Utility-based dynamic resource optimization in mission-oriented wireless sensor networks.

摘要

Mission-oriented wireless sensor networks present an interesting class of applications, such as battlefield monitoring and emergency disaster response scenarios, that are different from conventional wireless sensor network (WSN) applications. These applications often require the network to be operational for relatively short time periods (i.e., days rather than months), and employ sophisticated, high data rate sensors (e.g., video, short aperture radar and acoustic sensors). In such mission-oriented, multi-hop WSN environments, both bandwidth and energy are critically-constrained resources that must be used judiciously. This necessitates the use of adequate bandwidth- and energy-management algorithms that optimally control the data flows of multiple applications (called missions in this work) for ensuring that all missions obtain adequate volumes of sensor data in a timely and efficient manner. Furthermore, the missions may have different utilities, i.e., one mission may be more valuable or important than another. In such scenarios, the resource adaptation process must ensure that the resources dedicated to a mission's flows is proportional to its utility. The need for such application-level discretion, coupled with the unique challenges offered by mission-oriented WSNs render the existing resource adaptation protocols inadequate.;In this thesis, we address these problems by developing a unified, utility-driven resource optimization framework that aims to regulate the bandwidth and energy usage of flows by dynamically adapting flow characteristics such as rate and degree of compression, in order to meet the real-time data needs of dynamic, competing missions. We develop resource adaptation algorithms that are capable of jointly optimizing under several interesting conditions that are important and unique to mission-oriented WSNs, such as prioritized missions with resource demands, and temporally-dynamic missions and network conditions. Furthermore, the algorithms are fully distributed with practically realizable protocols, which provably improve the performance of the network. As a result, we have a powerful, extensible framework, along with a suite of adaptation protocols, that is capable of proactively and comprehensively optimizing the resource usage in a mission-oriented WSN, while explicitly factoring in the application-level utility of the individual missions.