Battery-Aware Transformations in Mobile Applications

Author(s): J. Cito, J. Rubin, P. Stanley-Marbell, M. Rinard
Venue: Automated Software Engineering
Date: 2016

Type of Experiement: Case Study
Sample Size: 30
Class/Experience Level: Other
Participant Selection: Most popular applications
Data Collection Method: Observation


This study focused on balancing the desires of mobile application users and developers. In this study, the desire of mobile application users was prolonging the battery life of their device, while the desire of mobile application developers was maintaining mobile advertisements and analytics (A&A) services. These services are crucial to developers as they are a source of revenue and a source of insight into user behavior for developers. However, the presence of these services in current applications are a major contributor to battery drain. The study mentions that removing these “services completely would reduce the energy usage of a mobile device by up to 16%”.

To satisfy both users and developers, the researchers proposed the following solution: adjust the frequency of recurring A&A requests based on the current state of the battery. Their approach involved identifying A&A requests of the same type, which they defined as recurrent A&A requests. As the battery life of the device decreased, the number of recurrent A&A requests would decrease.

In their experiment, they downloaded 30 of the most popular applications available on Google Play and ran each application for 30 minutes with no interaction. They performed this twice, once to determine the amount of battery drained without their battery-aware transformation algorithm and once with it, to determine the difference. The results of their case study displayed that applications after the battery-aware transformation decreased energy consumption by 5.86%.

The study presented the following limitations: the applications tested may not have been representative. The researchers only used applications that could be entered without creating or logging into an account, therefore the applications used in the study are not representative for all applications. Also the researchers did not interact with the applications, this limits the potential findings of further recurrent request patterns.