Resource Allocation for High-Speed Train Communication Based on Deep Reinforcement Learning

Abstract

With the advent of high-speed railway era, users' demands for high link capacity and high reliability are increasing. In this paper, a resource allocation algorithm based on deep reinforcement learning (DRL) is proposed in the high-speed railway (HSR) scenario. The communication link between high-speed train and high-speed train (T2T) is taken as an agent, and the link capacity and packet transmission success probability (PTSP) are taken as optimization objectives, which are mapped into the reward function. The agents gain experience by constantly interacting with the HSR communication environment to update the neural networks in DRL and eventually learned the optimization policy. The simulation shows that when the trains' speed is 350km/h and the transmission packet size is 2385Bits, the high reliability of T2T link can still be maintained, and the capacity of high-speed train to Infrastructure (T2I) link is improved by 1Mbps compared with the baseline, which proves that the proposed algorithm can mitigate the interference introduced by spectrum reuse and doppler shift.