Predicting smartphone battery usage using cell tower ID monitoring

Personalized power management in mobile devices is a critical issue in handling the diversity of smartphone usage. In particular, usage prediction is important for the efficient use of remaining battery capacity. In this paper, we propose a system that predicts the amount of battery usage required in the future. Our key insight is that a high degree of correlation exists between battery usage and a user's movements. We design an everyday location monitoring system that only uses cell-tower connections, without additional energy consumption. The technical challenge is eliminating the ping-pong effect in a series of cell-tower transitions to determine the mobility status, especially with limited access to the list of neighboring cell towers. We construct a graph from the sequence of recorded cell towers and recognize the points of interest using a partial clique graph. We use the Markov predictor to estimate the required battery level depending on the user's movements. We demonstrate the accuracy of battery usage prediction using real traces of participants collected over a period of four weeks. The result shows that the proposed system correctly predicts the battery usage of smartphones, with an 8.1±7.5% margin of error.