Research On 5G Converged Network Resource Allocation Algorithm Based On Networks

From 1G to 5G,mobile communication networks have experienced rapid development.The coexistence of multiple networks is the current status of the entire mobile communication network.With the increase of people's requirements for mobile communication network performance indicators,a single network can no longer meet people's needs.Network convergence can not only improve the utilization of network resources,but also meet people's diverse business needs and improve user experience.However,while converged networks bring benefits,there are also problems.The interference management and resource allocation of the network are facing challenges.In addition,the current optimization methods of wireless network resource allocation also have some shortcomings.The traditional mathematical optimization method is difficult to solve the non-convex optimization problem in complex scenarios,and in many distributed scenarios it cannot be used due to lack of global information.Based on these issues,this paper considers relevant scenarios in 5G converged networks,and applies reinforcement learning methods to wireless network resource allocation,thereby improving network performance.The main research contents of this paper are as follows:Firstly,the paper first studies one of the important communication methods in 5G converged networks,that is,device-to-device(D2D)communication.In order to better reflect the real wireless environment,this paper models the channel as a Finite State Markov Channel(FSMC)to reflect the time-varying channel environment.However,this also increases the complexity of the problem.Traditional optimization methods face many difficulties in solving this problem.At the same time,general reinforcement learning methods cannot be effectively trained due to the large state space in a time-varying environment.This paper combines the powerful sensing capabilities of deep learning with the decision-making capabilities of reinforcement learning,considers the D2D network as a multi-agent system,and abstracts the state,action,and reward functions from the wireless network environment.This paper proposes a centralized Power control algorithm based on Deep Reinforcement Learning(DRL).In addition,this paper also proposes a new reward function that can better improve the quality of user experience in communication.Secondly,in view of the shortage of authorized frequency bands and the difficulty of collecting accurate channel information in research content 1,the paper introduces unlicensed frequency bands in WIFI networks to D2D communication in LTE networks,forming a converged network scenario where WIFI and LTE coexist.This paper proposes a distributed deep reinforcement learning-based mode selection and joint resource allocation algorithm to maximize the throughput of cellular and D2D users while ensuring the quality of service(QoS)requirements of cellular,WIFI,and D2D users.