Distributed Tracking in Wireless Ad Hoc Sensor Networks

Target tracking is an important application for putation and thus the processing has to be distributed, with Abrelsstract ho T etsor t ngtworks. i eas e an i a th a nfrgy nodes communicating processed data instead of sensor data. wireless ad hoc sensor networks. Because of the energy Distributed tracking for sensor networks was first invesand communication constraints imposed by the size of the tigated in the early 1980's to understand how sensors with sensors, the processing has to be distributed over the distributed but overlapping coverage can be used to detect sensor nodes. This paper discusses issues associated with adtaktres ic hnteehsbe uhrsac distibued ultple argt tackng fr a ho sesor and track targets. Since then there has been much research distributed multiple target tracking for ad hoc sensor networks and examines the applicability of tracking algoon distributed tracking [2,3]. However, energy constrained netwrksandexaine t li iliy c g g pr ces ing and communication has not been a main concern rithms developed for traditional networks of large sensors. and the research emphasis has been on association and When data association is not an issue, the standard prefusion of tracks from multiple processing nodes. dict/update structure in single target tracking can be used Distributed tracking in sensor networks is a multito assign individual tracks to the sensor nodes based on Discipnyea that invles har ware, si a prcsin their locations. Track ownership will have to be carefully disciplinary area that involves hardware, signal processing, theigrlocatedsions Tora onerhpe wilhavetion bven c ore estimation and inference algorithms, and computer science. msingrat, usmingz fr nexaped iforcommuniation drsensr Researchers in computer science and signal processing have tasigeto m inie then need fssorc ommunationisnee n when undertaken most of the recent work [4,5]. While feasibility targets move. When data association is needed in tracking has been demonstrated for simple scenarios in simulated multiple interacting targets, clusters of tracks should be and real test beds, the development of algorithms for more assigned to groups of collaborating nodes. Some recent exasplesi tof gros typof collaboratingnodes. Some reent difficult and realistic tracking scenarios is just beginning. examples of this type ofdistributed processing are given. This paper starts with the standard tracking algorithms and investigates how processing can be mapped to a sensor