Predictive Power Conservation

点击翻译按钮获取中文摘要

The goal of this project is to investigate the use of adaptive techniques to minimize the power consumed by peripherals in mobile computing systems. Since battery life is one of the most critical resources in the mobile computing environment, good energy conservation methods can dramatically increase the utility of mobile systems. The investigators have previously demonstrated that new adaptive techniques from the machine learning community are very effective in solving disk spin-down problem (deciding when to shut down the disk in order to maximize power savings) on traces of disk activity. The project will continue this work in two directions. The first direction is to compare the performance of the new techniques with that of other adaptive algorithms for the spin-down problem. The second direction involves applying similar machine learning methods to wireless modems (radios). In addition to the power savings gained by turning off the radio when the network interface is not needed, adaptively choosing the broadcast power level may also lead to significant energy savings.