Date of Award

Spring 1-1-2014

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical, Computer & Energy Engineering

First Advisor

Li Shang

Second Advisor

Qin Lv

Third Advisor

Michael P. Hannigan

Fourth Advisor

Alan Mickelson

Fifth Advisor

Gary Phillis

Abstract

Personal computing devices have gone through more than four decades of evolution. The form factor of computing devices has gone down dramatically, while computation workload for each computing device has increased significantly. Moreover, people spend much more time interacting with computing devices than before. In short, everyone has a pocket supercomputer.

Delivery of intelligent services today is only possible thanks to rich context information from wearable devices, an emerging personal computing platform. Power efficiency is one of the determinant factors for the adoption rate of wearable devices, as people expect it to work 24/7.

Most existing works often pay careful attention to the energy and processing cost from the component level and show significant power efficiency gain by utilizing device level power management. Even with an abundance of such work, power optimization on wearable devices remains as an open problem, and available solutions only manage to provide weeks of continuous user experience before a battery recharge. To this end, this dissertation systematically measures and quantifies power characteristics of a wearable computing system, then explores adaptive approaches to best balance power consumption and user experience. A methodological application agnostic, yet practical, power optimization framework built upon adaptive sensing and communication management, is proposed and tested in several research projects. On top of that, we propose Gazelle, a personalized running analysis wearable system. Gazelle has been tested in the real world on more than fifty users over a year. It provides one order of magnitude better battery life comparing to other commercialized wearable platforms. The methodology and proposed framework can be readily extended to other application dependent wearable computing platforms.

Share

COinS