Skip to Main Content
Dense deployment of sensor and the nature of the physical phenomenon being observed (sensed) result in spatio-temporal correlation in the observed data. Data fusion plays an important role in sensor networks for exploiting this correlation. However, fusion performance depends on parameters such as the reliability model of the sensors, sensor observations, and a priori information. Individual sensors transmit likelihood functions to the fusion center to produce a single posterior distribution or estimate. Here we proposes a new fusion method, reliable likelihood opinion pool (RelOP), for aggregating likelihoods to produce a reliable estimate. It is based on a Bayesian framework. The performance of RelOP is compared with the commonly used opinion pools through simulations. We further propose a multi-sensor fusion architecture that follows from application of the RelOP rule.