MIMO感知多跳无线网络的跨层优化研究

发布时间：2018-06-09 04:52

本文选题：多跳无线网络 + MIMO技术　；参考：《浙江理工大学》2017年硕士论文

【摘要】：多输入多输出(Multiple Input Multiple Output,简称MIMO)技术因其能够显著改善传输容量限制和提高通信的可靠性,给无线通信领域带来了重大突破,与此同时,仍然有很多的应用场景等待深入探索。另一方面,多跳无线网络得益于其鲁棒性好、结构灵活、带宽高、可分布式部署等优势在无线网络通信领域得到了广泛应用。如果能够将两者有效结合,充分发挥其潜力,将大大改善现有通信质量。本文在充分了解目前国内外关于多跳无线网络资源优化分配相关研究的基础上,深入研究了MIMO技术在多跳无线网络中的调度问题,利用多天线带来的空间自由度优势来提供每条传输链路上的MIMO信道模式,并针对MIMO多跳无线网络这一场景,分别引入了预测队列和双层队列理念来设计改善网络模型并给出了分布式实施方案,本文的主要研究工作如下:1)提出基于信道模式的分布式跨层优化。针对MIMO多跳无线网络在长时间平均下网络效用最大化这一优化问题,提出了基于MIMO信道模式的动态感知调度模型,使得网络中的每个节点能够感知当前网络状态从而选择合适的MIMO信道模式来进行数据传输以满足通信需求。相比传统MIMO多跳无线网络,这种动态感知MIMO信道模式并进行调度决策的方式更加智能,也可以获得更好的网络效用,结合李雅普诺夫优化算法,保证了网络运行的稳定性,接着通过对偶分解算法将耦合项分离,最终实现整个跨层资源优化分配问题的分布式实施。2)提出基于预测队列的分布式跨层优化。在(1)的基础上,引入预测服务模型,即在原始队列模型中加入预测队列,网络中的每个节点基于一个预测窗口进行预测并发送未来一定范围内时隙的数据包。通过对未来数据的预测和提前决策规划,可以使整个网络在保证效用的情况下,有效减少数据包的时延。本文采用等效队列的方式,将真实数据队列和预测队列在宏观上先等效为一个求和队列结合李雅普诺夫漂移理论和对偶次梯度算法实现分布式求解,再根据具体的映射及更新规则分拆为每个具体优化项的决策。3)提出基于双层队列的分布式跨层优化。在(1)的基础上,引入双层队列模型来弥补经典背压式算法的不足。将整个网络的架构进行分离,在网络层和数据链路层分别构建队列,网络层部分负责每个节点的路由选择决策,数据链路层部分负责节点在链路上的调度决策,从而使得原来需要联合优化的方案可以分离。本文通过李雅普诺夫优化算法和对偶分解算法实现了网络效用最优并可分布式实施。
[Abstract]:Multiple Input Multiple Output (MIMO) technology has brought great breakthroughs in the field of wireless communication because it can significantly improve the transmission capacity limitation and improve the reliability of communication. At the same time, there are still many applications waiting for further exploration. On the other hand, multi hop wireless networks benefit from its good robustness, The advantages of flexible structure, high bandwidth and distributed deployment have been widely used in the field of wireless network communication. If it is possible to combine the two effectively and give full play to its potential, it will greatly improve the existing communication quality. This paper is based on a thorough understanding of the research on the optimal allocation of multi hop wireless network resources at home and abroad. The scheduling problem of MIMO technology in multi hop wireless networks is studied. The MIMO channel pattern on each transmission link is provided by using the spatial freedom advantage of multiple antennas. The predictive queue and double queue concept are introduced to improve the network model and give the distributed reality for the scene of MIMO multi hop wireless network. The main research work of this paper is as follows: 1) put forward the distributed cross layer optimization based on channel mode. Aiming at the optimization problem of MIMO multi hop wireless network with long time average network utility maximization, a dynamic perception scheduling model based on MIMO channel mode is proposed, so that each node in the network can perceive the current network. The state then selects the appropriate MIMO channel mode to carry on the data transmission to meet the communication needs. Compared with the traditional MIMO multi hop wireless network, this dynamic perception of MIMO channel mode and scheduling decision can be more intelligent, and can also obtain better network utility. Combined with the Li Ya prize optimization algorithm, the stability of the network operation is guaranteed. On the basis of (1), the prediction service model is introduced, that is, the prediction queue is added to the original queue model, and each node in the network is based on one, based on the (1). The prediction window predicts and sends data packets in a certain range of time slot in the future. Through the prediction of the future data and the early decision planning, the whole network can effectively reduce the time delay of the packet under the condition of guaranteeing the utility. In this paper, the equivalent queue is used to equip the real data queue and prediction queue at the macro level first. A summation queue combines Lyapunov drift theory and dual gradient algorithm to realize distributed solution. Then, based on the specific mapping and updating rules, the distributed cross layer optimization based on double queue is proposed based on the decision.3 of each specific optimization. On the basis of (1), a double queue model is introduced to make up for the classic back pressure calculation. The structure of the whole network is separated, the queues are constructed in the network layer and the data link layer respectively. The network layer is responsible for the routing decision of each node, and the data link layer is responsible for the scheduling decision of the nodes on the link. Thus, the original scheme which needs joint optimization can be separated. This paper through lyapuno is used in this paper. The optimal algorithm and dual decomposition algorithm achieve the best network utility and distributed implementation.
【学位授予单位】：浙江理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.3

【参考文献】