基于遗传算法的三单播网络对齐及优化

Network alignment and optimization for three unicast network based on genetic algorithms

针对三单播有向无循环网络,考虑无法得知网络内部拓扑图结构,且内部节点数量无法估量的情况。应用渐近干扰对齐技术研究各用户发送不同数据流m,n,p时的预编码矩阵,联合应用图论中的最短路径遗传算法和网络线性性质,分析其预编码网络对齐(precoding-based network alignment,PBNA)的可行性条件,该条件最终可简化为2个方程,实际检验复杂度较低,且每对单播会话能够达到1/2的自由度。

This paper considers the 3-unicast problem of directed acyclic network,the internal network topology structure not being known,and the number of internal nodes immeasurable. When different data streams m,n,p are sent by each user,these precoding matrixes are studied by using the asymptotic interference alignment technique,further,the set of feasibility conditions for precoding-based network alignment（ PBNA） can be reduced to 2 by using two shortest path genetic algorithm and graph-related properties. Significantly reduced computational complexities are obtained and achieved 1 /2 degrees of freedom（ DOF） of the individual source-destination.