未来异构云网络中的自优化智能决策模型研究

In the future heterogeneous cloud networks, it is a popular and urgent open problem of how to realize effective and efficient self-optimization decision-making technology, in order to meet the demands from the complicated environment and the users, improve the resource utility and decrease the operation cost. This project will tackle this problem by setting up a layered intelligent model, with the help of machine learning theory, and the consideration on computing resources and storage resources as well as the traditional communication resources. The research targets are as follows: 1) Research on the overall architecture and function of the proposed model, which is divided into separate macro and micro decision-making layers. This could suit different optimization scale, granularity and goals. 2) In the macro decision-making layer, propose the macro self-optimization decision-making strategy based on joint multiple objects. 3) In the micro decision-making layer, propose the micro self-optimization decision-making algorithm based on complex environment. By the thorough research on the proposed model, the communication scenario could be sensed in real time, the parameters could be adjusted intelligently and the resources could be allocated effectively. The user experience and system performance would be enhanced for the overall heterogeneous cloud networks.

在未来异构云网络中，为了应对复杂多变的通信环境和用户需求，提高资源利用效率，降低运营成本，如何实现高效灵活的网络自优化决策，已成为信息通信领域广受关注的前沿开放性问题之一，亟待深入研究。对此，本课题拟创建适用于未来异构云网络的自优化分层智能决策模型，采用机器学习理论的先进成果，将传统通信资源和计算资源、存储资源纳入统筹考虑，实现以下研究目标：1）研究该模型的整体结构和功能模块，拟分为宏观决策和微观决策两层，分别适用于不同的网络范围、优化任务和调控粒度；2）在宏观决策层，提出基于多目标联合的自优化宏观决策机制；3）在微观决策层，提出适用于复杂环境的自优化微观决策算法。本课题拟通过对该模型的深入研究，最终实现通信环境实时感知、参数调整智能决策、网络资源高效配置，从而全面提升未来异构云网络的用户体验和系统性能。

为了应对未来异构云网络中复杂多变的通信环境和用户需求，如何实现高效灵活的网络自优化是亟待解决的重要问题。本课题的研究紧密围绕未来异构云网络中的自优化决策展开，主要工作如下：首先，将移动边缘计算、区块链和无线能量收集等技术引入未来异构云网络中，结合机器学习等先进算法，研究适用于不同典型场景的自优化智能决策模型。然后，针对具体的优化目标和限制条件，研究数学建模与理论分析方法，在此基础上设计新型的基于多目标联合的自优化宏观决策机制和复杂环境下的自优化微观决策算法，实现高效灵活的网络自优化分层智能决策。最后，对提出的算法进行仿真验证，对仿真结果展开深入分析讨论。研究成果表明，在未来异构云网络的典型场景中，所提算法提升了系统性能和资源利用率，为保证优质的用户体验和高效的网络传输提供了具有重要参考意义的解决方法。