An Improved Low-orbit Satellite Routing Algorithm Based on DQN
To deal with the problem of fast-changing link-state between satellite nodes due to high-speed movement of satellite nodes,a satellite routing algorithm based on deep reinforcement learning is studied,and an improved satellite routing algorithm based on DQN is proposed.The algorithm adopts the idea of virtual nodes,and sets the number of hops and distance as the related parameters of reward function based on the principle of minimum hops.Meanwhile,a priority experience replay mechanism is set up to train the algorithm by learning samples with the highest value.Finally,the network is set in parameters and is trained.The simulation results show that there are significant improvements in network transmission delay,system throughput,and rate of packet loss,effectively adapting to high dynamic changes in link-state between satellite nodes.
国家科技期刊平台
NSTL
万方数据
维普
