Power Management of Smartphones Based on Device Usage Patterns

Abstract

Smartphones provide rich applications and offer many crowdsensing services to end users. However, the power consumption of smartphones is still a primary issue in green computing. This paper presents a general power management framework including a data logger, an unsupervised learning algorithm of classifier, and a power-saving decision maker. The framework gathers and analyses the usage patterns of smartphones and separates end users into non-active and active ones, and their mobile devices into low-power ones and high-power ones using an unsupervised learning algorithm. If the smartphone of a non-active user belongs to the high-power group, we observe abnormal usage behaviour. The framework provides recommendations, e.g. as a power-saving notification to the user. We collected device usage and power consumption attributes on two kinds of Android–smartphones and evaluated the framework in experimental studies with 22 users. The results show that our framework can correctly identify non-active users’ devices consuming much more power than active users by recognizing reasons of the abnormal usage behaviour of non-active users and providing recommendations for adjusting the device towards power saving.