Enhancing DTN Routing Strategies with Deep Reinforcement Learning in Disaster Recovery Networks

The application of Delay Tolerance Network (DTN) technology in emergency situations has gained significant attention. This paper proposes a routing protocol designed to facilitate communication and coordination among various rescue departments during emergencies, such as post-disaster scenarios, to enable rapid search, rescue, and transportation of trapped individuals and essential supplies. The network's mobile nodes are classified into two categories, and specific post-disaster routing strategies are developed for the messages carried by these nodes. The paper employs deep reinforcement learning methods in a multi-agent setting, allowing each agent to autonomously select suitable nodes for information transmission in local environments. Additionally, a CVR voting mechanism is used to evaluate the quality of information collected by multiple nodes. Finally, simulations conducted using the ONE simulator validate the feasibility of the proposed method. Results indicate that the proposed scheme performs well in terms of network load, message delivery rate, and transmission delay over time.