Research Of Energy Efficient Human-Mobile Web Interactions Based On Event Rate Learning

Smartphones play an important role in the age of Mobile Internet.As the computational ability of smartphones continues to improve,extending the short battery life has become an important problem for smartphones.Since the battery technology cannot achieve the revolutionary improvement within a short period of time,improving the energy efficiency of mobile apps has become the common approach to extend the smartphones' battery life.The mobile browser which is one of the most popular mobile apps,can be considered as the information portal for smartphones.However,it suffers from the energy inefficiency during the runtime.Recently,a number of techniques have been proposed to optimize the power consumption during the process of webpage loading including webpage downloading,parsing,and rendering,while the power consumption of mobile Web interactions,especially after the webpage has been loaded,has received comparatively little attention.Through an empirical study on commodity smartphones,we observe that the power consumption of smartphones during the scrolling and pinching operations can be increased by 250%-350%.We analyze the source codes of mobile browsers and motivation experiments,and find that the power consumption of scrolling and pinching operations is proportional to the interaction event rate.As simply reducing the interaction event rate will compromise the user experience of mobile Web interactions,we propose a quadratic model and a linear model as user preference models to adaptively adjust the interaction event rate of scrolling and that of pinching operations,respectively.To enhance the user experience of interaction event rate adjusting,we design an interaction event rate learning strategy based on Support Vector Regression(SVR).Through collecting user preference data periodically,we provide users with personalized interaction event rate models.To validate the effectiveness of our proposed interaction event rate learning strategy,we implement eBrowser based on the Chromium open-source project.eBrowser is an energy-efficient mobile Web interaction framework which comprises a remote cloud side and a local browser side.Our detailed experiments are designed as follows:(1)we evaluate the energy saving of eBrowser by measuring the power consumption during browsing different webpages.(2)We evaluate the user experience of eBrowser by collecting user experience ratings.(3)We evaluate the performance overhead of eBrowser by measuring the network traffic,computational delay,and power consumption during one round of model training.We conduct extensive real-world experiments with 100 mobile users,and the experimental results show that eBrowser reduces the power consumption of mobile Web browsing by up to 43.8%with negligible runtime overhead,while guaranteeing user experience of mobile Web interactions.