Research On 5G Network Resource Management With Non-Ideal Backhaul

With the rapid development of mobile communication in the past decades,user data has increased dramatically.High data rates and low latency have become th e urgent requirements for the fifth generation(5G)mobile communication technologies.In order to fulfill these requirements,5G heterogeneous dense networks emerged with capability of inter-site cooperation,which could achieve higher spectrum utilization and higher data rate than the traditional networks.Due to the dense distribution of small base stations,the deployment of backhaul links needs to be considered in combination of wired and wireless.However,because of its open transmission environment,the performance of wireless backhaul are restricted,thus affecting the overall performance of the 5G communica.tion networks.Nowadays,lots of resear-ches are devoted to the resource management in 5G communication networks with ideal backhaul.However,in the practical scenarios,when considering the constraints of wireless backhaul,resource management in the 5G networks needs to be further studied.Meanwhile,the service requirements for users in 5G networks are diverse,which increases the optimization difficulty of the traditional resource management solutions.Therefore,this thesis focuses on the researches of 5G network resource management with non-ideal backhaul,which mainly includes the following three aspects:A critical issue caused by backhaul delay in 5G networks is that the information transferred between cooperative base stations is non-realtime and thus causes system throughput degradation.To solve this problem,a robust channel prediction algorithm combined by Echo State Network(ESN)and Genetic Algorithm(GA)is proposed,and schedule decisions are made based on the predicted Channel State Information(CSI).The proposed algorithm uses an off-line trained ESN to predict current CSI based on the historical CSI,then GA is utilized to finetune the predicted value with the goal of maximizing system throughput.Compared with the traditional linear prediction algorithm,the proposed algorithm can achieve higher prediction accuracy,because it combines the characteristics of certain memory of ESN and the robustness of GA mutation process.The simulation results show that the proposed algorithm can achieve better system throughput performance than the traditional scheduling algorithm.With non-ideal backhaul links,the amount of data transferred between base stations in the 5G heterogeneous dense networks is limited.Meanwhile,system delay may be too large to meet users' requirements.To solve this problem,a user association scheme is proposed based on iterative minimization of weighted delay increment.A parameter is introduced to identify the delay requirements of user.When allocating the base station resources,the target is to minimize the weighted system delay.The suitable base station for each user is obtained by reordering the priority of user selection through multiple iterations.The simulation results show that the proposed scheme can achieve a lower delay than the traditional user association scheme with maximum Signal to Interference plus Noise Ratio(max-SINR).Therefore,it could better meet the users'delay requirements.In terms of unbalanced network load in the 5G heterogeneous dense network,a load balancing scheme is proposed for optimization of both system throughput and system delay with limited backhaul.The proposed scheme is based on Q-learning and the improved ?-greedy strategy.Each small base station learns to obtain a bias factor table that reflects the throughput performance and service quality.When allocating load,three factors are jointly considered including the user's Signal to Interference plus Noise Ratio(SINR),delay increment and base station expansion bias.Different from the traditional scheme,the proposed scheme is based on the specific requirements of user's services in terms of throughput and delay,taking into account the impact of the limited backhaul.Therefore,it is more in line with the actual network capability during the process of load allocation.Compared with the traditional scheme,the proposed scheme can obtain obvious performance gain on both the system throughput and the system delay,while achieving a load balance of the network to a certain extent.