Acoustic cost functions for autonomous adaptive sampling

Abstract

Recent advances in ocean modeling have provided the opportunity to forecast ocean conditions out to 48 h. Accuracy of these models is determined primarily by the input conditions (boundary, as well as internal state). An automated algorithm for adaptively sampling the ocean environment has been developed with the goal of providing sampling paths for regions where input drives the model solutions the most. In this paper, a set of acoustic cost functions is presented, which should indicate regions where acoustic propagation is most sensitive to model uncertainty and/or model‐predicted temporal variability. The four cost functions explored will be: Matched mode coherence, mode coupling coefficients, transmission loss variability, and detection range coverage. This technique will be applied to the data‐assimilated model results from the Shallow Water 2006 experiment, where oceanographic measurements consisted of a large array of thermistors, six gliders operating continuously, and a ship towed scanfish. Acoustic measurements included fixed source/receivers, AUV mounted sources, as well as sonobuoy receivers. Sampling guidance based upon the acoustic cost functions will be used in an observational systems simulation experiment (OSSE), in collaboration with Pierre Lermusiaux and his group at MIT. For the ocean modeling, data assimilation and sampling suggestions see http://modelseas.mit.edu.