Application of Sigma Point Particle Filter Method for Passive State Estimation in Underwater

Abstract

Bearings-only tracking (BOT) plays a vital role in underwater surveillance. In BOT, measurement is tangentially related to state of the system. This measurement is also corrupted with noise due to turbulent underwater environment. Hence state estimation process using BOT becomes nonlinear. This necessitates the use of nonlinear filtering algorithms in place of traditional linear filters like Kalman filter. In general, these nonlinear filters utilize the assumption of measurements being corrupted with Gaussian noise for state estimation. The measurements cannot be always corrupted with Gaussian noise because of the highly unstable sea environment. These problems indicate the necessity for development of nonlinear non-Gaussian filters like particle filter (PF) for underwater tracking. However, PF suffers from severe problems like sample degeneracy and impoverishment and also it is tedious to select an appropriate technique for resampling. To overcome these difficulties in PF implementation, the strategy of combining PF with another filter like unscented Kalman filter is proposed for target’s state estimation. The detailed analysis of the same is presented in comparison with other particle filter combinations using the simulation results obtained in Matlab.