Upcoming Event: PhD Dissertation Defense

Advancements in Single- and Multi-Target Filtering: Using Posterior Estimates to Update Gaussian Mixture Weights

Dalton Durant,

Abstract

A major focus of this Ph.D. defense is on computing the weights of Gaussian mixture filters for single- and multi-target filtering. Computing the weights is a crucial part of the filtering process because they determine the contribution of each Gaussian mixture component to the overall distribution. Traditionally, weights are approximated by linearizing the nonlinear measurement model about the prior state estimates. But, in nonlinear measurement scenarios, using the prior estimates to compute the weights is only an approximation, and one that ignores the latest information available, which can lead to subpar filtering performance and inconsistent state estimates. Therefore, this thesis argues that the weights should instead be computed using posterior state estimates because the posterior provides a better characterization of the system after receiving the latest measurement data. In this defense, some new weight formulations using posterior estimates are proposed, as well as a new kernel-based multi-target filtering algorithm.

Biography

Dalton is an aerospace engineering Ph.D. candidate at the University of Texas at Austin (UT) under Renato Zanetti. Dalton received the B.S. in aerospace engineering from the University of Maryland, College Park, MD, USA, in 2020, and the M.S.E from the University of Texas at Austin, Austin, TX, USA, in 2022. His research interests include nonlinear estimation, data assimilation, and multi-target tracking.