Channel Selection Algorithm based on Machine Learning for Multi- Medium/Multi- Bandwidth Communication in Underwater Internet of Things

Although Internet of Things (IoT) on the ground has been widely developed, there are a lot of challenges for Underwater Internet of Things (U-IoT) because of the hostile environments. Nevertheless, U-IoT is a potential field and it can be a platform for providing abundant services by interworking with terrestrial networks. In this paper, we introduce Multi-Medium and Multi-Bandwidth (MM/MB) communication for the connectivity that is one of the important constituent parts in U-IoT. The MM/MB communication is to use all possible methods for underwater communication. The main technology of MM/MB communication is channel selection, which aims to select suitable medium and bandwidth among various communication channels according to current states. This paper is intended as an investigation on a channel selection algorithm based on Machine Learning (ML). For ML, we collected datasets through several experiments and extracted seven features from the results that can be used for a softmax regression model training and testing. The model we trained performed with an accuracy of 96.5%.