Open-source 5G RAN Deterministic Latency Communication Methods

The rapid development of mobile communications and internet technology has brought tremendous changes to the society,especially the application of the 5th Generation Mobile Communications Technology(5G).Benefiting from its high bandwidth,low latency,and high reliability,a group of industry applications centered around smart manufacturing,low-carbon efficiency,and IoT have emerged.Particularly in vertical industries such as industrial IoT and smart transportation,there has been a large number of applications with differentiated demands based on latency as a key indicator.At the same time,the demand for deploying artificial intelligence algorithms in these vertical applications on the network is also increasing day by day,and the latency of computing tasks has become an important indicator.These services require support for key characteristics such as full-domain network latency measurement,computing latency control,and air interface latency determinacy.However,the evolution of mobile communication technology has encountered problems such as closed and rigid network architecture,long research and development cycles,and high costs,which have led to a lack of unified data openness and intelligent control capabilities in current 5G networks,making it difficult to meet the above-mentioned needs.Open-source 5G technology aims to introduce software and hardware decoupling protocol implementations,flexible network architecture,open data and network interaction,unified device interfaces,and other characteristics into mobile networks through an open and collaborative research and development model.Leveraging the advantages of open-source 5G technology is expected to provide the ability of latency-deterministic communication for vertical industry applications.Open-source 5G technology is currently in the development phase,particularly in three aspects:high-precision absolute time synchronization in access networks,deterministic control of computing task latency,and efficient air interface latency determinacy,there are still technical gaps and shortcomings.This article focuses on the study of latency-deterministic communication methods in the open-source 5G system to address these issues.The main work of this dissertation is summarized as follows:(1)The open-source 5G high-precision absolute time synchronization architecture,air interface timing method and corresponding hardware and software design have been studied.A sub-microsecond timing prototype system has been realized,which provides basic capability support for delay deterministic communication.This dissertation firstly aims at the overlapping characteristics of multiple synchronization domains in the open-source 5G network topology,and proposes a high-precision absolute time synchronization architecture design.Secondly,the air interface timing error source model is established.Aiming at the errors of each link,a delay estimation and compensation method based on the air interface physical layer signal is proposed,and the air interface timing process design is proposed.Under the premise of minimizing the function changes of the protocol stack and the timing overhead,the sub-microsecond level absolute time synchronization of the whole scene is realized.At the same time,a timing observation control algorithm based on Allan variance optimization,a dual-threshold stabilization mechanism and a multi-signal error perception mechanism are proposed to improve the robustness of the timing process.Aiming at the dynamic channel situation of the terminal,an automatic control algorithm of timing parameters based on deep reinforcement learning is designed.Finally,the software and hardware of the base station timing module and the terminal timing module were designed and developed based on the open-source 5G system,forming an absolute time synchronization prototype system.Tests show that the proposed timing-related algorithms and designs can be introduced into existing open-source 5G systems as independent functions.The average timing accuracy of the prototype system can reach 20 nanoseconds,and the worst timing accuracy is only 200 nanoseconds,which is improved by 2 to 3 orders of magnitude compared with the existing methods.The multiple stable timing mechanisms proposed comprehensively realize the robustness of timing maintenance in various channel environments.(2)The deterministic calculation method of delay based on intelligent management and control of open-source 5G network is studied.Firstly,based on the data openness and network interaction capabilities of opensource 5G,this dissertation designs the intelligent management and control computing engine architecture and protocol,so that the open-source 5G system can natively realize the environmental encapsulation,management and interaction of the intelligent computing tasks it carries.Secondly,a deterministic calculation principle of time delay based on parallel polling is proposed.It sets the delay deterministic control target by estimating the network communication and calculation delay,and then expands the decision-making execution sequence of the intelligent algorithm into the state space of the algorithm accordingly,so that the target algorithm can predict its action execution in the learning process.The relationship between communication and calculation delays ensures the convergence performance of the algorithm,thus realizing effective deterministic calculation delay control in open-source 5G networks.Finally,the efficient and low-latency management strategy of heterogeneous computing resources in the open-source 5G network is studied to reduce the average computing delay and optimize resource utilization and computing energy consumption under the premise of deterministic computing delay.Compared with the existing learning and training methods for intelligent computing tasks in 5G networks,the proposed scheme can effectively support the deployment,update and retraining of algorithms in actual networks with random communication and computing delays,and can robustly implement deterministic delay computation.(3)Key technologies and methods such as resource deterministic scheduling,air interface seamless switching and hybrid redundant transmission required for delay deterministic communication in open source 5G access networks are studied.This dissertation first studies how to solve the problem of large amount of calculation and long solution time caused by NP-hard of traditional static scheduling algorithm,aiming at the requirement of dynamic allocation of deterministic resources brought about by terminal mobility.In this dissertation,a multi-stage scheduling mechanism is proposed to adapt to different mobility states,and a fast dynamic scheduling algorithm combining clustered integer linear programming and adaptive genetic algorithm is designed to realize fast response of static scheduling,real-time response of incremental scheduling and Dynamic reconfiguration optimizes the scheduling effect of resource overhead,and its overall performance is better than that of existing optimal static scheduling methods.Secondly,in view of the problem that the switching of deterministic resources may cause service interruption during cell switching,a sparsity evaluation model of deterministic resources is proposed,and an auxiliary cell selection method based on deep reinforcement learning is designed.Compared with the traditional coordinated multi-point transmission The program has a lower probability of outage.Finally,in order to solve the problems of high overhead and low spectrum efficiency of the existing air interface redundant transmission technology,a time-sensitive channel capacity model is proposed.Based on the analysis of various blind retransmission methods,hybrid blind retransmission mechanisms based on cluster-head-relay and orbital angular momentum communication are proposed respectively.The former uses blind retransmission of D2D communication to reduce the resource overhead of the base station,and the latter uses the pattern nature of electromagnetic waves to improve the spectral efficiency of blind retransmission.The proposed method greatly increases the capacity of the network for deterministic services while using the same air interface resources.