The Research On Satellite Dynamic Resource Allocation Techniques In Aerospace Networks

With the explosive growth in ubiquitous communication demands for mobile communication and Internet accessing,and the continuous development of global covering and high throughput communication services,the emergence and development of aerospace networks are promoted.As an important infrastructure and a predominant component of aerospace networks,the deployment of the satellite system is becoming a new hotspot due to its fast development of manufacturing and launch technology in recent years.Although the development and applications of multibeam antennas,frequency reuse,onboard processing(OBP),etc.have relaxed the intensity status of satellite resources to some extent,the satellite system resources are still extremelly insufficient because of high throughput and large capacity multimedia services requirements and developments in aerospace networks.It is an important problem to improve the utilization of satellite systems resources such as spectrum,storage,power,etc..The research on dynamic resource allocation(DRA)in satellite systems has great significance.In summary,this dissertation focuses on satellite dynamic resource allocation techniques in aerospace network.The main innovative results are as follows:(1)In view of the low flexibility proplem of current DPA methods when adapting.to the dynamic variation of the satellite-ground channel conditions,the traffic demands,etc.,an online Deep Reinforcement Learning-Dynamic Power Allocation(DRL-DPA)algorithm is proposed in this dissertation.The proposed algorithm solves the online-sequential decision problem in dynamic and changeable environments through the deep reinforcement learning method.And,based on the key idea of the match of power and buffered data,the online DPA decisions are mainly made depending on the caching data in each beam instead of traffic demands.The proposed DRL-DPA algorithm can increase the system throughput in 5.3%comparing with state of art DPA algorithms.(2)In allusion of the problem that cache optimization is not involved in existing dynamic channel allocation(DCA)algorithms,a DCA method under constrained cache for multibeam satellite systems is proposed in this dissertation.To change the deficiency of heavy occupying or buffer overflow due to the throughput increasing in exsiting algoritms,the proposed algorithm adopts the idea of Multi-objective Optimization Problem(MOP)with both spectrum and cache considered,and realizes the approximation of the Pareto solution of user satisfaction degree and spectrum efficiency under restrict cache resources.The results show that the proposed algorithm can save the buffer capacity in 39.16%,increase the service demands satisfaction in 16.87%and improve the spectrum utilization in 20.51%compared with other traditional solutions.(3)In order to solve the problem that the existing DCA and DPA solutions can not guarantee the accumulative performance after sequential decisions,a Deep Reinforcement Learning-Multi-objective Optimization Problem(DRL-MOP)algorithm is proposed for dynamic satellite channel and power allocation in this dissertation.With considering the normalized weighted sum of spectrum efficiency,energy efficiency and satisfaction index as the optimization goal,the proposed DRL-MOP algorithm models the dynamically changing system environments and user arrival mode,and realizes the optimization of the accumulative performance in satellite systems based on the DRL and MOP idea.The computational complexity of the proposed algorithm is also analyzed.The results show that the proposed solution has lower complexity and it can improve the spectrum efficiency in 50.51%,promote the energy efficiency in 21.82%and increase the service demands satisfaction in 12.78%compared with classical metaheuristic solution.