一种基于Dueling DQN改进的低轨卫星路由算法

A Low-Orbit Satellite Routing Algorithm Based on Improved Dueling DQN

卫星网络具有高动态性、节点处理能力不足,流量负载不均等问题。现有的地面路由算法并不能很好的解决卫星网络存在的问题。针对此问题,提出一种改进Dueling DQN的低轨卫星路由算法。首先,在路由算法中引入决斗网络的思想;然后在经验回放进行改进,将随机经验采样和优先经验采样进行融合,设置分层采样方法来进行采样;最后对网络进行参数的设置并且进行训练。从仿真和分析表明,从网络传输时延、系统吞吐量、丢包率方面有明显的提升。

Satellite networks face challenges such as high dynamics,limited node processing capacity,and uneven traffic load.Existing ground routing algorithms cannot effectively solve these problems.To address this issue,an improved low-orbit satellite routing algorithm based on Dueling DQN is proposed.First,the concept of Dueling Networks is introduced into the routing algorithm.Then,the experience replay is improved by fusing random experience sampling and prioritized experience sampling,and a layered sampling method is set up for sampling.Finally,the network parameters are set and the model is trained.Simulation and analysis show significant improvements in network transmission delay,system throughput,and packet loss rate.