A decentralized Bayesian algorithm for identification

The problem of identification using a fully decentralized sensing network is considered. Given that an object or situation can be defined based on the values of several parameters, identification involves accurately determining that state based on uncertain observations of each parameter. A fully-decentralized network is made up of several locally intelligent sensors which act not only as transducers but also as decision makers. Locally obtained results are communicated to the other sensors for further processing, allowing each sensor to achieve a global result. Such a system has many benefits in terms of modularity, speed, and survivability. A Bayesian information fusion system is presented that allows sensors to calibrate information from other sensors before fusion occurs. A metric to determine disagreement between sensors is also described. The authors maintain that a decentralized system is the most appropriate sensing configuration for this problem, and describe implementation work currently in progress