A Study Of Power Management Mechanism And Estimation Model For The Android Smartphones

The diversity and social universality of mobile terminals have brought a profound transformation to the development of Internet. Smartphones have been dominant in the market, as a representative of the mobile terminal products. According to market analysts, more than 80% of smartphones install the Android system. Because of the increasing of the user experience requirement, understanding and optimizing the power consumption of smartphones has become an important area of research. The previous work has proposed various power models to estimation the power consumption. However, these models lack granularity and accuracy. In this paper, we study the Android power management mechanism and the construction of power model. The main research results obtained are as follows:1. Researching on the Android power management mechanisms. We analyze the source code implementations of Android power management module, specifically, the wakelock mechanism, power management and estimation mechanisms. The results show that the Android system can’t give the specific power consumption value within a specific running time.2. Aiming at the defects of Android power management mechanism, we put forward the dynamic power estimation model considering the hardware utilization and delay power. The model is based on a nonlinear regression structure. Firstly, we determine the model by making the relevant parameters modular from each hardware components of smartphones. Then we design test cases for identifying the model, confirm the final coefficients concerning the power consumption, and give the specific relational expression. The model has the following advantages. We propose that wireless communication equipment have the issue of delay power, and then we establish the corresponding delay power model. The Flash, Camera and Bluetooth are added to the power model, which makes the existing power model based on hardware utilization more perfect. Therefore the model can achieve the goal to improve the accuracy of the overall power estimation.3. Validating the above power model using the popular smartphones. The results showthat the average absolute error between the actual power and the estimated power is less than 4.6% in common scenarios. Depending on the above model, we have developed an Android power monitoring tool, which can display the system power trends, and acquire the power of the individual hardware components and the overall model in real time.