| With the development of wireless communication technologies,the capability of the equipments like user terminal and access point has been improved significantly,which also has promoted the realization of the future network with the goals of global coverage and seamless handover.The increasing demand of traffic in people’s daily life has been satisfied through deploying cellular base stations with high density.In order to cope with the burst traffic scenarios in the public places where human activities are concentrated,the Unmanned Aerial Vehicle(UAV)base stations have been intro-duced into the low-altitude heterogeneous networks as a supplementary communication technology for terrestrial cellular networks.Besides,to provide effective network cov-erage for those sparsely populated areas such as oceans and deserts,various types of satellites serve as space base stations in the space network.The large number and wide variety of serving base stations have caused that the complexity of network management increases exponentially,especially when mobile user equipments handover between dif-ferent base stations.It is an urgent problem to guarantee the quality of service on user equipments when the handover occurs.Facing the challenges in network management which caused by the ultra-density nodes,this dissertation focus on optimizing the performance of mobility management procedures,and studies the handover issues of mobile user equipments in different net-works,which include the terrestrial cellular networks,the low-altitude heterogeneous networks and the Space-Terrestrial Integrated Networks(STIN).In this dissertation,the characteristics of the wireless channels in different network sceneries have been stud-ied,which are used for designing the related mobility management solutions.Due to the stochastic changes of the wireless network environment is the main infection on the channel characteristics,we have adopted the reinforcement learning theory to make prediction on handover decision based on the statistic of history information,further to optimize the performance of the mobility management solutions.The main contribu-tions of this dissertation are summarized as follows:1.This dissertation has proposed a mobility management scheme under the Ultra-Dense Networks(UDNs),based on a reinforcement learning framework and multi-armed bandits theory.The density deployment of base stations in terrestrial cellular networks has in-troduced the frequent handovers and ping-pong handover problems.In order to solve that,we have proposed the cascading bandits based handover management algorithm to choose the base station with the optimal estimated long-term performance as the han-dover target,rather than candidate base station with the highest instantaneous channel qualities.Through the proposed algorithm,we can improve the performance of mobility management in UDN in terms of handover delay and dropped-call rate.In addition,we have proved the equivalence between 3GPP handover protocols and the ∈-greedy bandit algorithms.Then,we have performed the system simulation experiments to build the practical communication scenarios,and have applied the proposed handover algorithm together with the competitive solutions.Compared with the traditional 3GPP handover protocols,the simulation results show that the proposed algorithm improves the han-dover performance in terms of handover delay and call drop rate effectively.Finally,the robustness of the proposed algorithm has been verified in different scenarios with delayed and missing feedback.2.This dissertation has addressed the problem of optimizing the neighbor cell lists in low-altitude heterogeneous networks.In order to improve the accurate of evaluating candidate base stations,we have analyzed the channel characteristics between different types of base stations and user equipments.Specifically,the transmission power and load capacity of the candidate base stations has been estimated through historical information,and the results have been adopted in predicting the handover target for optimizing handover performance.On this basis,we have combined the cascading model with the bandits theory,and proposed the cost-aware cascading bandit NCL optimization(CCB-NCL)algorithm in this dissertation.Then,we have built the practical communication scenarios through conducting the system simulation experiments,and applied different NCL optimization algorithms.The simulation results show that the proposed CCB-NCL algorithm out-perform the exist solutions in terms of optimizing the length of presented NCL and the order of base stations in NCL,which reduces the signaling overhead during transmis-sion and the number of scanned candidate base stations during the preparation phase of handover procedure.3.This dissertation has accomplished the optimization of node location man-agement in STIN.By separating the Global Unique Identifier(GUID)from the Network Address(Network Address,NA),the location of mobile nodes could be maintained with less complexity in STIN.Based on the GUIS/NA mapping pair,we have designed the do-main and cluster area division mechanism,and established the distributed UID/NA map-ping resolving service system.With the novel STIN architecture,this dissertation has conducted two tasks in location management.Firstly,we have designed the sparse con-fidence upper bound algorithm,to allocate the mapping pair information of the network nodes into the distributed mapping resolving service.The simulation results of corre-sponding experiment show that the proposed algorithm can improve the content hit rate of inter-domain requests significantly.Secondly,we proposed the location updating strategy to deal with the challenge introduced by the continuous changes of the network topology.We have also carried out several simulation works to compare the perfor-mance between the proposed algorithm and exist updating solutions.The simulation experiment results show that the proposed algorithm can reduce the update cost effec-tively,through avoiding the satellite-ground links and unnecessary nodes in the update path. |
