Design And Implementation Of Low Power Android Mobile Sink Based On CPU Load Prediction

In the Wireless Sensor Networks (WSNs), mobile sink node can help resolve the problem of energy holes, and can use the sensor nodes in WSNs more reasonably by taking an optimized move path. In this way, mobile sink node will extend the lifetime of WSNs. While the mobile sink node, on which has power hungry modules like LCD and coordinator node, is powered by battery, so it is significant to manage the power hungry modules effectively by operating system software to reduce power consumption.As a very popular operating system in mobile device field, Android has a positive power management strategy, which guarantees that the most power hungry models on mobile device will be shut down the first time they are idle. This positive strategy will lower down the power consumption as soon as possible. Besides the Android applications are all running on Dalvik virtual machine, which gives the Android application a very strong plantation character.For the advantages of Android operating system, the mobile Sink node based on Android will get the same advantages. However, after a serious analysis of the power management mechanical in Android operating system kernel, we find out that the standard Android operating system has only three power states (AWAKE, NOTIFICATION and SLEEP), and it does not provide a good enough fine-grained power management mechanical for specific mobile device. Besides, the CPU load prediction algorithm in Android kernel is PAST, which will not predict the CPU load precisely. In order to reduce the power consumption of mobile sink node based on Android operating system, this paper will improve the standard Android system in two aspects:The first one is that a new state—PER NOTIFICATION—is added to the standard Android system, which has only three power states (AWAKE, NOTIFICATION and SLEEP). Under the state of PER NOTIFICATION, USB devices on the mobile sink node can be added to or removed from the system by the sys file interface according to users’specific demand without plugging in or out manually, realizing a more fine-grained power management.The second one is that an improved CPU load prediction algorithm called Self Adaptive Weight Dynamic Linear Prediction (SAWDLP) is proposed to replace the default one in ondemand governor, which is the default governor in Android operating system. Compared to the PAST algorithm, SAWDLP takes one more history CPU load and the weight of history loads can vary with the history loads to keep it optimal. Through this algorithm, a more accurate CPU load prediction can be done and then a more appropriate frequency can be chosen by DVFS technology to reduce the CPU power consumption.Finally, experimental results show that the PER_NOTIFICATION works well in managing the USB devices. And SAWDLP algorithm gets 85.20% higher accuracy in CPU load prediction and 0.40% lower power consumption of the whole mobile sink in comparison to PAST algorithm, accomplishing the goal of low power design and run time extension.