Endowing underwater networks with channel awareness: a discussion on computational complexity and information size issues

In this paper, we consider the problem of transferring channel state information (CSI) to the nodes of an underwater network. As the CSI is derived by simulating the behavior of the underwater channel via ray tracing, we first discuss a parallel implementation of the ray tracing software obtained using the CUDA architecture, that helps compute the CSI faster. We then consider the problem of making the channel state information compact, and suitable for transmission over underwater acoustic channels. Our results indicate that a feedforward artificial neural network achieves both good accuracy and good compression of the CSI.