:ai
ABSTRACTIn recent years, the study of systems with multiple sensors has been an active area of research. In this paper, we focus on the computational, i.e., architectural, algorithmic, and synchronization issues related to competitive information integration in a Distributed Sensor Network (DSN). The proposed architecture of the DSN consists of a set of binary trees whose roots are fully connected. Each node of the tree has a processing element and one or more sensors associated with it. The information from each of the sensors has to be integrated in such a manner that the communication costs are low and that the real time needs are met. We present an information integration algorithm which has a low message cost (linear in the number of nodes of the network ) and a low distributed computation cost.In a distributed environment there is no central clock which regulates the activitie&< of each node. Further, the clock at each node is typically not accurate. The estinates from each of the sensors which need to be integrated have to be temporally "close to each other', however. We consider the problems associated with synchronizing information to be integrated in the presence of imperfect clocks. .. We derive a) the relationships between the clocks of the processing elements in the network for proper information integration and b) an upper bound on the period between consecutive resvnchronizations of a processing element's clock with the central time server. -Finally, we discuss the fault tolerant features of the network and the integration algorithm.
KEYIVORDS and PHRASES
I'
INTRODUCTIONIn recent years the study of multisensor systems has been an active area of research [14,4,2, 11, 10]. Our interest is in multiple sensor systems which consist of tens or even hundreds of sensors. Such a system is usually organized as a distributed sensor network (DSN) which consists of a set of sensors, a set of processing elements (PEs), and a communication network interconnecting the various PEs. One or more sensors is associated with each PE. A need for DSNs arises in diverse applications such as intelligent robotic systems, aircraft navigation, systems which monitor the activities on an industrial assembly line, etc.[4, 10, 15]. A significant advantage offered by the integration of information from disparate sensors is that the reliability and the fault tolerance of the sensor system are both enhanced. Another advantage is the suppression of the effects of noise. This is because the noise measured by different sensors tends to be uncorrelated, while the signal of interest remains correlated. Also, by careful selection of disparate sensors, one can compensate for the shortcomings and peculiarities of particular types of sensors.The method of integrating information from the various sensor outputs, called information integration, depends on whether the sensors provide a) competitive information or b) complementary information [4,2]. In the former case, each sensor ideally provides identical information. In reality, however,...
