【作者】 张巍；

【导师】 王晟；

【作者基本信息】 电子科技大学 ， 通信与信息工程， 2015， 硕士

【摘要】 带宽测量在网络信息领域中是一项非常重要的技术。它通过测量出网络路径在单位时间内能传输的最大数据量,来指导网络系统的Qo S管理,拥塞控制,以及路由选择访问等。2006年,Ekelin S,Nilsson M等研究学者提出了在带宽测量领域中使用卡尔曼滤波器的测量方法BART,在保持高测量精度的同时,提高了对变化环境的跟踪性能,不仅扩大了算法的使用条件,也大幅度降低了计算功耗。2008年,采用了BART设计思路的基于概率模型的ABEST算法也被提出,拓展了卡尔曼滤波器在带宽测量领域的使用。不过,在BART和ABEST中均只提出了卡尔曼滤波器的使用方法,当链路环境改变时,若使用相同的参数设置则可能导致测量的结果会千差万别。本文在研究卡尔曼滤波器的过程中,先以BART算法为基础,通过模型,理论,实验分析,引入了卡尔曼滤波器的自适应参数设定,提高了算法对不同链路环境的适应性,并提高测量的精度和跟踪性能。接着以研究ABEST算法为基础,根据标准状态方程的完整性,提出了动态状态方程的改进思路,设计了双探测流发包策略,提高了算法的跟踪性能。为了补充卡尔曼滤波器在带宽测量领域中的应用,以及考虑结合两种改进算法,本文在基于探测间隔模型IGI中,建立卡尔曼滤波器以脱离对链路容量C的依赖,并尝试将自适应参数设定和动态状态方程结合起来,进一步提高卡尔曼滤波器系统的测量精度,跟踪性能,稳定性,以及实用性。更多还原

【Abstract】 Bandwidth measurement is very important in the field of network information. It plays a major role in the Qo S of network system management, congestion control and the routing access by measuring the largest amount of data of a link or a network path in the unit to transfer in the course of time. In 2006, Ekelin S and Nilsson M put forward the BART, which using Kalman filter method in the field of bandwidth measurement.While maintaining the high measuring accuracy, the BART improves the tracking performance of the changed environment. And it not only greatly reduces the computational power, but also expands the using conditions of the algorithm. In2008,the Probability-based Model ABEST, which uses the design of BART, extend the using of Kalman filter. However, the BART and ABEST just presents the method of Kalman filter in the context. It can’t be used in the other different link condition because of its parameter Settings ways.Now this paper presents a new adaptive filtering system according to the basis of the model, the theory, the experimental analysis of BART. It improves the measurement accuracy and tracking performance proved by the experiments. Then, this paper presents a new measurement “dynamic state equation” based on the probability-based model ABEST according to the standard state equation. By putting forward the double probe flow measurement strategy, It effectively improves the filtering performance. In the study of measuring the method based on the probability-based model IGI, the paper tries to combine the adaptive filtering and the dynamic state equation to make the Kalman Filter more stability and more practicability, so that it do not need the capacity C of bottleneck link while measuring more accuracy and faster.更多还原