随着各种新的无线技术的广泛应用，无线电频谱成为了当今社会最宝贵的资源之一。现今频谱平均使用率相当低并且使用分布不均匀。认知无线电技术能够在不影响授权用户的条件下允许非授权用户使用授权频谱，可显著提高频谱利用率。频谱决策技术根据用户QoS需求分配最佳频谱资源，是认知无线电技术的重要组成部分。

本文对树形认知无线电分簇子网模型进行了研究，设计了该子网模型的路由策略和动态频谱管理方式。该子网模型采用多簇并行方式工作，由子网中主站进行集中式频谱分配。针对分簇子网提出了频谱决策问题，该频谱决策涉及到了子网容量、吞吐量和子网稳定性三方面因素。

针对子网容量最大化和子网速率最大化两项QoS，提出了子网拓扑搜索算法，该算法基于簇结构和簇生长度构造无重复的搜索空间，引入子网容量动态门限进行剪枝。接下来对子网拓扑搜索算法进行了时间性能改进，提出了时限子网拓扑搜索算法和快速子网拓扑搜索算法。时限子网拓扑搜索算法通过限制子网拓扑搜索算法的迭代次数以提升算法实时性，快速子网拓扑搜索算法引入了在搜索空间中动态调整搜索路径的方法，极大地提高了子网拓扑搜索算法的实时性。

本文分别对子网拓扑搜索算法、时限子网拓扑搜索算法和快速子网拓扑搜索算法在相同的仿真参数下进行了实验，对比分析了三个算法的时间复杂度。实验结果表明快速子网拓扑搜索算法在相应频谱空间和子网规模等约束条件下，能够获得最优解且满足实时性需求。最后总结全文并且给出了未来进一步的研究方向。

As various new wireless communication technologies widely used, wireless spectrum resource has become one of the most precious resources in current society. Recently, the average utilization rate of licensed spectrum is very low and it has an unbalanced distribution. Cognitive radio technology allows secondary users using licensed spectrum without interfering the primary users, it significantly improves the utilization rate of licensed spectrum. Spectrum decision making is used to allocate the best spectrum via QoS(Quality of Service). It is an important part of cognitive radio technology.

This paper studies on tree-based clustering cognitive radio subnet. This subnet’s routing strategy and method of dynamic spectrum management are proposed. This subnet employs multi-cluster parallel operation mode. The master node of the subnet allocates the spectrum centrally. The spectrum decision making is proposed based on such subnet, and it involves subnet capacity, subnet throughput and subnet stability.

Subnet Topology Search Algorithm（STS） is proposed for two QoS: subnet capacity maximization and subnet throughput maximization. STS created a search space without generating duplicated nodes via cluster structure and cluster growth degree. The search space was pruned at dynamic subnet capacity threshold. Two improved versions of STS was also proposed. Time Limited Subnet Topology Search Algorithm（TLSTS） improves the time performance of STS via limiting the solution iterations. Fast Subnet Topology Search Algorithm（FSTS） uses the method of adjusting the searching path dynamically to improve the time performance of STS by half.

The experiments of STS, TLSTS and FSTS were done at the same simulation parameters. The results show that FSTS could obtain the optimal solution and meet real-time requirement under a specified subnet scale and spectrum space constraints. Finally, summary of the main points and direction for further research are given.