WiFi centric power modeling of smartphones

Previous work of WiFi power modeling proposes the powerthroughput curve [2], the power-transmission/reception airtime curve [1], or the state transition between high power level and low power level according to the packet rate [3]. These models cannot accurately estimate the energy consumption of the smartphone because they either do not take into account the power reduction from the power saving operation of the components in the smartphone or model the active power level too simplistically, thus resulting in larger estimation error. The smartphone, which is one of the battery sensitive devices, essentially has power saving algorithms in its components to reduce energy consumption by turning off the components when they are not in use. In this sense, two types of traffic, which turn out to be the same average throughput, airtime, or packet rate, may result in different total energy consumption if one does not allow the smartphone to turn off its components but the other allows it. We propose a packet interval based power model which takes into account the power saving operations of the WiFi module so that it accurately estimates power level of the smartphone when the smartphone transmits/receives packets using its WiFi interface. Furthermore, we manage the power state of the smartphone with the Finite State Machine (FSM) to accurately estimate the power level of the smartphone considering other components in the smartphone, e.g., CPU and display.