【摘要】 针对源—目的(OD)流量估计解的不稳定性和求解方法的复杂性,将广义线性反演应用于大尺度网络流量矩阵估计,降低求解方法的复杂性;采用奇异值分解(SVD)获得阻尼系数,提高了求解的稳定性;采用历史均值和链路测量数据相结合决定反演初值,降低多解性。Abilene网络的实际数据仿真结果表明,提出的算法能够在保证较高实时性的前提下,提高计算精度。更多还原

【Abstract】 To reduce the estimation’s complexity,an algorithm for traffic matrix estimation based on generalized linear inversion theory was proposed.For the purpose of improving the algorithm stability,a Singular Value Decomposition(SVD)decomposition method was used to find an optimum damping coefficient.Moreover,the historical average and link data were used to get the prior information so that the number of solutions can be reduced.Simulation results using Abilene network’s actual data show that the proposed algorithm can guarantee good real-time capability while the computational accuracy can also be improved significantly.更多还原