ACBTC: ADVANCED COMMUNICATION BASED TRAIN CONTROL USING DEEP REINFORCEMENT LEARNING

High unwavering quality and low idleness are urgent for metropolitan rail travels. In this paper, we present correspondence techniques for Advanced correspondence based train control (ACBTC) frameworks utilizing long-term evolution for metro (LTE-M) to improve the dependability and inactivity. A multiple input multiple outputs (MIMO) helped handoff (MAHO) plot for ACBTC framework is proposed to lessen the handoff dormancy in this undertaking. During the handoff technique the versatile station speaks with the serving access point (AP) and up-and-comer AP simultaneously with handoff flagging communicated by one radio wire and data bundles sent by another, so the station can handoff from one to the next without intruding on data transmission. Initial, a novel metropolitan rail travel remote correspondence model is set up utilizing FlashLinQ-based Train-to-Train (T2T) interchanges. At that point, we present a novel psychological control conspire dependent on LTE-M with T2T correspondence to improve the nature of administration and the strength of multi-train ACBTC frameworks. In the presented conspire, Q-learning is utilized to create ideal control methodologies considering both remote correspondence boundaries adaption and train control boundaries. Broad reenactments are directed and the outcomes show that the versatility of ACBTC frameworks can be upgraded utilizing the presented plot. Besides, utilizing the presented conspire, not just the holes in ideal speed versus distance bend are more modest, yet in addition the impromptu footing and breaking are decreased also.