基于FMB和元胞自动机的实际业务流性能预测方法

Prediction Method of Actual Traffic Performance Based on FBM and Cellular Automaton

为了有效刻画实际业务流性能状态,结合分形布朗运动模型(Fractional Brownian Motion,FBM)和元胞自动机提出一种新的预测方法 TSPCA(Traffic State Prediction method based on Cellular Automaton).该方法首先基于FBM模型推导了平均队列长度和平均时延的数学表达式,同时利用定义的元胞演化规则对估算结果进行修正,以提高预测精度.最后,通过NS2和MATLAB进行仿真实验,深入分析了影响该方法的关键因素,发现缓冲区较小时流量性能将由短相关特性支配,而缓冲区较大时性能由长相关支配,重置效应和截断效应对业务流性能影响较大.并且对比FARIMA和ARIMA的预测结果,证明该方法具有较好的适应性.

In order to effectively describe the performance state of actual traffic, a novel prediction method TSPCA （ Traffic State Pre- diction method based on Cellular Automaton ） is proposed with Fractional Brownian Motion （ FBM ） model and cellular automaton. The method has deduced the mathematical expression of average queue length and average delay based on the FBM model, and has revised the estimated results with the cellular evolution rules that have been defined so as to improve the prediction accuracy. At last, a simulation experiment is conducted through NS2 and MATLAB, which analyses key factors that affect the method. And it is found that, the traffic performance is mainly related with short-dependence when buffer size is lower, and the traffic performance is mainly related with long-dependence when buffer size is bigger, which is affected by resetting effects and truncating effects. Compared to FARIMA and ARIMA, the results shows that TSPCA has better suitability.