A Fast Variational Bayesian Adaptive Extended Kalman Filter for Robust Underwater Direction-of-Arrival Tracking

Underwater direction-of-arrival (DOA) tracking using hydrophone array is an important research subject in passive sonar signal processing. In this study, the robust underwater DOA tracking with the unknown measurement noise results from underwater environmental noise is concerned. To deal with the unknown measurement noise, the existing variational Bayesian adaptive extended Kalman filter (VB-AEKF) for robust underwater DOA tracking jointly estimates the covariance matrix of unknown measurement noise and the DOA by using variational Bayesian approach in the framework of extended Kalman filter. However, the iteration process of VB approach and the nonlinear measurement model based on hydrophone array signal result in significant computational complexity of the VB-AEKF for DOA tracking. To better solve the problem of robust DOA tracking with unknown measurement noise covariance matrix, a novel computationally efficient version of the existing VB-AEKF, which is named the fast variational Bayesian adaptive extended Kalman filter (FVB-AEKF), is proposed. An equivalent variational iteration process to estimate the measurement noise covariance matrix is derived. The proposed FVB-AEKF is identical to the VB-AEKF but reduces the computational complexity of the algorithm. The proposed method is tested via Monte Carlo simulations of an underwater DOA tracking scenario and an experiment in the South China Sea in July 2021. The results of the simulations and the experiment verified the accuracy and the computational efficiency of the FVB-AEKF.