Abstract: Target tracking involves the solution of problems which can roughly be broken into four components: sensor data alignment, track estimation, association, and track fusion. A collection of improvements which touch on each of those areas will be proposed to improve the overall accuracy and efficiency of a tracker.

An improvement to the global nearest pattern algorithm will eliminate the need for a fudge-factor mentioned in the literature and allow scale-invariant bias correction. Track estimation via interactive multiple model filtering can be improved with augmentation of the inputs via a neural network. Chernoff fusion can enable more accurate track-to-track fusion algorithms for nonlinear transformations of Gaussian distributions, and one novel application of that capability will be shown. Association using a variant of the multiple hypothesis tracking algorithm can be made more efficient with a deterministic sampling algorithm that efficiently covers the state space. And finally, the utility of random graphs for evaluation of distributed fusion algorithms will be discussed.

Adviser: Dr. Margaret Cheney
Co-Adviser: Dr. Chandrasekaran Venkatachalam
Non-ECE Member: Dr. Michael Kirby, Mathematics
Member 3: Dr. David Crouse, Electrical and Computer Engineering
Addional Members: NA

Publications:
NA

Program of Study:
ECE514
MATH520
MATH620
MATH633
ECE556
MATH676
ECE799
GSTR600