User support for battery management of continuous sensing applications

Abstract

Continuous sensing applications have been increasingly emerging as a new class of mobile applications. In the background, they continuously monitor diverse user contexts such as location, physical activity, emotion, and conversation. Continuous sensing applications provide what users need in the right place at the right time. Due to continuous sensing and processing, it is inevitable to persistently consume the significant amount of power of a smartphone. To overcome the problem of the excessive power consumption, extensive efforts have been performed from diverse perspectives which mainly develop system-level supports to help save the energy consumption. However, these supports do not fully address power issues of continuous sensing applications, especially in terms of users’ battery management. Even with such system supports, continuous sensing applications still consume a considerable amount of the power in the background, thereby causing unexpected, sudden battery depletion on users’ smartphone. In this dissertation, I newly propose a comprehensive approach to address users’ battery concerns of continuous sensing applications. Its key idea is to provide users with power impact of continuous sensing applications and help them manage their phone’s battery themselves. Actually, users are the most important stakeholders for battery management of smartphones. They are those who experience inconvenience with the low battery and can also proactively take direct action prior to the battery depletion, e.g., recharging the phone in advance, minimizing the phone use, and selecting an energy-efficient continuous sensing application. Thus, if users are well aware of the battery drain of continuous sensing applications, they can better manage their phone’s battery. First, I address the new battery-impacting factors of continuous sensing applications which are prominent enough to outdate users’ existing battery model in real life. I explore an initial approach to help them count the new cause and effect between their physical activity and their phones’ battery lives. I expose the macroscopic impact that today’s commercial sensing applications have already been exhibiting on our smartphone battery experiences. I present Sanda, a novel mobility-aware smartphone battery information advisor and study its potential to help users redevelop the battery models in their mind with the new factors of continuous sensing applications. Second, I present PowerForecaster, a system to provide an instant, personalized power estimation of a continuous sensing application at pre-installation time. It provides power estimation at pre-installation time, removing the hassle of installing and using applications one after another until users find power-efficient applications. The estimation is highly personalized to reflect an individual user’s activity and phone use pat-terns, achieving higher estimation accuracies. It also supports a wide-variety of sensing applications without requiring any changes in the applications or additional information from developers. Finally, I introduce current practices of battery use and management for smartwatches. With processors, a screen, sensors, communication modules, and a battery well-integrated in small form factors, smart-watches have much potential to provide new user experiences and enable enriched continuous sensing applications to run in real life. However, despite their potential, we know little about the current practices for battery use and management of smartwatches. I conduct the user studies targeting real smartwatch users and discuss key findings and the implications for continuous sensing applications.