Author's School

School of Engineering & Applied Science

Author's Department/Program

Computer Science and Engineering

Language

English (en)

Date of Award

January 2010

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Chair and Committee

Chenyang Lu

Abstract

Wireless sensor networks are poised to change the way computer systems interact with the physical world. We plan on entrusting sensor systems to collect medical data from patients, monitor the safety of our infrastructure, and control manufacturing processes in our factories. To date, the focus of the sensor network community has been on developing best-effort services. This approach is insufficient for many applications since it does not enable developers to determine if a system's requirements in terms of communication latency, bandwidth utilization, reliability, or energy consumption are met. The focus of this thesis is to develop real-time network support for such critical applications. The first part of the thesis focuses on developing a power management solution for the radio subsystem which addresses both the problem of idle-listening and power control. In contrast to traditional power management solutions which focus solely on reducing energy consumption, the distinguishing feature of our approach is that it achieves both energy efficiency and real-time communication. A solution to the idle-listening problem is proposed in Energy Efficient Sleep Scheduling based on Application Semantics: ESSAT). The novelty of ESSAT lies in that it takes advantage of the common features of data collection applications to determine when to turn on and off a node's radio without affecting real-time performance. A solution to the power control problem is proposed in Real-time Power Aware-Routing: RPAR). RPAR tunes the transmission power for each packet based on its deadline such that energy is saved without missing packet deadlines. The main theoretical contribution of this thesis is the development of novel transmission scheduling techniques optimized for data collection applications. This work bridges the gap between wireless sensor networks and real-time scheduling theory, which have traditionally been applied to processor scheduling. The proposed approach has significant advantages over existing design methodologies:: 1) it provides predictable performance allowing for the performance of a system to be estimated upon its deployment,: 2) it is possible to detect and handle overload conditions through simple rate control mechanisms, and: 3) it easily accommodates workload changes. I developed this framework under a realistic interference model by coordinating the activities at the MAC, link, and routing layers. The last component of this thesis focuses on the development of a real-time patient monitoring system for general hospital units. The system is designed to facilitate the detection of clinical deterioration, which is a key factor in saving lives and reducing healthcare costs. Since patients in general hospital wards are often ambulatory, a key challenge is to achieve high reliability even in the presence of mobility. To support patient mobility, I developed the Dynamic Relay Association Protocol -- a simple and effective mechanism for dynamically discovering the right relays for forwarding patient data -- and a Radio Mapping Tool -- a practical tool for ensuring network coverage in 802.15.4 networks. We show that it is feasible to use low-power and low-cost wireless sensor networks for clinical monitoring through an in-depth clinical study. The study was performed in a step-down cardiac care unit at Barnes-Jewish Hospital. This is the first long-term study of such a patient monitoring system.

Comments

Permanent URL: http://dx.doi.org/10.7936/K73F4MPK

Download

Share

COinS