| People's communication mode has changed due to mobile communication.The positioning of 5G is not only that it can provide users with a more services and better experience,but also connect to the industry's network.The widening of 5G in the global application field has made the network be more complicated in the process of the system design and optimization,and further expanded its support capabilities in network applications to reliability,delay,and data transmission rate,connection density and other key performance indicators(KPI)[1].Artificial Intelligence(AI)technology is also providing opportunities and possibilities for the design and optimization of 5G communication systems that far exceed traditional concepts and communication performance[2].In order to cope with the diversified user service demands and network load pressure brought by the explosion in the number of communication access devices and mobile applications in the future,the network needs to introduce multiple technologies such as the heterogeneous and dense structure of cellular networks,intelligent management of base stations,and sharing of spectrum resources in the same frequency band.Based on the existing base stations and wired links,it is considered to combine millimeter wave communication with wireless backhaul solutions to form a millimeter wave hybrid backhaul network under dense small cell deployment.Such a network can not only meet the performance requirements of 5G communications,but also reduce network deployment costs and increase flexibility.At the same time,algorithms in traditional research methods(such as optimization theory and game theory,etc.)lack the ability to respond in real time in a dynamic environment,and have gradually become unsuitable for more complex heterogeneous cellular networks.This article will use the efficient data processing capabilities and powerful decision-making capabilities of intelligent algorithms to solve the problems of base station deployment and backhaul link design in millimeter-wave heterogeneous cellular networks based on the optimization method modeling.The main research contents are summarized as follows:First of all,under the architecture of the millimeter-wave heterogeneous cellular network based on hybrid backhaul,the performance index of cost efficiency is used as the basis for the deployment of the millimeter-wave backhaul access network,and the goal is to maximize cost efficiency.The theoretical model of the deployment framework of small cells in the network is presented.In this model,the characteristics of cluster distribution of small base stations are considered,and the theoretical tools of random geometry are used to derive the mathematical expressions of deployment cost and network capacity.For this complex nonlinear optimization problem,an improved simulated annealing(SA)intelligent algorithm is used to verify the correctness of the model and the validity of the proposed base station deployment framework.The research results in this paper provide useful guidance for the optimal deployment of base stations in millimeter wave mobile network planning.In addition,in the mmWave Integrated Backhaul and Access Network(mmWave IBAN),the resource requirements and performance at the backhaul and access levels have mutual coupling characteristics.At the same time,changes in user mobility and quality of service(QoS)requirements will also bring about continuous changes in network traffic load status.These variability make the millimeter wave wireless backhaul link design more complicated and difficult to control.Therefore,it is very necessary for us to dynamically change the resource allocation between the backhaul and access links based on the analysis and prediction of user behavior,and design the wireless backhaul link based on the results.Each base station in the area will transmit information with associated users.Modeling is based on data such as user equipment location updates and data transmission rate requirements.Improved long and short-term memory(LSTM)machine learning algorithms are used to analyze and predict this type of data with strong time series.It is verified by simulation experiments that the algorithm proposed in this paper can obtain more accurate prediction results than traditional LSTM algorithms.The results obtained by this algorithm can more effectively reflect the area's demand for network resources,so that the quality of service can be guaranteed under the condition that the total network resources remain unchanged.The user's network requirements and mobility not only affect the resource allocation of the access network,but also indirectly affect the performance of the backhaul network and even the entire network system.Using the accurate prediction results of user behaviors to pre-allocate wireless network resources can optimize network resource configuration as a whole,and has a better effect on improving the performance of mmWave IBAN.Finally,we studied the multi-user multi-base station spectrum resource allocation strategy in mmWave IBAN on the basis of user behavior prediction.After considering the user's mobile path and network service requirements in the network,the prediction results are used to dynamically change the resource allocation between the backhaul and access links,which can better meet the user's immediate demand for dynamic traffic on the network.In this paper,under the condition that the user's quality of service QoS and network throughput requirements are met,the system energy efficiency is maximized through the allocation of spectrum resources.This paper uses machine learning ideas to model the spectrum resource allocation problem of mmWave IBAN as a Model-free Markov decision process.Then the deep reinforcement learning algorithm DRQN combined with recurrent neural network LSTM is used to solve this problem.A dynamic resource allocation solution that is dynamic and flexible and can interact with the environment in real time is obtained to solve the problem of spectrum resource scheduling in millimeter wave access and backhaul networks. |
PDF Full Text Request