Research On Dynamic Beam Management Technology Of Satellite Network

The multi-beam LEO satellite constellation has attracted extensive attention due to its wide coverage,high system capacity,and high resource utilization.However,with the increase in the number of satellites in the LEO satellite constellation system,the unreasonable use of various communication resources such as frequency and bandwidth will inevitably cause serious interference.It seriously restricts the system access capability of the LEO satellite constellation system in the service downlink.Therefore,studying the beam management method of the LEO satellite constellation with system interference is the key technology to improving the access capacity of the constellation service downlink.However,due to the high-speed movement of LEO satellites,the satellite network topology and communication channels will change in real-time.It makes the spatial-temporal characteristics of interference in the LEO satellite constellation complex,which leads to great challenges in interference analysis.Meanwhile,in the case of considering the interference in the constellation system,resource allocation results will affect the situation of network interference,so that there is a coupling relationship between the system states of adjacent time slots.In response to the above challenges,this thesis focuses on the multi-beam LEO satellite constellation scenario,which considers system interference.Starting from the interference analysis of LEO satellite constellations,the relationship between system access capacity and satellite orbit height,inclination,constellation size,and configuration is analyzed in detail.Furthermore,the access capacity of the constellation system is further improved by designing a resource allocation method.To this end,the thesis carry out the following work:(1)For the scenario of a multi-beam LEO satellite constellation,this thesis proposes an interference analysis method based on system access capacity and further analyzes the spatio-temporal evolution characteristics of interference.This thesis first describes the basic system model of the LEO satellite constellation,analyzes the influencing factors of signal transmission during satellite communication,and specifically describes the LEO satellite antenna model,free space attenuation,and rainfall attenuation.Secondly,this thesis studies the possible interference situations in the scenario and analyzes the influence of interference on the access capacity of the LEO satellite constellation system from the perspectives of orbit height,constellation size,and orbital plane inclination.It is found that the system access capacity will increase first and then decrease with the increase of the constellation size,and the constellation configuration with relatively uniform distribution is more conducive to improving the system access capacity.The thesis analyzes the spatio-temporal evolution characteristics of interference in the LEO satellite constellation system and reveal that the interference changes periodically with time.The intensity of interference is independent of the longitude of the region and increases with latitude.It then guides the beam management strategy to further improve the system access capacity.(2)Based on the multi-beam LEO satellite constellation scenario in the interference environment,this thesis further improves the system access capacity by beam management and proposes a dynamic power allocation algorithm based on deep reinforcement learning.Firstly,the dynamic power allocation problem is constructed as a dynamic decision problem,then proposes a dynamic power allocation algorithm based on Q-Learning.To solve the complex and huge environment state space in the interference environment,based on the above algorithm,the thesis proposes a dynamic power allocation algorithm based on deep reinforcement learning.Finally,the algorithm is verified by co-simulation on Satellite Tool Kit(STK)and Python platform,and the results show that both algorithms can improve the access capacity of the network.In addition,the dynamic power allocation algorithm based on deep reinforcement learning is better than the one based on Q-Learning.