基于遗传算法的分层覆盖网络组播树构建

A Layered Multicast TreeModel Based on Genetic Alogrithem

依据NICE网络协议组织网络节点，提出一种分层覆盖网络组播树模型，其基于K-Mediods和遗传算法对组播服务节点MSNs选择，构建覆盖网络分层组播树。根据覆盖网络中组播服务节点的特点，引入基因差异控制和变异精英控制策略，对遗传算法中的交叉和变异算子进行修正，限制适应度差的个体生成，在缩小搜索空间、加快收敛速度的同时，提高算法的全局寻优能力。理论分析和仿真结果也表明，该模型不仅有效克服了传统K-Mediods算法模型易陷入局部极小值的特点，而且明显避免了对初始中心选值敏感的问题。

After analyzing the characteristics of overlay network, a source-specified multicast tree with hierarchy structure is built to manage the multicast service nodes. A layered multicast tree model （LMTM） is proposed based on the algorithms of K-Mediods and genetics, which is used to choose the multicast service nodes （MSNs） among network nodes. Compared with the traditional K-Mediods model, LMTM not only avoids converging to local minimum value, but also is robust to initialization. Also, during the evolution, according to actual features of MSNs, the evolutional control schemes is used to enhance the local search ability, and increase the convergent speed.