Data Fusion in Decentralised Sensing Networks

This paper briefly describes the results of a ten year, and still on-going, research program in decentralised sensing systems. This program covers both the theoretical development of data fusion methods appropriate to networks of decentralised sensors and the practical implementation of these in both civilian and military contexts. The methods employed for studying decentralised data fusion problem are based on the information-filter formulation of the Kalman filter algorithm and on information-theoretic methods derived from Bayes theorem. This theory is briefly described in context of a number of practical implementations of decentralised data fusion methods in surveillance and control applications. The paper describes specific theoretical tools developed to address such issues as; decentralised communication management, model distribution, decentralised data association and fault detection, sensor control (information gathering, target cuing and hand-off), decentralised picture compilation and map building. Finally, we describe our current work toward deployment of decentralised, large-scale, ’systems of systems’ demonstrations.