文摘
Research in wireless sensor networks (WSNs) has created many approaches to conserve scarce energy resources while accomplishing the critical tasks of data collection and message routing. However,it remains difficult for non-specialists to leverage the understanding of WSN application design. One recent area of research focuses on the definition of a re-useable software module to encapsulate energy management concerns. There have been a variety of approaches,but none allows complete flexibility for optimization using application defined utility measures without imposing significant overhead costs by flooding status report messages or collecting all node status information to a central location for decision-making. This paper proposes a middleware,a Dynamic Framework for Energy Management (DFEM),that allows users to describe application utility,and then solves for the global optimal policy for the network using only information from immediate neighbor nodes. The basis for this optimization is a dynamic programming formulation of the sensing and routing problem of an arbitrary network,which is adapted into a distributed algorithm that solves the dynamic programming through policy iteration. In addition,DFEM provides a skills gradient for different target user groups. New sensors and task descriptions can be extended from the library code with only basic programming experience,and deployment and configuration require no special programming or wireless networking knowledge. Extensions of the basic DFEM are investigated to relax the lifetime constraint description,to allow user queries to modify the application description,and to include low-latency reporting of rare events as a network responsibility. Both simulation and physical deployment results are discussed and compared.