【作者】 陆文彦；

【导师】 贾维嘉；

【作者基本信息】 中南大学 ， 计算机应用技术， 2008， 博士

【摘要】 作为一种新兴的无线宽带网络,无线城域网具有高速度,远距离,高效率的优点,近年来成为了无线通信研究的热点。为了提供QoS支持,无线城域网采用了请求/授予模式来发送数据,即用户站想要发送数据时,首先必须向基站发送带宽请求。由于带宽请求是数据收发的前提,因而有效的带宽请求接入控制对无线城域网的性能具有重要的影响,它是无线城域网的一个基础问题。目前IEEE 802.16协议定义了轮询、捎带、借用、轮询位带宽请求发送方法,针对广播轮询时,协议推荐采用基于二进制指数回退算法的竞争解决方案,这些方法和方案为带宽请求的接入定义了框架和基础。但是各种方法的性能特征,适应范围,以及各种方法之间的选择都不在标准的定义范围。为了最大限度地使用无线资源并保证QoS的支持,论文研究了无线城域网中基于二进制指数回退算法的竞争解决方案的性能,给出了一种中心调度的竞争解决方案,针对组播轮询提出了一种二元组播轮询,讨论了单播轮询和广播轮询的选择算法,最后结合竞争和时分复用多路接入协议的特点,提出了一种n模的多路接入协议。这些研究对于无线城域网的组网和运营具有重要的指导意义。论文的具体工作为:（1）基于IEEE 802.16 MAC层的特征以及它与802.11的差别分析,论文给出了传送机会利用率u,带宽请求延时d等性能评价指标,并通过概率分析的方法给出这些性能指标的计算方法,通过理论和模拟结果分析,讨论了初始化竞争窗口w,最大竞争窗口W_max、每帧中提供的传送机会数目N_to,用户站总数目N,活动用户站数目n等参数对系统性能的影响,基于这些分析,给出了无线城域网中设置系统参数的原则和方法。（2）通过对IEEE 802.16竞争性能的分析,论文发现基于二进制指数回退方法在某些情况下性能并不是最优的,在分析导致性能不能达到最优的原因的基础上,根据IEEE 802.16的特点,论文给出了一种中心调度的竞争解决方案,它的基本思想是,基站在每个时间帧把预测的当前活动用户站数目广播给所有的用户站,用户站每次以该数目参与竞争。理论分析和模拟结果表明,虽然基站对用户站点预测存在误差,但是中心调度的方法依然能达到比较好的性能,在很多指标上,它优于基于二进制指数回退算法的方案。（3）二进制指数回退方法和中心调度方法理论上性能都只能达到（1-1/n）^n-1,这一结果在n比较大时并不理想,因此本文分析了无线城域网中单播轮询和组播轮询的性能特点,给出了一种二元组播轮询,通过将用户站两两分组,来提高竞争的成功率,理论分析表明,二元组播轮询在用户站点数目低于0.75倍总站点数目时,比单播轮询性能要好。结合广播轮询的特点,本文还给出了基站在选择广播轮询和单播轮询时的原则和方法。（4）带宽请求的发送本质上是一个多路接入点问题,多路接入有竞争和非竞争两种方式,竞争方式适合于用户站站点数目不多的情况,而非竞争方式则正好相反,为了有效地结合竞争方式和非竞争方式的优点,论文给出一种n模的多路协议,它的思想是:协调者（基站）将当前竞争的状况以n数广播给所有的站点,各个站点通过将自己的编号与n求模得到一余数r,然后站点在第r个时隙发送自己的数据。理论分析表明,这种方法无论是轻载荷还是重载荷时都有比较理想的性能。更多还原

【Abstract】 As an up and coming network, Wireless Metropolitan Area Networks （WMAN） which is possessed with outstanding advantages such as long distance transmission, high speed and efficiency, has become the hot topic in wireless communications in recent years. To support QoS, the request/grant mechanism has been used in WMAN. Under this mechanism, when a subscriber station wants to send data, it needs to send a bandwidth （BW） request to base station firstly. So the admission control of BW request has a great impact on the performance of WMAN and is treated as a fundamental problem in WMAN.At present, there are several BW request methods in IEEE 802.16 which include unicast polling, multicast polling, broadcast polling, piggyback and so on. As for broadcast polling, the mandatory contention resolution that shall be supported by IEEE 802.16 is based on the truncated binary exponential backoff algorithm. All these methods and resolutions have defined the framework for subscriber station to send BW request. While the performance of each method and how to choose the right method according to different parameters of system are out of the range of standard. To fully use the wireless resource and guarantee the QoS, this thesis studies the performance of the contention resolution provided by IEEE 802.16 and presents a centralized contention resolution and a binary multicast polling method. The choice between unicast polling and multicast polling is also discussed and an n modulus multiple access protocol is presented. All these studies are useful for building and running WMAN. The works of this thesis in detail are as followings:（1） Based on the analysis of the differences of the MAC layer between IEEE 802.16 and IEEE 802.11, the metrics to evaluate the performance of BEB based contention resolution, which include the utilization of transmission opportunity u, the delay of BW request d and drop probability of BW request p_d are calculated by using the probability method. With theoretical analysis and extensive simulations, the effects of contention parameters such as initial window w, the Maximum contention window W_max, the number of transmission opportunities N_to, the number of total SSs N and the number of active SSs n are discussed. The principle and method of setting system parameters are also presented.（2） With the performance analysis of contention resolution of IEEE 802.16, we found that the BEB based resolution can not get optimal performance in some situations. So a novel centralized scheduling contention resolution （CSCR） is presented. The basic idea of CSCR is: in each time frame, by estimating the number of active SSs, the BS broadcast an optimal window to all SSs, each SS then individually participates with this window in transmission. Theoretical analysis and extensive simulation results show that even there are errors between the really and estimated number of active SSs, CSCR outperforms BEB based resolution on many aspects.（3） Since the maximum utilities of TO under both BEB and CSCR based contention resolution can only be （1-1/n）^n-1, which is not so good when n is large, we turn to the unicast and multicast polling method and present a binary multicast polling method which divides the whole SSs into N/2 groups, each group consists of two SSs, each SS only contest with the SS in its group. The theoretical analysis indicates that when the number of active SSs is less than 0.757V, the binary multicast polling is better than unicast polling. We also provide an algorithm to make a choice between unicast and broadcast polling.（4） The BW request is a multiple access problem. Typically there are contention based and contention free ways for this problem. For the contention based method, it has good performance when there are only little active SSs, while the contention free method has the contrast situations. To full use the benefit of both types of protocol, an n modulus multiple access protocol is presented. In this protocol, the coordinator broadcasts the number of active nodes to all nodes. Each node divides this number with its own ID number and gets a reminder r and then sends its data at rth slot. The theoretical analysis indicates that the n modulus protocol has a good performance in both heavy and light load.更多还原