A Load Balancing Algorithm In Satellite-terrestrial Integrated Network

With the development of Fifth Generation Mobile Communication Technology,there are more data transmission requirements in remote regions.In these regions base stations are likely to overload,and it is difficult to find an appropriate offloading target based on the existing load balancing algorithms.Fortunately,in the pre-research of the future network,the satellite-terrestrial integration network architecture has attracted much attention.The seamless coverage provided by satellite base stations can obviously help solve the overload problem of terrestrial base stations.So this thesis studies the load balancing algorithm in the satellite-terrestrial integration network.The existing load balancing algorithms take the signal reception as the single metric,ignoring the dynamic changes of the users,which influences the long-term performance of these algorithms.Therefore,this thesis considers designing a metric that considers multiple factors.Based on the modeling and numerical simulation,this thesis first designs the Fuzzy Evaluation Metric(FEM).FEM introduces the fuzzy logic mechanism to avoid the difficulty of designing explicit mathematical relations.FEM is used to evaluate the impact of users on the overload of base stations both at present and in the future,which is the basis of the subsequent algorithm design in this thesis.Then,this thesis proposes the Fuzzy Logic and Reinforcement Learning based Access(FLRL-AC)algorithm where the Fuzzy Deep Deterministic Policy Gradient(FDDPG)model is proposed by introducing FEM.FLRL-AC preprocesses the input data with FEM,and makes access decisions in consider of the future impacts.Compared with several widely used reinforcement learning models,the user access decisions given by the proposed algorithm can better avoid overload.Then,this thesis proposes the Fuzzy Logic based Offloading(FL-OL)algorithm,which uses FEM to find the most suitable users and the target base stations to be offloaded.Compared with the existing algorithms,FLOL extends the time interval of the next overload.And while keeping the computational complexity,FL-OL reduces the overhead of calling the algorithm multiple times and triggering user switching.Finally,this thesis combines the above two algorithms to obtain the Fuzzy Logic based Load Balancing(FL-LB)algorithm.FL-LB introduces active load balancing before overload to reduce the probability of overload.Also,FL-LB retains passive load balancing after overload to provide further protection.It solves the overload problem in the satellite-terrestrial integration network from two aspects and ensures the service quality of the satellite-terrestrial integration network.