A variational Bayesian approximation based adaptive single beacon navigation method with unknown ESV

The localization performance of single beacon navigation system is affected by the accuracy of effective sound velocity (ESV) which is difficult to precisely know. The state augmented method and expectation-maximization (EM) based method, the two existing state-of-the-art single beacon navigation methods which can deal with the unknown ESV, are sensitive to the noise statistic parameters and vehicle initial position offset, respectively. This paper proposes a variational Bayesian (VB) approximation based adaptive single beacon navigation method to deal with these deficiencies. The ESV is treated as a random variable with unknown statistic parameters, and the state vector, ESV and ESV uncertainty parameters are simultaneously estimated by VB approximation. Numerical studies indicate that the proposed VB approximation based navigation method can overcome the deficiencies of both state augmented and EM-based navigation methods, achieve better localization and ESV estimation performance than the existing state-of-the-art methods.