Dynamic Load-Balancing Spectrum Decision for Heterogeneous Services Provisioning in Multi-Channel Cognitive Radio Networks
Dynamic Load-Balancing Spectrum Decision for Heterogeneous Services Provisioning in Multi-Channel Cognitive Radio Networks
复制标题
DOI：
10.1109/twc.2017.2717403
复制
发表时间：
2017-06
影响因子：
10.4
通讯作者：
Huijin Cao;Hongqiao Tian;Jun Cai;A. Alfa;Shiwei Huang
中科院分区：
文献类型：
--
作者：
Huijin Cao;Hongqiao Tian;Jun Cai;A. Alfa;Shiwei Huang
本文研究了一种认知无线电网络（CRN）的动态负载均衡频谱决策问题，该网络将从辅助用户（SU）的数据包动态分配到不同可用的主信道。我们在SU中考虑了两种不同的服务类别，即延迟敏感（DS）和最佳努力（BE）服务，并为延迟敏感（DS）服务分配了更高的优先级。我们应用优先级排队模型来解决CRN中的优先级问题。在排队模型的基础上，建立了两个马尔可夫决策过程（mdp），其目标是在保证DS服务优先级的同时使两个服务的平均延迟最小。在交通和信道特征未知的情况下，应用强化学习方法寻找最优解。为了解决MDP解决方案的计算复杂性问题，我们提出了一种基于估计数据包停留时间的近视眼方法，该方法通过制定相位类型分布来推导。仿真结果验证了所提算法对负载均衡频谱决策的有效性。研究结果还表明，该方案在降低低优先级BE服务延迟性能的前提下，可以显著降低计算复杂度。
In this paper, we study dynamic load-balancing spectrum decision for a cognitive radio network (CRN) that dynamically distributes packets from the secondary user (SU) to different available primary channels. We consider two different classes of services at the SU, i.e., delay sensitive (DS) and best effort (BE) services, and assign a higher priority to the DS services. We apply priority queuing model to address this priority issue in the CRN. Based on the queuing model, two Markov decision processes (MDPs) are formulated with objectives to minimize the average delay of both services while guaranteeing the priority of the DS services. Reinforcement learning is applied to find the optimal solutions when the traffic and channel characteristics are unknown. To address the computational complexity issue in the MDP solutions, we propose a myopic method based on the estimated packet sojourn time, which is derived by formulating a phase type distribution. Simulation results demonstrate the effectiveness of all proposed algorithms for load-balancing spectrum decision. It also shows that the proposed myopic scheme can achieve significant reduction on computational complexity with a cost on the delay performance of low priority BE services.