【作者】 刘健；

【导师】 黄玉兰；

【作者基本信息】 吉林大学 ， 通信与信息系统， 2016， 硕士

【摘要】 IEEE 802.15.4描述了无线个人域局域网的物理层(PHY)和媒体接入控制子层(MAC)规范,具有成本低、功耗低、组网简单和数据率低等特点。基于这些性能特点,802.15.4已经被广泛应用于多种无线网络,如传感器网络、工业控制网络、智能家居、智能农业和环境监测等。然而,实践表明802.15.4的MAC子层仍存在诸多缺陷。无论是竞争接入阶段(CAP)的基于冲突避免的载波侦听多点接入机制(CSMA/CA),还是无竞争接入阶段(CFP)的基于先到先服务的保护时隙预留机制,均不提供差异化服务和优先级保证。在机器类通信(MTC)蔚然兴起的背景下,802.15.4网络中业务类型和Qo S需求表现迥异。实时业务具有固定的传输时间限制,超过时延期限将失去时效性,如视频流,工业控制信号和灾难预警信号;非实时业务具有节点数量庞大,数据包短小,突发度高和接入尝试频繁的特点。因而在802.15.4网络中,优化信道接入和时隙分配机制,提供异质业务的差异化服务,并保证Qo S具有重要意义。本文以保证实时业务在802.15.4网络中的传输时延为研究目标,提出CAP和CFP阶段的优化算法。首先,在CAP阶段,实时节点的GTS申请帧和非实时节点的数据帧同时竞争信道资源,由于标准协议提出的竞争接入机制并不提供优先级保证,因而如何提升实时节点的接入成功概率是本文拟解决的第一个问题。马尔科夫链理论(Markov Chain)是随机接入过程建模中最常用的理论。本文利用马尔科夫链理论建模异质节点的接入状态和信道状态转移过程,分别分析不同的接入参数,如竞争窗口、退避指数范围和退避次数,对网络性能的影响。基于该分析结果,本文以保证GTS请求帧的传输时延为优化目标,合理地有区别地改进实时节点和非实时节点的MAC子层参数,提供GTS请求帧和普通数据帧的差异化服务,以保证GTS请求帧具有更高的接入优先级。其次,在CFP阶段,多个实时流同时竞争数量有限的保护带宽资源,如何最大化带宽利用效率是本文拟解决的第二个问题。标准协议提出的先到先服务的时隙分配机制是静态的,缺乏公平性,带宽利用率低。马尔科夫决策过程(MDP)是时延敏感且资源有限系统中,研究最佳资源分配方式的最有效理论。本文以保证实时业务的Qo S,并且最大化CFP阶段的带宽利用效率为优化目标,将CFP阶段的时隙分配问题构建为MDP问题,提出最佳保护时隙分配机制(MOS)。MOS是一个跨层优化机制,同时考虑业务流的状态信息(包括时延限制、到达率和突发度)、信道状态信息和可用的带宽数量等MAC层和PHY层信息。基于MOS的最佳时隙分配策略是当前时刻收益和未来可能收益的权衡,它不仅考虑使当前时刻系统收益最大,还考虑在未来获得可能更大的收益。本文还讨论了MOS在802.15.4协议中的实施方法,保证MOS与标准协议的兼容。更多还原

【Abstract】 IEEE 802.15.4 described the specifications of physical layer(PHY) and media access control sublayer(MAC) for wireless personal area LAN. This protocol has distinctive characteristics of low cost, low power consumption, simple networking and low data rate. Because of these advantages, 802.15.4 has been widely used in various communication scenarios, such as wireless sensor networks, industrial control networks, smart home, intelligent agriculture and environmental monitoring. However, practice shows that its MAC sublayer protocol has many defects. Whether the collision avoidance based carrier sense multiple access mechanism(CSMA/CA) in CAP, or the inflexible first-come-first-serve(FCFS) based guaranteed slot reservation mechanism in CFP does not provide differentiated services and priority guarantee. In the context of springing up of machine type communication(MTC), the traffic types and Qo S requirements in 802.15.4 networks differ greatly. Real-time traffics have a fixed time constraints. As exceeding time limits, these traffics will lose timeliness, such as video streaming, industrial control signal and disaster warning signal. Oppositely, non-real-time traffics, which are generated by majority of nodes, have characteristics of short packet size, burst arrive and frequent access attempts.In this paper, we are inspired to present optimization theory for 802.15.4 to ensure the transmission of real-time traffics. First, in CAP, GTS request commands of real-time nodes compete with data frames of non-real nodes for channel resource. Since standard protocol can’t provide distinguishing priority, enhancing the access probability of real-time node is our first purpose. Markov chain theory is the most common theory for modeling random access procedure. In this paper, we utilize Markov chain theory to establish the probability transmission mode of channel states and node states, and reveal the influence of different access parameters, such as contention window, backoff exponents and number of backoff stages, on network performance. Then, to guarantee the transmission of real time traffic, we configure different access parameters for heterogeneous traffics. As a result, we can provide differentiated services between GTS request commands and data frames, and ensure the GTS request commands getting higher access priority.Secondly, in CFP, huge number of real time traffics request for the limited number of GTS resources. How to maximize bandwidth utilization is another purpose of this paper. The FCFS based GTS allocation strategy is a static method, and lacks of fairness, which may result in reducing bandwidth utilization. Markov Decision Process(MDP) is the most efficient theory for optimal resource allocation problem in delay sensitive and resource limited systems. For the purpose of maximizing the bandwidth utilization in CFP, we formulate the GTS allocation problem during CFP in IEEE 802.15.4 as a MDP problem, and propose a dynamic GTS allocation scheme, namely Markov decision process-based Optimal Scheduling(MOS). The MOS is a consideration of cross-layer information consisting of flow specification information(such as random burst arrivals, delay requirements, channel qualities), channel state information, and the available bandwidth resources. Furthermore, the optimal policy is a tradeoff between current immediate reward and possible future reward. Namely, we consider not only how to maximize the reward at present, but also the how to get greater reward in the future. Finally, we provide some implementation consideration of MOS to ensure backward compatibility with the standard.更多还原