Research On The Application Of LTE-U Technology In Train-ground Communication Of Urban Rail Transit

The Specification for Train-Ground Integrated Communication System of Urban Rail Transit(LTE-M),which is based on Long Term Evolution(LTE)technology to carry information transmission of urban rail transit,requires that the communication system need to be deployed in the licensed spectrum band of 1785MHz～1805MHz.Due to the increase in safety requirements and the development of smart subways,the data that need to be transmitted between train and ground has increased significantly in recent years.The spectrum bandwidth with 20 MHz at present is gradually difficult to meet the transmission requirements.With the development of unlicensed spectrum domain Long Term Evolution(LTE in unlicensed,LTE-U)technology,more and more unlicensed spectrum resources are available.Therefore,This paper proposes a scheme which is deploying the train-ground communication system of urban rail transit in the unlicensed frequency band.Firstly,this paper analyzes channel access mechanism of the Wireless Local Area Network(WLAN)system and train-ground communication system of urban rail transit,and finds that the deployment of train-ground communication system of urban rail transit in unlicensed frequency domain will lead to the channel be occupied in long-term.That will lead to the WLAN system which is already deployed in this unlicensed frequency domain lost the authority of channel access.Therefore,a spectrum resource allocation scheme based on the duty cycle is proposed.With the allocated spectrum resources,WLAN system and train-ground communication system of urban rail transit alternately obtain the channel to realize their data transmission respectively.Secondly,in order to use resources efficiently,this paper uses Q-learning algorithm,a model-free algorithm in Reinforcement Learning(RL),to achieve dynamic allocation of spectrum resources.The dynamic allocation strategy is a Markov decision process.The process is mathematically abstracted to turn it into a solvable state value function.Finally,the optimal strategy is found through the state value function.According to this strategy,spectrum resources are allocated for train-ground communication system of urban rail transit and WLAN system.The simulation results show that the spectrum resource allocation scheme based on the Q-learning algorithm is significantly better than the fixed duty cycle scheme in terms of throughput and user satisfaction.Finally,in order to better perceive the business needs of WLAN users,Deep Reinforcement Learning(DRL)technology with capabilities of perception and decision-making is introduced,and the Deep Q Network(DQN)algorithm in DRL is used to achieve resource allocation of unlicensed frequency domain.The DQN algorithm completes the extraction of state features such as spectrum resources and users business requests by introducing a neural network into Q-learning,and the final resource allocation scheme is output through the deep Q network.The simulation results show that this scheme can achieve a higher WLAN data packet delivery rate while ensuring the data transmission of the train-ground communication system of urban rail transit,and complete the data transmission of the two systems in the unlicensed frequency band with higher quality.