Center for Embedded Network Sensing

Large and bright screens on today's mobile phones account for significant energy demand on phones' batteries. In this paper we propose an algorithm that, given the energy profile of the screen, finds the optimal schedule to minimize screen energy dissipation when the phone is idle. We profile the screen energy consumption of two popular smartphones, Nokia N95 and E71, through carefully designed micro-benchmarks. Our energy measurement results suggest that the default screen schedules on these phones are far from optimal - on average our algorithm performs 50% better than default. We also find that on the E71 not using the dim state of the screen and directly turning it off is more energy-efficient. We improve the performance of our screen scheduling algorithm by considering the history of each user's interaction with his/her phone. We study the interaction patterns of six volunteers with their smartphones. The results suggest that the distribution of the length of idle times for each user does not change over time. Therefore, the screen scheduler can learn this distribution during a learning phase and use it to improve screen management. We show that the screen energy consumption can be further reduced by up to 60% using this technique.