A Novel Bayesian-Based Adaptive Algorithm Applied to Unobservable Sensor Measurement Information Loss for Underwater Navigation

In complex underwater environment, it is hard to acquire the accurate position estimation when the sensor measurement information is lost. Under this situation, the measurement loss is unobservable, and meanwhile, the statistical information of system noise is unpredictable and time-varying, so existing algorithms have poor performances. To address the problem, this article proposes a Bayesian-based adaptive Kalman filter with Gaussian-inverse-Wishart mixture (BAKF-GIWM) distribution. By using BAKF-GIWM, the measurement loss is judged through the process of sensor measurement, and then, the system state vector and statistical information of system noise are together determined according to the variational Bayesian (VB) method. Meanwhile, the estimation results of measurement loss status are regarded as evaluation criteria to improve the efficiency of the BAKF-GIWM. The simulations and the real trial results illustrate that the proposed algorithm can efficiently improve the state estimation accuracy in complex underwater environment.