Research On Multiple Access Protocols For Next-generation Wireless Distributed Networks

With the rapid development of wireless communication networks and mobile Internet technologies,a distributed wireless network composed of smart terminals has become an important part of the next generation wireless communication networks.Because the next-generation wireless network has the characteristics of many kinds of service and a large number of users,these characteristics increase the randomness of service arrival and the randomness of service transmission,and lead to more difficulty for users to share channel resources in a distributed network.Therefore,it is necessary to study a new type of multiple access protocol suitable for multi-user wireless distributed networks to solve the shortcomings of existing protocols in applications.This paper analyzes that the multiple access problem of distributed wireless networks is affected by the randomness of service arrival and the randomness of service sending,and adopts deep learning algorithms to predict business traffic and study the random characteristics of the business arrival process.According to the service arrival characteristics,the deep reinforcement learning algorithm is used to study the service delivery strategy,and a new service delivery strategy is proposed.The specific content of this article is as follows:1.Aiming at the scenario of base station traffic prediction,this article firstly studies the actual traffic information obtained at the base stations on a time scale.This paper secondly analyzes the physical meaning of the various signaling information collected at the base station,and uses the Pearson correlation coefficient to measure different information.The degree of correlation between the information and the flow can be judged to determine the influence of different signaling information on the flow prediction,and the use of CNN to verify that the introduction of signaling information improves the accuracy of the flow prediction.Based on the influence of different signaling information on prediction,this paper designs a multi-head CNN model using different convolutional layer parameters to separate and sample different signaling information to maximize the extraction of effective information and reduce redundancy.The influence of the remaining information has improved the accuracy of the forecast.2.Further considering the spatial correlation between multiple base stations,this paper firstly uses KL divergence to measure the similarity of the traffic of different base stations,and secondly designs a CNN model based on the migration learning idea,using multiple base stations traffic information.In this method,the pre-training the model is used to capture the common characteristics of traffic changes.The traffic information of the target base station to fine-tune the model is to obtained.The individual traffic characteristics is used to realize the joint spatio-temporal prediction of communication traffic.This method can further improve the accuracy of the prediction.3.Aiming at the multi-user channel resource allocation scenario,on the one hand,The service transmission problem in the single channel and light service load scenario is studied in this thesis.The time-slot resource allocation scheme using the prior information of the random service arrival process is designed.The deep reinforcement learning based on the independent multi-agent structure is used in channel resource allocation scheme.There is no the information exchanging between the two nodes in sending process.According to its own business conditions,combined with the historical information of channel occupation,the sending process is autonomously decided by the agent.The experimental results show that,in general,the algorithm makes the nodes that are initially in the random sending state gradually appear a certain regularity in the service sending process.The each node individually decides to send by predicting the probability of the channel being occupied.Because this resource allocation scheme reduce the conflictions in resource occupation,the network throughput is greatly improved.The problem of service transmission in multi-channel and heavy service load scenarios is also studied in this thesis.An time-frequency resource block allocation scheme using independent multi-agent structure is proposed in the distributed OFDMA network.The experimental results show that the algorithm makes the node with initially selecting a resource block at random tend to select different time-frequency resource blocks for service transmission during the sending process.This method reduces the probability of collisions and improves the throughput of the network.The virtual traffic is introduced in this thesis.The virtual can maintain a good throughput and enhances the robustness of the algorithm when the business load fluctuates.