Multipath Data Fusion With Recursive ML-PDA and Generative ML-PMHT for VLO Targets in Underwater Environment

Abstract

We discuss and extend two approaches to target detection and motion parameter estimation in a very low observable (VLO) underwater multipath environment. The maximum likelihood probabilistic data association (ML-PDA) algorithm considers all possible path–measurement hypotheses, whereas the maximum likelihood probabilistic multiple hypothesis tracker (ML-PMHT) assumes independence among the returns’ origins leading to a simpler likelihood formulation. In particular, we present a recursive form of the multipath ML-PDA algorithm for improved computational efficiency and a generative multipath ML-PMHT algorithm that accounts for the prior probability distribution of the number of received measurements. By simulation of an active sonar tracking scenario, we show that while the two algorithms perform similarly for few returns, the added fidelity of the ML-PDA outperforms the ML-PMHT in parameter estimation as returns accumulate over multiple scans and dimensions. However, the computational cost of the ML-PDA discourages its use beyond simple scenarios. This cost, coupled with the comparable detection performance of the two algorithms, suggests that the simpler ML-PMHT is the better choice for practical use. We also show how both the ML-PDA and ML-PMHT can be applied to realistic ocean environments using ocean acoustic modeling software. This article is an extension of previous work on comparing the standard versions of the ML-PDA and ML-PMHT algorithms.