项目名称: 基于云计算的动态分布式多目标粒子群算法研究
项目编号: No.61503086
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 自动化技术、计算机技术
项目作者: 陈霓
作者单位: 广东工业大学
项目金额: 20万元
中文摘要: 本项目针对传统多目标粒子群算法在求解大规模复杂多目标优化问题时存在的求解速度过慢及解集质量劣化的问题,提出通过问题分解技术和分布式处理以加快算法求解速度,通过多种群协作提升解集质量的解决方案。并利用云计算这一先进计算模式,开展云计算环境下的多目标粒子群算法的研究。首先,针对分布式多目标进化算法中种群维护策略设计的问题,提出一种基于问题-子问题的二阶结构组织子种群,从而提升算法的全局搜索能力和解集质量。然后,结合云环境的动态特性,提出了基于解集分布状态感知的参数和拓扑结构自适应控制策略,以增强算法的普适性,提高算法求解能力。最后,结合种群协作策略和自适应机制两方面的的成果,实现基于云计算的高性能多目标粒子群算法,并在运筹调度领域的大规模工程实践问题中对算法进行应用检验。本项目的研究为解决大规模多目标优化问题提供了新型而高效的途径。
中文关键词: 多目标优化;大规模优化;规划与调度
英文摘要: Many large-scale multi-objective optimization problems emerge in the practical applications. Traditional centralized multi-objective particle swarm optimizers (MOPSO) are subject to slow convergence speed and degenerated solution quality. To solve the above problems, this project proposes a distributed multi-population MOPSO algorithm in the cloud computing platform. First, this project makes researches into the cooperation mechanism of sub-populations for the distributed MOPSO, and proposes a double-level-based srategy for population cooperation in order to enhance the solution quality; then proposes the adaptation strategy for distributed MOPSO, so as to enhance the scalability of the algorithm and utilize the dynamic resource in the cloud environment; and at last makes researches into the application of distributed MOPSO to a large-scale scheduling problem in real-world applications. This project is to provide an efficient way for solving large-scale and complex multi-objective optimization problems.
英文关键词: multi-objective optimization;large-scale optmization;programming and scheduling