非静止轨道卫星网络中Hopfield神经网络路由算法

应用Hop fie ld神经网络路由算法能够解决全球无缝覆盖的W a lker de lta卫星通信网络中的路由问题。模拟退火算法作为一种全局优化的方法能够解决神经网络迭代运算中的振荡环问题,消除Hop fie ld神经网络计算局部稳定状态。Hop fie ld神经网络路由算法需要大量的神经元,消耗大量的运算时间。本文提出一种适合卫星网络运算的神经元数量减少的神经网络路由算法,根据卫星网络半径,削减神经元数量、加快运算时间。根据中轨和低轨卫星网络卫星数量的差异,在仿真过程中选择Hop fie ld神经网络的最优参数,比较不同源-目标节点对、不同时刻和不同收敛判决门限时的算法性能。

Hopfield Neural Network Routing Algorithm in Nongeostationary Satellite Communication Networks

Hopfield neural network routing algorithm can be used in seamless Walker delta satellite communication networks to solve the routing problem.The simulation annealing algorithm can solve the pro-blems of the oscillation ring,thus eliminating the local stable state.However,the Hopfield neural network is a large scale network,which needs large quantity of neurons and consumes much time for calculating.This paper presents a neuron-reduced Hopfield neural network routing algorithm,and the algorithm needs fewer neurons and fastens the speed of the routing calculation.The optimal parameters for Hopfield-network in two satellite networks with different altitudes are selected,and the performances of the new Hopfield routing algorithm are compared with different origination-destination-pairs at different time and with different thresholds.