Research On Resource Allocation Algorithm In High-speed Mobile Communication Scenarios

With the advent of smart railway era,more and more links will be established between infrastructure,passengers and trains.Automatic train driving,railway Internet of Things,etc.,are emerging as potential applications.High-speed trains,especially inter-city trains,will realize the transportation organization mode of "high density,large capacity,small group and public transit",and the train distribution will be more intensive.High-speed railway(HSR)communication,which introduce technologies such as artificial intelligence,5G will be safer and more reliable.The high mobility of HSR makes it more stringent on time delay.D2D technology and MEC technology promoted by 5G can effectively reduce the delay loss from terminal to base station or cloud server and improve the service quality of users.In HSR communication,the contradiction between the limited resources and the ever-increasing service demand is increasingly prominent,so it is urgent to study the resource allocation algorithm for high dynamic scenarios.This paper is supported by the National Natural Science Foundation of China(Grant No.6182001)and the Key Project of the National Natural Science Foundation of China(Grant No.61531007),the resource allocation problem in the HSR communication scenario is studied.This paper considers the application of D2D communication technology and MEC technology in T2T scenario and T2I scenario of HSR communication.Traditional resource allocation algorithms are mostly centralized and take a long time to iterate,so it is difficult to guarantee their effectiveness under high dynamic environment.In order to efficiently and reliably meet the service requirements under different scenarios,this paper proposes a reasonable distributed resource allocation algorithm based on deep reinforcement learning to improve system performance.The specific innovation work is summarized as follows:(1)The main demand of T2T links in HSR communication is periodically sharing security information,and T2T links sharing spectrum resources with T2I links to improve the spectrum utilization.In order to meet the reliability requirements and delay requirements of T2T links and reduce the interference to T2I link,this paper proposes a DDQN resource allocation algorithm based on deep reinforcement learning.In the algorithm,each T2T link is regarded as an agent,selecting the spectrum sub-band and transmit power in a distributed manner,which can be based on its locally observed channel state and data transmission situation without waiting for global information of the network.The model has been properly trained under the guidance of the reward function.The simulation results show that compared with other schemes,the proposed algorithm can ensure the superior packet transmission successful probability of the T2T link and mitigate the interference to the T2I link.The algorithm has can robustness,and can achieve good performance under different data packet size and moving speed.(2)The main demand of T2I scenario in HSR communication is high-data task calculation.In order to provide resource access for the calculation tasks generated by trains and passengers on demand,MEC servers are mounted on base station and unmanned aerial vehicles(UAVs).In order to complete as many computing tasks as possible under the premise of satisfying QoS,A multi-agent DDQN resource allocation algorithm based on deep reinforcement learning is proposed.In the algorithm,MEC servers are regarded as agents,making access decisions and allocate spectrum resources,cache resources and computing resources to each computing task with different delay requirements in a distributed way,which can be based on their local observed train positions and task requirements.The simulation results show that compared with other schemes,the proposed algorithm has higher QoS satisfaction rate and resource occupancy rate.