AN ARCHITECTURE FOR DECISION SUPPORT IN AD HOC SENSOR NETWORKS

The intelligent jobsite is becoming a reality as applications using sensors and mobile computing devices are being developed and deployed commercially. This creates an opportunity for workers to access data in an ad hoc manner as they move through a site. However, retasking and reuse of sensors in a dynamic setting presents significant challenges including ad hoc identification of sensors in a local environment, generalization of data, and the use of dynamic information for decision support. To achieve a generalized approach to using local data, we describe an architecture that abstracts functionality into three tiers: a layer for physical communication among devices; a layer for data processing and abstraction; and a top layer for decision support. At each tier, the level of abstraction increases, allowing for development of decision support applications at the top level that are not directly tied to specific devices. The industry vision for the intelligent jobsite is driving considerable research and development in mobile technologies. However, the bulk of this development is application specific and directly ties hardware (e.g., RFID tags, sensors) to data processing applications. We seek to decouple sensors from computing hardware and application development by providing a flexible and robust middleware. Realization of the middleware will enable more flexible reuse of data to make it available to a range of decision support applications, while at the same time speeding development of such applications. This paper describes an architecture informed by a working first generation prototype. Details of the prototype, lessons learned, and specific advancements are detailed. Future commercial implementation of the architecture will make construction-specific visions for ubiquitous computing possible by enabling flexible and robust discovery and use of data in an ad hoc manner.