Prioritized Content Determination and Dissemination Using Reinforcement Learning in DTNs

Abstract

In a battlefield, several groups of soldiers are deployed with different mission goals by the command and control center (CC). To continue the missions appropriately and get a better understanding of the situation, the soldiers, as well as the CC, need to collect information of interest generated in different battle zones. However, due to the damaged network infrastructure in the hostile areas, it is a challenge to determine the topics of interest associated with the events and missions, and efficiently forward the associated content to the CC. Hence, the devices of the soldiers (nodes) generate, store and forward content hop by hop using a Delay Tolerant Network (DTN). While forwarding content, nodes avoid congestion so that meaningful content related to prioritized mission goals can be disseminated. In this dynamic surrounding, any sudden but important event-related content should also be sent to the CC with the help of intermediate nodes regardless of their own mission interests. We design a scheme to forward contents generated by the nodes to the CC using Reinforcement Learning (RL) while maximizing the number of interesting data in the respective nodes' buffer, and avoiding congestion. In this forwarding process, we focus on identifying the trending topics/keywords among changing missions and their related data at the node level, and the changes of interest of the nodes based on their mobility and connectivity patterns. Experiments are conducted using real datasets and ONE simulator to show the effectiveness of Reinforcement Learning (RL) on the prioritized content dissemination in a DTN.