中文摘要：

针对移动用户将应用迁移至云端处理会引起大量的数据传输导致高能耗的问题,提出了一种任务联合执行策略（Task Collaborative Execution Policy,TCEP）.首先,在云端和移动终端联合执行移动应用的前提下,将应用考虑为一系列的串行任务,根据任务的计算负荷、输入和输出数据量,把云端与移动终端联合移动应用的优化问题建模为最小化移动终端的能耗问题,并得出结论该优化问题属于NPC（Nondeterministic Polynomial Complete）问题.接着,按移动终端需向云端迁移任务的次数来划分迁移策略集,并利用串行任务仅能一个接一个地执行的特点,给出了一次迁移最优特性.然后,通过对比串行任务与染色体的相似点,采用遗传算法来处理文中优化问题,并在简单遗传算法（Simple Genetic Algorithm,SGA）的基础上,利用一次迁移最优特性来设计交叉操作和变异操作,以便进一步提高算法性能.最后,通过仿真验证了所提策略及算法的性能,仿真结果表明,改进后的遗传算法具有良好的收敛性能,能够保证新个体具有仅向云端迁移一次的特性,与现有方法相比,所提策略可有效地减少搜索最优解的运算时间,能在满足应用执行时间要求的同时最小化移动终端的能耗.

英文摘要：

The user can save the energy consumption on Mobile Devices （MDs ） by offloaded mobile application to cloud, but it adds the transmission energy consumption of MDs. Focusing on this problem, a Task Collaborative Execution Policy （TCEP） is proposed in this paper. Firstly, mobile applications are considered as a series of tasks which can be executed by cloud or MD. Using the calculated load, the amount of input and output data of the task, we model the optimization problem as energy consumption minimization of MD for scene which the cloud and the MD jointly carry outa mobile application. Based ontheanalysis, onecandrawaconclusionthat the optimumenergy consumption of MD belong to NPC（Nondeterministic Polynomial Complete） problem. Next, considering the mobile application be jointly executed by the cloud and the MD, we divide the policies class into various set of the migration policy based on the number of offloading.According to the serial task just be executed one by one, we obtain the optimal feature of application offloading. Then, considering the serial task similar to chromosome, we turn to genetic algorithm solving the optimumenergy consumption. Using this offloading property, we improved the performance of genetic algorithm by designed the crossover operation and the mutation operation.At last the simulation validates the feasibility of the proposed policy and algorithm. Simulation results show that the proposed algorithm can ensure the new individual executing offloading no more than one time. Compared with existing methods, the proposed method can effectivelyimprove the rate of convergence and reduce the operation time. Moreover, it can significantly save the energy consumption on the MT while meeting the application deadline.