DATE: Thursday, November 30, 2000
TIME: Noon - 1 pm
PLACE: Hammerschlag Hall D210

Dushyanth Narayanan

Ph.D. Student
Computer Science Department, CMU

Using History to Improve Mobile Application Adaptation

Mobile interactive applications such as augmented reality, speech recognition, and language translation make heavy demands on system resources such as CPU, network bandwidth, and battery energy. At the same time, users demand quick interactive response and long battery life.

An effective way to achieve low latency and long battery life is to adapt application fidelity to the available resource levels. This is one of the core ideas in Odyssey, a platform we are building for mobile applications. Effective fidelity adaptation requires us to know the resource consumption of the application at any fidelity. In this talk, I describing the work we have done in Odyssey to monitor, log, and learn application resource consumption as a function of fidelity.

I will briefly describe our "logging and learning approach" and provide some preliminary results. I'll conclude with a look at future work: how we can improve prediction accuracy, and what other capabilities Odyssey will need to support truly effective adaptation.

Dushyanth Narayanan is a 6th year Ph.D. student in Computer Science. He works on the Odyssey mobile computing project.

For Further Seminar Info: