基于PA的卫星网络动态带宽分配在线算法及仿真

为解决卫星网络带宽分配问题,基于离散事件动态系统理论中的PA方法,提出了一种在线算法.该算法把动态带宽分配问题描述为约束优化问题,以一次计算机仿真的数据为基础,通过PA方法求出网络平均延时相对于分配带宽的梯度,以优化网络平均延时为目标,构造带宽约束条件下的迭代公式,在线地分配带宽,有效解决了一般计算机仿真中需要进行多次重复仿真而导致的大量机时问题,克服了一般算法须假定信息流的统计特性的限制.仿真结果表明,该算法能明显改善网络的平均延时性能.

Online Dynamic Bandwidth Allocation Based on Perturbation Analysis for Satellite Networks

To solve the dynamic bandwidth allocation problem in satellite networks, based on the Perturbation Analysis method, a new online algorithm was proposed. The algorithm regarded the dynamic bandwidth allocation problem as one minimization of the performance index (specially, the mean delay) with the decision variable (specially, the bandwidth) under some limitations. PA technique was used to derive sensitivity estimators for the performance index, thus provided online gradient descent algorithms capable to optimally minimize the mean delay. The algorithm decreased the consumption of computing power and avoided the assumption on the statistical characteristic of the information flow. Simulation results illustrate that the algorithm may improve the performance of the network mean delay.