Smartphones power

Power modeling and management techniques in portable devices have become major design concerns in recent years; rapid advances in chip design and hence their power requirements and slow advances in battery technologies forced designers to focus on power reduction rather than battery technology. Our work is an attempt to develop user-behavior based power models for smartphones where the premise is the power consumed by the device is directly related to its user´s activities which in essence constitute the workload. This can be of great assistance to the designers of smartphones´ hardware and software; having a user-driven power model of a device will pave the way for an optimal design. We collected users´ related data using an in-house logging tool and a selected set of parameters is identified to develop a Neural Network (NN) based power model. Many trials are conducted in order to identify the suitable NN structure and training algorithm. Results demonstrate that NNs models can provide suitable platforms for studying the energy usage of smartphones.