Statistical Problems in Wireless Sensor Networks.

Wireless sensor networks (WSN) are a new technology with many applications, including environmental monitoring, surveillance, and health care. The dissertation concentrates on two critical aspects of a WSN: network design and information fusion. Our design strategy minimizes the overall network cost, explicitly incorporates sensor capabilities, and maintains coverage and connectivity constraints necessary for successful network operation. A new algorithm for local correction of sensor decisions, Local Vote Decision Fusion, is developed for the problems of target detection, localization, and tracking, and extended to multiple targets. The methodology is tested in simulations and on two case studies - an experiment involving tracking people and a project of tracking zebras. The local correction algorithm is further developed into a general framework for performance improvement for spatially correlated classifiers.Ph.D.StatisticsUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/63701/1/nkatenka_1.pd