Statistical, Graphical, and Learning Methods for Sensing, Surveillance, and Navigation Systems
Abstract
This final report summarizes our accomplishments over the four years of support under this grant. The first of two interrelated research areas focuses on scalable, high-performance inference algorithms for graphical and hierarchical models. This has clear applications to sensor exploitation applications such as tracking and distributed network fusion, including the development of message-passing algorithms for location-aware networks in complex (possibly GPS-denied)and often communications-limited environments. Our second thrust focuses on discovering graphical models not only relating different sensor observables but also discovering and linking them to higher-level hidden variables capturing the common context that relates them. One motivation here is to enhance both lower-level sensor processing (e.g., for object recognition) and higher-level context discovery through these models. A second motivation is the discovery of complex, possibly coordinated dynamic behavior exploiting emerging methods of Bayesian nonparametric modeling.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 28, 2016
- Accession Number
- AD1011580
Entities
People
- Alan S. Willsky
- Moe Z. Win
Organizations
- Massachusetts Institute of Technology