Modeling and prediction of large-scale temporal variation in underwater acoustic channels

Abstract

Adaptive operation in underwater acoustic networks relies on channel prediction. This work models and predicts the large-scale variation of underwater acoustic channels by taking the channel signal-to-noise ratio (SNR) as a quality indicator, which is defined as the received SNR with a unit transmission power. The channel SNR process is modeled as a summation of an environment process which can be represented as a linear combination of a set of measurable environment parameters, and a Markov latent process that accounts for the contribution from unknown or unmeasurable physical mechanisms. Based on historical SNR measurements and available environment parameters, a recursive algorithm is developed to estimate the latent process and the combinational coefficients of environment parameters, which are then used for channel prediction. The algorithm is further extended to seasonal channels, where both a sequential channel prediction algorithm and a joint channel prediction algorithm are developed. The proposed channel model and prediction algorithms are validated via extensive simulations and experimental data analyses.