Research On Resource Management Of 5G Internet Of Things Based On MEC

With the commercialization of the fifth-generation mobile communication technology(5G),the rapid emergence of Internet of Things applications and the explosive growth of the number of terminal devices have caused problems such as network congestion and high response delay.The application of mobile edge computing technology(MEC)is one of the keys to breaking through the bottleneck of future IoT computing and transmission.However,the current MEC network still faces challenges in resource management.First,it is necessary to optimize the MEC application deployment strategy to balance the network load.Second,it is necessary to design a new physical layer resource allocation scheme to cope with the surge in the number of terminals and the diversification of business scenarios.Based on the above background,combined with the progress of standardization research,this thesis conducts research on MEC-based 5G IoT resource management.The main contributions are as follows:Aiming at the problem of unbalanced load of edge node due to delay and resource constraints when deploying MEC applications in edge networks,an online deployment framework for MEC applications based on deep reinforcement learning is proposed.The scheme considers a multiuser dynamic MEC network,using a policy-based deep reinforcement learning algorithm.First,the MEC application deployment decision is generated through the actor neural network and rewards are calculated according to the constraints;then,the periodic policy update is performed using the critic neural network to minimize the load difference.The simulation results show that,compared with the baseline scheme,the proposed scheme can reduce the sum of the difference of the load ratio between edge nodes to 19.7%.Aiming at the unreliable Quality of Service(QoS)of Reduced Capability devices in edge networks,a random access channel(RACH)resource isolation scheme and QoS control strategy based on network slices is proposed.First,define an independent RACH resource pool isolated from existing resources,and add a new field in the system message to indicate the location;then,define a field in the network slice identification information as a resource activation indicator;finally,define the 5G QoS feature identification field in the network slice identification information to obtain the corresponding QoS parameters The simulation results show that the proposed scheme can effectively improve the access success rate by 70.4%and reduce the access delay by 54.3%compared with the baseline scheme.Aiming at the problem that the single bandwidth part(BWP)and numerology in the edge network sidelink cannot match the business requirements,a Multi-BWP scheme based on resource allocation Mode 2 is proposed.First,multiple BWPs with different numerology are defined through pre-configuration;then,the appropriate BWP is selected according to the service type,and the terminals in the sidelink group perform resource awareness to assist resource pool selection;finally,by defining the switching mechanism,the terminal cluster BWP can be dynamically adjusted to adapt to changes in services and scenarios.The performance evaluation and simulation results of the proposed scheme show that flexible numerology settings can well adapt to application scenarios with different speeds and coverages.