<![CDATA[Reliable communication is essential for effective disaster response; however, conventional IP-based networks often fail when the network infrastructure is damaged. Disaster communication networks need adaptive forwarding strategies that maintain reliability under rapid topology changes, various link qualities, and resource constraints. This research proposes a Q-Learning-based Adaptive Forwarding (QLAF) strategy designed to enhance reliability in heterogeneous disaster emergency communication networks. QLAF implements reinforcement learning into the NDN forwarding plane, enabling each router to autonomously learn optimal forwarding faces based on multiple performance metrics: Round-Trip Time (RTT), throughput, and link stability. The proposed strategy was implemented in the Named Data Networking Forwarding Daemon (NFD) and evaluated using the MiniNDN emulator over a BRITE-generated 25-node disaster topology that integrates terrestrial, cellular, and satellite links. We compared QLAF and Adaptive Smoothed RTT-based Forwarding (ASF), Access strategy, and Self-Learning. Experimental results show that QLAF achieves a Packet Delivery Ratio (PDR) of 99.91%. These results show that QLAF gives a robust solution for reliability-sensitive disaster communication, guaranteeing high data delivery performance under unstable network conditions. However, its latency overhead limits its applicability to real-time scenarios.]]>
Copyrights © 2026
