A Comparative Study of IMM Kalman Filters and Neural Computing for Underwater Bearing Only Target Motion Analysis

Bearing only target motion analysis (BOTMA) is essential for tracking and locating underwater targets, particularly in situations when conventional approaches such as range-based observations are not feasible or accessible. This paper offers a comparative investigation of interacting multiple model (IMM) Kalman filters and neural computing modelling for analyzing the motion of a target in underwater environment using only bearing data. The study aims to assess the effectiveness of above methods in efficiently analyzing motion features of an underwater target using just bearing observations obtained from two passive observers. The proposed methodologies address the challenges given by noisy and limited bearings, providing a substantial improvement in the analysis of the target's motion. To determine the performance of these two adaptive approaches, a supervised large turning track is developed for an underwater maneuvering target. Real time position, velocity, and turning features of a passive object are estimated in a cluttered marine atmosphere. The objective is to minimize the mean square error (MSE) among true and estimated motion of the target. The performance of each technique is evaluated on the basis of accuracy and resilience through detailed analysis. The results clearly demonstrate superiority of neural computing modelling compared to typical nonlinear filtering tools, like interacting multiple model extended Kalman filter (IMMEKF) and interacting multiple model unscented Kalman filter (IMMUKF). Therefore, this comparative study offers useful insights into the merits and drawbacks of each method, providing recommendations for future advancements in underwater target motion analysis and tracking models.