面向移动服务计算的服务质量随机优化方法研究

With the rapid development of wireless technologies and the increasing popularity of mobile devices, Mobile Service Computing has been developed and widely used in People's daily life. The guarantee and optimization of Quality of Service (QoS) is an important research problem for Mobile Service Computing. However, the QoS optimization for Mobile Service Computing faces severe challenges. Firstly, there are a lot of uncertain factors in the mobile service environment, such as the unstable wireless communication quality, the mobility of users, and the burst arrival of service requests. The distributions of some random variables are non-typical and hard to predict precisely. It is quite difficult to formulate or solve the QoS optimization problem. Secondly, as services are widely used and the number of mobile devices is increasing dramatically, the scales of services and service systems are becoming larger and larger. It is critical to propose efficient solutions for the QoS optimization problem. This project conducts research on the stochastic optimization of QoS for Mobile Service Computing. (1) We study the modeling theory of stochastic optimization for QoS, propose the methods based on Approximate Dynamic Programming and Q-learning to solve the unconstrained optimization problem, and the methods based on transforming the dynamic problem to static problem to solve the constrained optimization problem. (2) We design efficient optimization methods and techniques based on problem partition to sub-problems, goal softening and approximation. This project is expected to expand the theories in stochastic optimization, provide new methods and techniques of QoS optimization, and offer reference and support for the design and improvement of mobile service system.

随着无线网络技术的发展和移动设备的普及，移动服务计算蓬勃发展并广泛应用于人类日常生产生活。其中，服务质量的保障和优化是重要的研究问题。然而，一方面，移动服务环境中存在无线通信质量的不稳定性、用户的移动性、服务负载的突发性等大量不确定性因素，随机变量存在非典型性分布难预测的情况，对服务质量优化建模和求解造成困难；另一方面，随着服务的广泛应用和移动设备数量的急剧增长，服务大规模特性对服务质量优化方法的效率提出了挑战。本项目围绕面向移动服务计算的服务质量随机优化问题，（1）研究服务质量随机优化建模理论，提出基于近似动态规划与Q-学习的非受限模型求解方法以及基于动静转化的受限模型求解方法；（2）设计和实现基于优化问题分割、目标放缩以及近似化简求解的高效优化方法及技术。本项目研究将丰富随机优化模型理论，促进和推动服务质量优化方法和实用技术的发展，为移动服务系统的设计和改进提供参考与支撑。

无线网络技术的发展和移动设备的普及使得移动服务计算蓬勃发展，服务质量的保障和优化成为重要的研究问题。本项目针对移动服务系统与服务环境的随机性、动态性及大规模特性，从建模理论、分析求解方法及实现技术三个方面出发，开展了服务质量的随机优化理论和方法的研究。针对移动服务系统及移动服务环境的随机性和动态性，研究了服务质量随机优化建模理论及求解方法，建立了形式化数学模型描述服务质量随机优化问题，根据是否受到客观约束条件限制，设计了非受限随机优化求解方法以及受限随机优化求解方法，并从理论上给出问题转换与动静转化的精确度的分析与证明。针对受随机因素影响的大规模复杂服务系统及移动服务环境，研究了高效的服务质量优化方法及实现技术，提出了基于问题分割和模型分块的低复杂度、高效率服务质量优化方法，将优化目标进行放松，形式化证明了优化效率与近似度，提出了求解效率与运行效率权衡的化简求解技术。项目研究成果在真实系统和大规模开源数据集上进行了应用验证。本项目相关研究丰富了随机优化模型理论，促进和推动了服务质量优化方法和实用技术的发展，为移动服务系统的设计和改进提供了参考与支撑。

内容获取失败，请点击重试