ECE Seminar Series Spring 2016
Monday May 2rd 1-2 PM, ITEB 336
Industrial Strength Real World Multi-Sensor Data Fusion
Frederick E. Daum
Abstract: We explain why multi-sensor data fusion is a difficult problem in the real world. We also describe several new algorithms that are designed to solve such problems, considering real world effects. The major real world issues for data fusion include: (1) unresolved sensor data; (2) residual sensor bias errors; (3) closely spaced multiple targets; (4) unresolved sensor data; (5) non-unity probability of detection and non-zero probability of false alarms from noise & clutter; (6) inconsistent covariance matrices; and (7) unresolved sensor data. Such problems often result in putting the wrong data into your favorite estimation algorithm (extended Kalman filter, particle filter, unscented Kalman filter, etc.). The best way to ruin the performance of a good filter is to put the wrong data into it. One of the best algorithms to mitigate such problems is called GNPL (global nearest pattern Levedahl), invented at Raytheon. This algorithm jointly estimates the residual relative sensor biases and the association of data or tracks between sensors. The key word is “jointly”. We show comparisons of GNPL vs. simpler algorithms that decouple the bias estimation and data association problems. For difficult scenarios, GNPL is superior to simpler decoupled algorithms. We also describe more advanced algorithms that promise better performance at the cost of higher real time computational complexity. Such algorithms actually model the correct relevant physics, and they also exploit recent advances in particle filters and GPUs and the theory of random sets. This talk is for normal engineers who do not have nonlinear filters for breakfast.
Short Bio: Fred Daum is an IEEE Fellow, a principal Fellow at Raytheon, a Distinguished Lecturer for the IEEE and a graduate of Harvard University. Fred was awarded the Tom Phillips prize for technical excellence, in recognition of his ability to make complex radar systems work in the real World. He developed, analyzed and tested the real time algorithms for essentially all the large long range phased array radars built by the USA in the last four decades, including: Cobra Dane, PAVE PAWS, Cobra Judy, BMEWS, THAAD, ROTHR, UEWR, and SBX, as well as many other systems (SPY-3 proposal, JLENS proposal, SPACE FENCE proposal, LRDR proposal, JADGE, Project Hercules, ADI concept A, C-RAM, C-MAR, AN/TPN-19, ASDE-X, DERD-MC, NATO Sea Sparrow, DLGN-38, GPS OCX and several sonar systems). These real time algorithms include: extended Kalman filters, radar waveform scheduling, Bayesian discrimination, data association, discrimination of satellites from missiles, calibration of tropospheric and ionospheric refraction, and target object mapping. Fred’s exact fixed finite dimensional nonlinear filter theory generalizes the Kalman and Beneš filters. Fred’s particle flow nonlinear filter is many orders of magnitude faster than standard particle filters for the same accuracy. He has published nearly one hundred technical papers, and he has given invited lectures at MIT, Harvard, Yale, Caltech, the Technion, Ecole Normale Superieure de Paris, Brown, Georgia Tech., Duke, Univ. of Connecticut, Univ. of Minnesota, Melbourne Univ., Univ. of Toulouse, Univ. of New South Wales, Univ. of Canterbury, Liverpool Univ., Xidian Univ., Univ. of Illinois at Chicago, Washington Univ. at St Louis, McMaster Univ., Boston Univ., Northeastern University, Huntsville, Colorado and Rutgers.