Improvements in the estimated time of flight of acoustic signals for AUV localization

The problem acoustic localization for autonomous underwater vehicles (AUVs) is discussed in the framework of kernel-based models. The approach consists of improving the measurement of time of fligth of acoustic signals with the help of the Auto-Associative Kernel Regression (AAKR). This makes localization less sensitive to perturbations in the acoustic channel improving the estimation filter convergence. Field tests with vehicle confirmed these claims.