| With the increasing demand for network communications in public life and the digital economy,wireless communication technologies are constantly evolving.In addition,emerging applications and services including automated driving,industrial automation,E-health,smart grids,and the Metaverse are also seeking the blessing of next-generation mobile-cellular networks.In 5G-advanced and 6G systems,Ultrareliable Low-latency Communication(URLLC)is essential for numerous applications with stringent reliability and latency constraints,as well as being the most challenging aspect of wireless communications.Therefore,the research of URLLC has been the focus of academia and industry in recent years.In scenarios such as the Internet of Vehicles(Io V),Industrial Internet of Things,and Tactile Internet,besides latency and reliability requirements,URLLC also demands spectral efficiency,throughput,and energy consumption.And with the spread of latency-sensitive services,URLLC traffic in mobile networks will proliferate.The consequent problems are shortage of wireless resources,severe internal interference,network congestion,and worsening queuing overhead.Therefore,with the limited wireless resources,an efficient wireless resource management method is urgently needed to guarantee the end to end(E2E)URLLC performances.Millimeter Wave(mm Wave)band is the mainstream frequency band for 5G-advanced and 6G,while Time Division Duplexing(TDD)has the advantages of flexibility in spectrum utilization and spectrum efficiency.Therefore,this thesis is started from the TDD-based URLLC wireless resource management and divides the E2 E URLLC optimization problem for mm Wave transmission into three parts to accomplish.Specifically,a new coding scheme for URLLC short packets is proposed in the physical layer,and then URLLC resource optimization is performed for downlink and uplink,respectively.The main innovations of this thesis are as follows:1.A method of optimizing URLLC spectrum resources based on Analog Fountain Code(AFC)is studied for short packet transmission in the physical layer of the mm Wave communication network.At first,a CRC-AFC cascade coding scheme with a feedback mechanism is proposed by analyzing the mm Wave channel model for finite data block length.In the scheme,the strong error detection capability of CRC is combined with the rateless characteristic of AFC.Moreover,to control the coding cost and improve the coding reliability,a once re-encoding mechanism is designed to dynamically adjust the CRC-AFC cascade code rate by analyzing the URLLC delay constraint and the AFC bit error rate(BER).Then,a dynamic spectrum allocation algorithm for URLLC based on CRC-AFC is presented by analyzing the dependencies among the URLLC latency constraint,the non-URLLC signal noise ratio(SNR)threshold,and the CRC-AFC BER.On the premise of ensuring the quality of nonURLLC service and spectrum efficiency,this algorithm dynamically allocates spectrum for URLLC according to code rate and data volume so as to reduce URLLC transmission latency.Simulation results show that the proposed algorithm can achieve high reliability and low latency performance in a wide range of SNR.The latency advantage of the proposed algorithm increases with the increase of URLLC users,and the reliability advantage increases with the decrease of channel quality.2.To better support critical-mission intensive communication scenarios,a Reinforcement Learning(RL)based hybrid spectrum allocation method for high-load URLLC is investigated for the URLLC latency optimization problem during downlink transmission in mm Wave hybrid spectrum access networks.At first,the active packet dropping and temporary dumping mechanisms are designed by considering the system latency constraint and long data preference problems under the high load resource demand of URLLC,respectively.On this basis,a greedy resource allocation algorithm based on time-varying resource state transfer is proposed,which is made more predictable and flexible by a hierarchical cache packet drop policy.Then,an RL-based hybrid spectrum resource allocation algorithm is proposed to reduce the computational overhead,where the temporary dumping mechanism can effectively compress the state space while decoupling action from time can compress the action space.To further reduce the computational overhead,a multipath-DNN is designed to approximate the optimal policy by reserving exits in each hidden layer of the Deep Neural Network(DNN).Simulation results show that the two proposed algorithms can both maintain high reliability and reduce URLLC latency under high load conditions.In particular,the RL-based resource allocation algorithm achieves a latency close to 1 ms and a packet loss rate below 17% when the overload is below 30%.3.For the coexistence of URLLC services and non-URLLC services with different performance metrics,a URLLC spectrum-power jointly optimization method in E2 E transmission of Io V is studied.At first,a joint iterative resource allocation algorithm with the objectives of maximizing non-URLLC throughput and URLLC delivery rate is proposed by analyzing the vehicle movement model and mm Wave interference model.And the above multi-objective optimization problem is converted into a single-objective iterative optimization problem by problem decomposition.Then,a Multi-agent Reinforcement Learning(MARL)resource allocation algorithm based on cooperative competition is proposed for the high mobility of vehicle movement and the difficulty of centralized resource management in the uplink,in which the vehicles are treated as agents.Every agent in the algorithm is only required to observe the local Channel State Information(CSI)and share the local CSI through a federated mechanism to optimize the global performance,while a penalty mechanism is implemented in the algorithm to enhance the importance of URLLC services.Simulation results show that the proposed scheme has better extensibility as well as superior non-URLLC throughput and URLLC delivery rate performance compared to other MARL-based comparison algorithms. |
PDF Full Text Request