Glenn T. Nofsinger holds a B.S. in physics from The College of William and Mary, and a Ph.D. in Engineering Sciences from Dartmouth College. His work centers on pattern detection and classification. The ultimate objective of many pattern detection problems is the recognition of high level behavior in large and noisy data sets. Bayesian statistical methods and machine learning techniques serve as excellent tools for such problems. With these fundamental techniques, he has designed applications in the disparate fields of distributed systems, sensor networks, social network analysis, image processing, and quantitative finance.
Multiple hypothesis tracking algorithms, hidden Markov models, unsupervised clustering, and reinforcement learning.
Learn more about his mathematical genealogy at the Mathematics Genealogy Project.
- Hidden Markov Models
- Machine Learning
- Social Modeling with System Dynamics
- Distributed Source Localization
This chemical simulation accounts for a pseudo-random wind field, and releases multiple sources at arbitrary times. Can be used for data generation in a simulated sensor network. Accounts for advection (wind) and diffusion.
This code assigns new observations to tracks, and maintains the most likely sequence of observations in real-time.
Based on collected observations and a collective Bayesian estimation, this code allows groups of observations to be inverted and form a situational map, or liklihood map of a region.
In this project we present a system of simple chemical sensors capable of inverting large numbers of low resolution observations into state estimates on the location(s) of chemical agent release. This novel method is based on Multiple Hypothesis Tracking (MHT) and does not assume the ability to invert distance to a source based on concentration. This application was one of many developed around an approach known as a Process Query System (PQS). The notion of a PQS is loosely based on a Wiener filter which matches incoming streams of real-time data for instances of a desired process. A process is any event that produces an observable. See this overview of plume tracking in sensor networks
In this project we address the problem of mobile targets in a static field of sensors. Independant of sensing modality. This could apply to mobile chemical plume sources, or seismic/acoustic sensors tracking a mobile vehicle.
Systems Dynamics modeling
allows the understanding of extremely complex systems over time. By introducing time delays and feedback loops one can begin the isolation of the main driving loops within a larger complex system. When coupled with other modeling paradigms such as agent based modeling, extremely dynamic systems emerge.
The Process Query System (PQS) technology developed at Dartmouth College is a new form of process filtering, with applications in tracking, detection, network security, video tracking, as well as my application of chemical plume detection. This is an active and ongoing project. Featured on the cover of IEEE Computer magazine Jan 2007. The main PQS group at Dartmouth College continues to apply this process detection to new application areas.
- Tracking Based Plume Detection. Glenn Nofsinger's PhD Thesis at Dartmouth College, 2006
- "Airborne Plume Tracking With Sensor Networks." Glenn Nofsinger and George Cybenko. In Proceedings of the SPIE, Defense and Security Symposium - Conference 6231 - Unattended Ground, Sea, and Air Sensor Technologies and Applications VII, Orlando, FL, April 2006.
- Distributed Chemical Plume Process Detection. Glenn Nofsinger, George Cybenko, In IEEE Military Communications Conference, MILCOM, Atlantic City, NJ, October 2005
Plume Source Detection Using a Process Query System. Glenn Nofsinger, George Cybenko, In Proceedings of the SPIE Vol. 5416, Defense and Security Symposium - Chemical and Biological Sensing V,Orlando, FL, 2004