부산대학교 도서관

The advent of the Internet of Things (IoT) is leding to huge volumes of data from millions of connected devices dedicated to collecting, transferring, and generating data from a wide variety of domains. Thereby, several approaches have been proposed in the literature for data aggregation, most notably that rely on adding an edge computing layer to offload the network of data transferred to the cloud for processing. However, this approach faces several challenges related to IoT network clustering and load balancing. As a result, different methods for clustering networks based on static typologies have been proposed; however, this static aspect is ineffective when the batteries of some devices are depleted at the expense of others. In this work, we propose a method based on reinforcement learning to find dynamically an efficient clustering that guarantees load balancing. Our approach remains adaptive and able to adjust itself to changes in the energy capacity of devices. From the results obtained, our proposal achieved a significant improvement in terms of load balancing in the presence of devices with depleted batteries.