认知超密集网络用户关联与资源分配联合优化遗传算法

针对下行的异构认知超密集异构网络（UDN）的多维资源配置问题，提出一种以毫微微小区用户最大吞吐量为目标的联合优化用户关联和资源分配的改进遗传算法。首先，在算法开始之前进行预处理，初始化用户可达基站和可用信道矩阵；其次，采用符号编码，将用户与基站以及用户与信道的匹配关系编码为一个二维的染色体；然后，将动态择优复制+轮盘赌作为选择算法，以加快种群的收敛；最后，为避免算法陷入局部最优，在变异阶段加入早熟判决的变异算子，从而在有限次迭代下求得基站、用户、信道的连接策略。实验结果表明，在基站与信道数量一定时，所提算法与三维匹配的遗传算法相比在用户总吞吐量方面提高了7.2%，在认知用户吞吐量方面提高了1.2%，且计算复杂度更低。所提算法缩小了可行解的搜索空间，能在较低复杂度下有效提高认知UDN的总吞吐量。

Joint optimization of user association and resource allocation in cognitive radio ultra-dense networks to improve genetic algorithm

Aiming at the multi-dimensional resource allocation problem in the downlink heterogeneous cognitive radio Ultra-Dense Network （UDN）， an improved genetic algorithm was proposed to jointly optimize user association and resource allocation with the objective of maximizing the throughput of femtocell users. Firstly， preprocessing was performed before the algorithm running to initialize the user’s reachable base stations and available channels matrix. Secondly， symbol coding was used to encode the matching relationships between the user and the base stations as well as the user and the channels into a two-dimensional chromosome. Thirdly， dynamic choosing best for replication + roulette was used as the selection algorithm to speed up the convergence of the population. Finally， in order to avoid the algorithm from falling into the local optimum， the mutation operator of premature judgment was added in the mutation stage， so that the connection strategy of base station， user and channel was obtained with limited number of iterations. Experimental results show that when the numbers of base stations and channels are fixed， the proposed algorithm improves the total user throughput by 7.2% and improves the cognitive user throughput by 1.2% compared with the genetic algorithm of three-dimensional matching， and the computational complexity of the proposed algorithm is lower. The proposed algorithm reduces the search space of feasible solutions， and can effectively improve the total throughput of cognitive radio UDNs with lower complexity.