Distributed sensing based on intelligent sensor networks

Network centric approach to complex systems has been attracting a lot of attention in the recent past. With specific applications in space, military and other information enterprises, such an approach is capable of yielding autonomy at the system level. This article describes a distributed approach to information processing in sensor networks. The network architecture entails a hierarchy of capabilities, information and control, where nodes in the network are thoroughly or partially autonomous. The network also consists of slave nodes dedicated to sensing and gathering information. Individual nodes are expected to possess resources for networking and computing and presume autonomy through multi functional modules for sensory processing and situation assessment. Data association, registration and fusion are formulated into a joint process, and an expectation maximization (EM) approach is developed to solve the three problems simultaneously. A fuzzy cognitive map (FCM) is proposed to perform situation assessment, while a genetic algorithm (GA) is applied for learning, to identify an optimal structure. The distributed architecture is applied to coastal surveillance with simulations on the west coast of Canada.