A Comparison of Kriging and Cokriging for Estimation of Underwater Acoustic Communication Performance

George P. Kontoudis, Daniel J. Stilwell

Abstract: Mobile underwater communication network nodes, such as autonomous underwater vehicles, can use estimates of underwater acoustic communication performance to anticipate where they are likely to be connected to the communication network. In this paper, we consider the challenge of estimating a spatial field that represents underwater acoustic communication performance from a set of measurements. Kriging, which is widely used in geostatistics, has been previously used to estimate the communication performance at unknown locations, by performing spatial extrapolation. We compare kriging to cokriging where the latter is a bivariate estimation method. The methodology yields estimates of communication performance at desired locations based on measurements acquired at other locations. Moreover, a variance measure is provided that characterizes the uncertainty of the estimation. We present the structure of the proposed estimation technique and its computational complexity. We evaluate the efficacy of the technique by considering an approximate linear-log model of the communication performance, environmental noise, and a direct comparison of kriging and cokriging results. We provide two sets of simulations in which the proposed multivariate cokriging framework outperforms the univariate kriging in the estimation process.