Date of Award


Document Type


Degree Name

Master of Science


Department of Electrical and Computer Engineering

First Advisor

Peter S. Maybeck, PhD


The task of tracking a target in the presence of measurement clutter is a two-fold problem: one of handling measurement association uncertainty (due to clutter), and poorly known or significantly varying target dynamics. Measurement association uncertainty does not allow conventional tracking algorithms (such as Kalman filters) to be implemented directly. Poorly known or varying target dynamics complicate the design of any tracking filter, and filters using only a single dynamics model can rarely handle anything beyond the most benign target maneuvers. In recent years, the Multiple Hypothesis Tracker (MHT) has gained acceptance as a means of handling targets in a measurement-clutter environment. MHT algorithms rely on Gaussian mixture representations of a target's current state estimate, and the number of components within these mixtures grows exponentially with each successive sensor scan. Previous research into techniques that limit the growth of Gaussian mixture components proved that the Integral Square Error cost-function-based algorithm performs well in this role. Also, multiple-model adaptive algorithms have been shown to handle poorly known target dynamics or targets that exhibit a large range of maneuverability over time with excellent results. This research integrates the ISE mixture reduction algorithm into Multiple-Model Adaptive Estimator (MMAE) and Interacting Mixed Model (IMM) tracking algorithms. The algorithms were validated to perform well at a variety of measurement clutter densities by using a Monte Carlo simulation environment based on the C++ language. Compared to single-dynamics-model MHT trackers running against a maneuvering target, the Williams-filter-based, multiple-model algorithms exhibited superior tracking performance.

AFIT Designator


DTIC Accession Number