Research On Intelligent Access Control And Resource Allocation Mechanism For Space-Air-Ground Integrated Network

Recently,space-air-ground integrated network(SAGIN)has emerged as an integrated information network with space networks as the main body,ground networks as the foundation,combined with the air networks.It provides land,sea,air,and space users with anytime access,global coverage,on-demand services,safe and reliable information services.With the aim of mastering global space resources in the era of information,it is of strategic importance in economy,society and military to develop the service capability of SAGIN to meet diversified business requirements.Meanwhile,it is also inevitable to establish global leadership in new technologies and industries and to realize global information service.With the rapid development of satellite communication technology and mobile communication technology,the demand for users and their multimedia services has exploded.Meanwhile,limited wireless resources,such as on-board spectrum and power,have always been scarce,which can lead to system performance degradation.More importantly,the nodes of SAGIN are wide-area distributed and highly dynamic,and the network itself is of resource heterogeneity with large propagation delay and severe fading.Such a complex and dynamic environment has brought great challenges to wireless resource management.Therefore,an efficient and flexible wireless resource allocation scheme has become a key issue to satisfy various requirements of users and improve system resource utility in the construction of SAGIN.Recently,AI(artificial intelligence)algorithms are being applied to the communication field.Especially,deep reinforcement learning methods show great potential in solving decision problems in complex environments,which shed light on solving the above problems with intelligent techniques.To this end,this paper is focused on intelligent access control and resource allocation,and our main contributions are summarized as follows:Firstly,we work on the inter-beam admission control mechanism.Due to the rapid relative movement between LEO satellites and user terminals,handovers are frequently required to switch service calls between beams.Based on traditional channel reservation strategy,we set different priorities and admission thresholds for a new call or handover call to multiple services.We propose a dynamic channel reservation strategy based on the Actor-Critic framework(AC-DCRS)for dynamic adjustment of admission thresholds,which reaches a balance among all services in terms of the system quality of service(Qo S),and improves the performance of both users' side and the network.The numerical results show that the AC-DCRS achieves better long-term overall system performance,average access success,and channel utilization under different service traffic and dynamic scenarios.Secondly,we studied the access selection mechanism under a multi-layer coverage.As ground terminals are often located in the coverage area of multiple communication nodes in the space and air networks,we need to decide where to access the network.We propose an intelligent access selection algorithm based on a multi-agent deep reinforcement learning algorithm named MADDPG,which overcomes environmental non-stationary in multi-agent scenarios.With the defined Qo S(defined as a comprehensive weighting of multiple performance metrics)as a long-term optimization objective,we achieve intelligent access selection for multiple user terminals.Besides,a reparameterization approach is used to solve the problem of disability to calculate derivatives by probabilistic sampling in discrete action space and to expand the exploration of action selection.The numerical results show that the proposed algorithm achieves a better average system Qo S,access success rate,system throughput,and improved overall system performance compared to baseline algorithms.Thirdly,the power allocation mechanism of GEO multi-beam satellites is studied.Due to the limited on board power resources and the severe influence of co-channel interference and rain fade in the Ka-band,efficient and rational power allocation across beams in a dynamic environment is required.We propose a dynamic power allocation algorithm based on Soft Actor-Critic(SAC)with the long-term optimization objective of maximizing system throughput,which achieves fine-grained and efficient real-time decision making in high-dimensional action space through continuous control in continuous state space.As the SAC algorithm has enhanced exploration capabilities to search for optimal solutions in the largest range,and thus more robust optimal strategies can be learned.The numerical results show that the algorithm achieves higher system throughput.