基于半Markov决策过程的手机节能策略研究
发布时间：2018-10-10 10:16
【摘要】：诸如智能手机、平板电脑等智能移动设备已成为人们日常生活中的必需品。随着智能手机上配备更多的硬件模块,开发者使用它们开发出各类软件,以最大限度提升用户使用体验。高能耗是此类软件所要考虑的重要因素。然而,当前电池密度的增长相对缓慢,电池技术无法满足移动设备的重度使用。为了解决这个问题,已有许多能耗管理技术被用来设计能量高效的移动系统。当前电源管理方案主要存在的问题包括过分重视单个子系统而忽视全局优化,占用大量计算资源,云技术泄漏用户隐私的风险等。由于电量是移动设备的一种极度稀缺资源,电源能耗管理已成为移动设备系统中必不可少的组成部分。针对当前电源管理方案的不足,本文提出了一种基于半Markov决策过程的动态电源管理方案。该方案基于半Markov决策过程建模,同时考虑了用户体验和设备能耗,与Boe方案相比需要更少的状态和计算时间。为完善模型,通过设备Power monitor监测华为G610-T00智能手机的使用状态并记录不同模块的功耗,设计Android控制软件以调节GPS更新速率和LCD屏幕亮度,根据开源软件Power Tutor二次开发以记录用户使用数据。此外本文给出了两种在线算法,Q学习和策略梯度估计算法,并分析了仿真结果。为了比较所得策略,制定了多种评估准则,以此验证节能算法的有效性。本文通过仿真验证了所建模型的可行性和节能算法的有效性。仿真结果显示本方案的最优策略与常用策略相比能延长53%的使用时长,增加51%的总用户体验。所给出的策略梯度估计算法能在不使用云端技术条件下更新策略,计算量小,不依赖于模型参数,同时用户可自由设置策略更新时间,便于实现。本文所提出的动态电源管理方案对现实手机节能有积极的指导意义。
[Abstract]:Smart mobile devices such as smartphones and tablets have become a daily necessity. With more hardware modules on smartphones, developers use them to develop software to maximize user experience. High energy consumption is an important factor to be considered in such software. However, the current cell density growth is relatively slow, battery technology can not meet the heavy use of mobile devices. To solve this problem, many energy management techniques have been used to design energy-efficient mobile systems. The main problems of the current power management scheme include paying too much attention to a single subsystem and neglecting global optimization, occupying a large amount of computing resources, and the risk of cloud technology leaking users' privacy and so on. Power consumption management has become an essential part of mobile device systems because of the fact that power consumption is an extremely scarce resource for mobile devices. A dynamic power management scheme based on semi-Markov decision process is proposed in this paper in view of the shortcomings of the current power management scheme. This scheme is based on semi-Markov decision process modeling, taking into account the user experience and device energy consumption. Compared with the Boe scheme, it requires less state and computing time. In order to perfect the model, Huawei G610-T00 smartphone is monitored by device Power monitor and the power consumption of different modules is recorded. The Android control software is designed to adjust the update rate of GPS and the brightness of LCD screen. Second development based on open source software Power Tutor to record user usage data. In addition, two online algorithms, namely, Q learning and strategy gradient estimation, are presented, and the simulation results are analyzed. In order to compare the strategies, a variety of evaluation criteria are developed to verify the effectiveness of the energy-saving algorithm. The feasibility of the proposed model and the effectiveness of the energy saving algorithm are verified by simulation in this paper. The simulation results show that the optimal strategy can prolong the service time by 53% and increase the total user experience by 51%. The proposed strategy gradient estimation algorithm can update the policy without using cloud technology. It has less computation and does not depend on the model parameters. At the same time, the user is free to set the policy update time, which is easy to implement. The dynamic power management scheme proposed in this paper has positive guiding significance for practical mobile phone energy saving.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.53
本文编号：2261401
[Abstract]:Smart mobile devices such as smartphones and tablets have become a daily necessity. With more hardware modules on smartphones, developers use them to develop software to maximize user experience. High energy consumption is an important factor to be considered in such software. However, the current cell density growth is relatively slow, battery technology can not meet the heavy use of mobile devices. To solve this problem, many energy management techniques have been used to design energy-efficient mobile systems. The main problems of the current power management scheme include paying too much attention to a single subsystem and neglecting global optimization, occupying a large amount of computing resources, and the risk of cloud technology leaking users' privacy and so on. Power consumption management has become an essential part of mobile device systems because of the fact that power consumption is an extremely scarce resource for mobile devices. A dynamic power management scheme based on semi-Markov decision process is proposed in this paper in view of the shortcomings of the current power management scheme. This scheme is based on semi-Markov decision process modeling, taking into account the user experience and device energy consumption. Compared with the Boe scheme, it requires less state and computing time. In order to perfect the model, Huawei G610-T00 smartphone is monitored by device Power monitor and the power consumption of different modules is recorded. The Android control software is designed to adjust the update rate of GPS and the brightness of LCD screen. Second development based on open source software Power Tutor to record user usage data. In addition, two online algorithms, namely, Q learning and strategy gradient estimation, are presented, and the simulation results are analyzed. In order to compare the strategies, a variety of evaluation criteria are developed to verify the effectiveness of the energy-saving algorithm. The feasibility of the proposed model and the effectiveness of the energy saving algorithm are verified by simulation in this paper. The simulation results show that the optimal strategy can prolong the service time by 53% and increase the total user experience by 51%. The proposed strategy gradient estimation algorithm can update the policy without using cloud technology. It has less computation and does not depend on the model parameters. At the same time, the user is free to set the policy update time, which is easy to implement. The dynamic power management scheme proposed in this paper has positive guiding significance for practical mobile phone energy saving.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.53
【参考文献】
相关期刊论文 前2条
1 章坚武;袁杰;;基于3G背景流量整形的节能方法[J];电信科学;2014年12期
2 梁桂才;;基于云代理的智能手机3G通信端口节能模型[J];计算机与现代化;2014年06期
，本文编号：2261401
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2261401.html
