Harnessing the Sensor Web Through Model-based Observation

This paper introduces the notion of model-based observing, a process of coordinating resources in a sensor web based on goals generated from Earth science or other investigations. Model-based observing will transform a sensor web into a cognitive web, a distributed, goal-directed sensing system. For a collection of remote sensors to be a cognitive web, an information infrastructure with the following features must be developed. First, there needs to be a representation of current observation goals and the tasks being performing to accomplish these goals. Second, there must be a means to recognize when changes to the sensing environment threaten to prevent the accomplishment of those goals. Third, a cognitive web must adapt to changes in its own configuration; it must be able to recognize opportunities for modifying its data acquisition plans in order to improve overall performance. It must act as a coordinator of resources, an optimizer of tasks, and a broker for data products. An architecture for a cognitive web is proposed that would utilize continuous planning and scheduling, plan execution and monitoring, and model-based goal generation. Portions of this architecture exist in extant systems.