Research And Implementation On Power Management Policy Based On Android Mobile Devices

With the high-speed development of mobile devices‘ software and hardware, and the popularity of the new generation 3G and 4G network, more and more people use mobile intelligent devices and enjoy their convenient service and leisure entertainment. Since the number of current global smartphone user has grown rapidly, Android has been also dominating the market share of mobile intelligent device system. Because of the high-performance, complicated applications and high level of games, mobile devices face difficulty of higher power consumption. But the development of battery technology is too slow, and the capacity of battery is obviously conditioned to its own volume. So how to reduce the power consumption has become one problem of the development of mobile devices.Dynamic power management has been proved to be an effective system-level technique of low power consumption. The traditional policies include time-out policy, predictive policy and the stochastic model policy, which have their own characteristics and the insufficiency. The Android system also provides its own power management scheme. Since the Android kernel is based on Linux, it also supports Cpufreq kernel subsystem for dynamic power management on CPU. Cpufreq defaults to using Ondemand governor, which will set CPU to the highest frequency when the load exceeds a certain threshold. This radical way fails to maximize energy saving effect. In addition, Android provides a Wakelock-based power management system to make it to enter the sleep state under waiting mode as far as possible. But research shows that the energy consumption of Android mobile device under waiting mode accounts for more than half of total energy consumption. Due to the openness of Android the third-party applications in the background abuse Wakelock resources through the network interface for data transmission.As the change of usage state of resource, the CPU load of Android will change largely, but the traditional predictive policy is difficult to predict effectively. This paper proposes a DVFS policy and its framework based on resource usage state, to dynamically manage the power of CPU more effectively. Based on the analysis of the relevance to the CPU load for foreground application and resource usage state, the policy defines the system state according to foreground application and resource usage state. Then when different foreground application is running, the system state is subdivided into several states. As for every system state, the policy will predict the CPU load in order to adapt to the change of the load. Meanwhile this paper provides the details on how to implement the proposed policy based on Android mobile devices.As for the high-consumption problem of mobile devices under waiting mode, the current methods mainly configure some fixed time and other conditions statically to open and close the network interface. In order to reduce the power consumption under waiting mode more intelligently and effectively, this paper proposes a power management policy and its framework based on context awareness, which makes decisions to predict whether you currently need the network interface according to the context-aware information of mobile devices, and intelligently manage the network interface to reduce the system energy consumption under waiting mode. We preprocess complex context-aware information, and optimize the way to obtain locational data. Decision algorithm is based on the improved K nearest neighbor algorithm. Distance calculation for samples uses the weighted Euclidean distance formula, and information gain value is used to select context-aware attributes and determine the weights of attributes. In addition, we also propose a sample pruned method based on maximum attribute domain to reduce the calculation when making decision. Meanwhile this paper also provides the details on how to implement the proposed policy based on Android mobile devices.Finally, this paper does the experiment and contrasts the experimental data based on Android mobile devices for the two proposed policies respectively. The results show that both the policies are more effective on power saving, which proves the validity of the proposed two policies.