摘要
针对嵌入式系统的多任务环境,提出了混合模型功耗管理算法,用于对服从一般分布的系统进行建模。首先,介绍了现有的动态功耗管理策略算法,阐述了算法需要改进的原因。然后,使用重标极差法(Rescaled Range Analysis,R/S)对非平稳服务请求下的时间序列进行长距离相关性分析;根据不同的分析结果选择相应的最大概率策略,即基于电池剩余电量的超时策略、模糊非标准PID策略和半Markov随机策略。最后,给出了策略参数的确定方法并通过实验的方法对本文提出的策略进行分析。实验结果表明,本文策略弥补了常规动态电源管理策略的不足,具有更广泛的适应性;在性能损失10%的条件下,系统平均功耗减少了37%,命中率大于60%,更稳定、有效地降低了功耗,有利于在嵌入式系统中应用。
For multitasking environment of an embedded system,an improved method called hybrid model for power management algorithm was proposed for modeling of system with general distribution.First,the dynamic power management strategy algorithm was introduced,and the reason why it needed to be improved was expounded.Then,the Rescaled Range Analysis（R/S）method was used to analyze the long distance correlation of non-stationary time service requests and the corresponding strategy was selected depending on the different results.These strategies are remaining battery power timeout strategy,fuzzy not quite PID strategy and semi-Markov random strategy.Finally,the method for determining the strategy parameters was given and the strategy proposed in this paper was analyzed experimentally.The experimental results show that this strategy makes up for the deficiency of conventional dynamic power management strategy and has more extensive adaptability.Under the condition of 10% performance loss,the average system powerconsumption is reduced by 37%,and the hit rate is more than 60%.The algorithm reduces power consumption more efficiently than other methods,and is applicable in embedded systems.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2014年第7期1929-1937,共9页
Optics and Precision Engineering
基金
国家科技支撑计划资助项目(No.2009BAE73B01)
关键词
功耗管理
重标极差法
半MARKOV决策过程
策略优化
嵌入式系统
power management rescaled range analysis semi-Markov decision processes policy optimization embedded system
