Locality-aware clustering for dynamic collaborative information processing in wireless sensor networks.

摘要

A wireless sensor network (WSN) is characterized by a collection of low-priced, battery-powered embedded sensing devices that have limited computation and communication capabilities and especially stringent energy constraints. Thus, the necessity of sensor collaboration has become a unique feature of WSN applications, and various forms of collaborations have been developed based on requirements of applications. In large-scale event-based applications, such as forest fire tracking, original locations of events are unavailable a priori. Hence, a sensor collaboration approach that dynamically organizes collaborative sensors in response to event changes, and efficiently performs Dynamic Collaborative Information Processing (DCIP) can greatly improve performance of a WSN in reporting updated status information of spontaneous events.;Clustering of sensor nodes has been shown to be an effective approach for DCIP in resource constrained WSNs to keep network traffic local in order to reduce energy dissipation due to long-distance transmissions. The key issue is to define the range and topology of clusters to achieve the minimal overall energy consumption. Most research in the area, however, aims to produce a small number of clusters by finding a Minimum Dominating Set or a Maximum Independent Set, and pays less attention on cluster locations and sizes. As a result, collaborative clusters may incur higher radio interference, and thus increase energy dissipation on data retransmissions. The novelty of our Locality-Aware Dynamic Sensor Collaboration (LA-DSC) approach is three-fold. First, upon occurrence of a phenomenon event, a Collaborative Agent Sensor Team (CAST) is dynamically established for all detecting nodes. The CAST network structure is constructed into clusters that hold three important properties to achieve energy saving on reduced interference: minimized overlapping cluster areas, approximately equal cluster sizes, and the solid disk property. Second, a two-tier Data Aggregation Tree is designed in LA-DSC to achieve energy efficient and scalable data gathering on the established CAST structure. Finally, a LA-DSC light-weight CAST reconfiguration approach is provided to reshape a CAST to report status information of dynamic phenomenon changes. In summary, LA-DSC provides an effective DSC approach to facilitate efficient and scalable collaborative sensor data processing with reduced overall energy consumption.