网格环境下分层并行多群体协作PSO框架设计与实现

针对利用广域范围内的计算资源参与PSO执行，从而提高工程最优化问题计算效率并降低计算成本，提出一个网格环境下分层并行多群体协作PSO（G-LPMCPSO）框架。首先给出一个适应负载不均衡和计算资源异构网格环境下的并行多群体协作PSO（PMCPSO）算法；然后着重阐述了如何利用标准的网格技术和PMCPSO算法设计并实现G-LPMCPSO框架，该框架具有一个扩展的GridRPC API用于隐藏网格环境的复杂性和一个元任务调度器用于无缝的资源发现和选取；最后，根据理论分析及实验结果，证明利用网格技术及PMCPSO可以提供一个可靠的框架用于加速解决科学工程最优化问题。

Design and realization of layered parallel multi-swarm cooperative PSO framework under grid environment

Concerning how to utilize computational resources in large field to join the implementation of Particle Swarm Optimization(PSO), thereby the computational efficiency can be enhanced and the computational cost can be reduced, the Grid-based Layered Parallel Multi-swarm Cooperative PSO(G-LPMCPSO) Framework was presented in this paper, as well as a Parallel Multi-swarm Cooperative PSO(PMCPSO) algorithm adapted to grid environment of load imbalance and the heterogeneity of computational resources. And then, how to make use of standard Grid technologies and PMCPSO algorithm to design and realize G-LPMCPSO framework was discussed in detail. The framework has an extended GridRPC API to conceal the high complexity of the Grid environment, and a meta-scheduler for seamless resource discovery and selection. At last, the theoretical analysis and the result of experiment indicate that the proposed G-LPMCPSO using Grid can offer a credible framework for providing a significant speed-up to optimization in science and engineering.