Using Learning to Predict and Optimise Power Consumption in Mobile Devices. By

Extending the life of a battery in mobile systems has always presented a challenge, many researchers have come up with efficient power optimisation methodologies. The complex-ity of modern systems is growing rapidly and such systems cannot be considered as single devices anymore; they must be observed as a system of inter-connected and inter-dependent devices. Applying a single power optimisation technique on a particular sub-system may not be sufficient, it may even have the opposite effect of wasting energy rather than saving energy. The power of the system must be managed entirely and a variety of energy saving techniques must be used to ensure maximum efficiency in the consumption of power at any point of time. Another aspect of modern mobile systems, is the power efficiency of applications running, which lies purely with the developer. Unfortunately development en-vironments and frameworks do not provide a facility to profile an application for debugging power inefficiencies. With this in perspective, the goal of the project is to for automated runtime power management and to provide a framework for power consumption testing. To achieve the goal, we present PowerRunner an automated power management softwar