| As a low-power wide-area Internet of Things technology,NB-IoT has greater coverage than GSM and LTE,and the protocol adopts a three-level coverage category,so that all terminals under the same coverage level use the same uplink coverage enhanced configuration.The coarse division granularity leads to the waste of some terminal resources.In this thesis,the optimal configuration scheme of uplink coverage parameters is studied according to the actual uplink channel quality of terminals.The research is of great significance for ensuring the reliability of NB-IoT uplink communication and reducing the delay of uplink communication.Deep reinforcement learning used for decision processing solves the problem of training instability in the field of strategy optimization,therefore,based on deep reinforcement learning technology,the optimization configuration of uplink coverage enhancement parameters on NB-IoT network is studied to ensure its coverage performance,and the following research work is mainly completed:(1)The existing research results of NB-IoT coverage enhancement are reviewed and discussed.It is pointed out that the number of uplink repeated transmissions,the interval and number of subcarriers,and the MCS(Modulation and Coding Scheme)should be considered in the optimization configuration of NB-IoT uplink coverage enhancement.At the same time,the advantages of the combination of deep reinforcement learning technology and wireless communication are discussed,and the idea of guaranteeing NB-IoT uplink coverage enhancement basing on deep reinforcement learning technology is proposed.(2)Aiming at the problem of coarse coverage enhancement settings in the existing NB-IoT protocol,a DDQN-based NB-IoT uplink coverage resource allocation scheme,DDQN-CE(Double DQN based Coverage Enhancement)is proposed.The scheme considers the uplink state of each UE,and changes the NB-IoT uplink coverage resource allocation problem to a deep reinforcement learning process.The optimization goal is to make the uplink data transmission reliable(BLER≤10%)and minimize the uplink communication delay.The agent trains and learns through information interaction with the NB-IoT communication environment,so that the neural network parameters of the agent converge,and finally realizes the optimal configuration of the multidimensional parameters of the NB-IoT uplink coverage.Based on the NS3-Gym open source framework,the scheme is simulated and analyzed.It can enhance the reliability of NB-IoT uplink communication,reduce the uplink communication delay,expand the uplink communication coverage,and significantly improve the overall performance of NB-IoT uplink communication coverage.(3)A prototype system of smart street lamps with enhanced NB-IoT uplink coverage is designed and implemented,which can realize remote monitoring and policy-based intelligent control of street lamps.Based on the NB-IoT uplink coverage enhancement optimization scheme,the relevant parameters of the smart street lamp terminal uplink coverage are configured,the uplink coverage performance of the smart street lamp terminal is improved,and The number of terminal connections in the cell covered by the base station has been further increased. |
