As sensor hardware becomes more sophisticated, smaller in size and increasingly affordable, use of large scale sensor networks is bound to become a reality in several application domains, such as vehicle condition monitoring, environmental sensing and security assessment. The ability to incorporate communication and decision capabilities in individual or groups of sensors, opens new opportunities for distributed sensor networks to monitor complex engineering systems. In such large scale sensor networks, the ability to integrate observations or inferences made by distributed sensors into a single hypothesis about the state of the system is critical. This paper addresses the sensor integration issue in hierarchically organized sensor networks. We propose a multi-agent architecture for distributed sensor networks. We present a new formalism to represent causal relations and prior beliefs of hierarchies of sensors, called Hierarchically Organized Bayesian Networks (HOBN), which is a semantic extension of Multiply Sectioned Bayesian Networks (MSBN). This formalism allows a sensor to reason about the integrity of a sensed signal or the integrity of neighboring sensors. Furthermore, we can also evaluate the consistency of local observations with respect to the knowledge of the system gathered up to that point.
Multi-target tracking in distributed sensor networks using particle PHD filters
