INFORMATION-DIRECTED ROUTING FOR MULTI-TARGET TRACKING IWIRELESS SENSOR NETWORKS

Abstract

With the development in wireless networking, and micro-fabrication technologies, it is now possible to develop large sensor networks. These networks, characterized by the micro nodes, are capable of wireless communication, sensing and processing data. Distributed sensor networks are characterized by limited battery power, variable data quality, and the unpredictable nature of the sensor nodes. To be used in tracking and classification applications, these systems have to rely on intelligent collaboration among distributed sensors. To further enhance the tracking accuracy and reduce the conserved power, multiple targets tracking at the same time should be considered.

The problem addressed in this thesis is how to dynamically query sensors, route data from multiple mobile targets in a network so that the information gain is maximized while the communication cost is minimized. The thesis proposes an algorithm to generalize the Information-Directed Routing Algorithm of Liu. Zhao, and Petrovic [LIU05] to handle multiple mobile stimuli. The key idea is to introduce an information utility measure to select which sensors to query and to dynamically guide data routing. Then the generalization of the algorithm to handle multiple stimuli is achieved using Steiner trees. Computing an exact Steiner tree is very expensive, so approximate algorithms are required. The thesis uses an approximate algorithm for the online Steiner tree problem, where the set of nodes in the tree changes over time to achieve the main goal. The performance of the information directed algorithm using the Steiner tree is analyzed and compared to the case using the Shortest Path tree with respect to the tracking accuracy, the communication cost and the number of sensors queried. The proposed scheme is shown to be a better candidate.