Font Size: a A A

Application driven sensor network architecture

Posted on:2007-08-08Degree:Ph.DType:Dissertation
University:University of California, BerkeleyCandidate:Szewczyk, RobertFull Text:PDF
GTID:1448390005960346Subject:Computer Science
Abstract/Summary:
Habitat and environmental monitoring is a driving application for wireless sensor networks (WSN). Motivated by a study linking environmental conditions with the behavior of Leach's Storm Petrel, a small ocean bird, nesting on the islands of the northern Atlantic, we develop a set of requirements for habitat monitoring applications. The application requires reliable performance over an extended deployment time; our system architecture aims to achieve these goals by building on principles of simplicity, availability, manageability and pervasive power awareness.; We evaluate the proposed architecture based on two implementation iterations. During the summer months of 2002 and 2003, we deployed this system on Great Duck Island off the coast of Maine. During the first season, we deployed 43 wireless sensors that monitored the bird nest occupancy and environmental conditions. During the following season, we deployed a refined system consisting of 150 wireless nodes. We discuss the physical elements of the application, from the node architecture, through the hierarchy of networks required by this deployment, to the base station design. We proceed to describe and evaluate the minimal set of software services required by the WSN.; We evaluate the two deployments based on over 2 million readings collected by the two deployments, each spanning a 4-month window. We analyze the node lifetime and reliability. We examine the networking performance over time: we analyze the link-level packet loss, as well as topologies and dynamic behavior of a long-lasting multihop network. Finally, we examine the quality of readings, and their suitability for life-science analysis. We use a number of analysis techniques that allow us to perform the analysis despite the missing data; we develop several filtering rules to distill the biologically meaningful data. We show that the sensor data is also useful for predicting system operation and network failures.
Keywords/Search Tags:Sensor, Network, Application, Architecture, System
Related items