Environmental-Parameter-Importance-Informed Robust Matched Field Processing in Uncertain Shallow Water

Matched field processing (MFP) compares the array-sampled sound field with the predicted field from a full-wave propagation model for source localization in shallow water. Its performance is highly limited by environmental uncertainties, that is, the environmental parameters are known only within certain value bounds—for example, the water depth is known within (102.5 ± 2.5) m. The prevalent method, the MFP with environmental perturbation constraints (MFP-EPC), achieves improved robustness by constraining the beamformer response on possible signal vectors regarding different realizations of all unknown environmental parameters. But it experiences large performance degradation when the uncertainties are extant. This work exploits the varying effects of the uncertainties in various environmental parameters on the sound field’s variations and proposes an enhanced MFP-EPC method. We find from numerous numerical studies that the uncertainty in water depth can cause much more notable variations to the possible signal vectors than the other environmental parameters in downward-refracting shallow-water channels. We also demonstrate that the MFP-EPC’s performance is proportional to the correlations among the possible signal vectors. Thus, we propose to separate the water depth from the full environmental uncertainty set and to match the received data with multiple sets of the MFP-EPC weights corresponding to different possible water depths within its value range. The resulting processor, called the environmental-parameter-importance-informed MFP-EPC, can achieve better performance over MFP-EPC. The performance superiority of the proposed method is verified by simulations in the general mismatched benchmark channel and is evaluated under different source and environmental conditions. The real data collected at the north of the island of Elba in 1993 also demonstrate the proposed method’s effectiveness.