s packet-level handling for implicit coordination. This design significantly reduces communication overhead while preserving coordination among agents. Simulation results demonstrate that the proposed framework achieves substantial improvements in delivery reliability, latency reduction, and retransmission minimization, along with enhanced scalability under diverse disaster conditions. © 2014 IEEE.

키워드