Topic: Information Theoretic Sensor Management
By: Jason L. Williams (MIT), John W. Fisher (MIT), Alan S. Willsky (MIT)
Inference in sensor networks involves the competing objectives of minimizing uncertainty while attempting to maximize the life of the network. Mutual information is an effective measure of uncertainty for complex multi-modal distributions, which can be conveniently used to trade off estimation performance for communication cost. This work will utilize such information theoretic methods in an approximate dynamic programming framework to address problems such as distributing the target state representation across sensors, coordinating dynamic hand-off of the representation from sensor to sensor, and determining strategies for selecting which subset of sensors should sense and/or communicate at each point in time.
Questions and comments to the webmaster. This page is copyrighted by the Massachusetts Institute of Technology.