This article examines a strategy to establish are liable underwater acoustic communication for navigation of autonomous underwater vehicles (AUVs). This work proposes a framework through the use of kernel function based models to make the task of locating AUVs less sensitive to channel fluctuations. For this, the Auto-Associative Kernel Regression(AAKR) and the Support Vector Data Description (SVDD) are integrated to the data fusion algorithm to improve the accuracy of the estimated time of flight (ToF) of acoustic signals.