Design of Vehicle-based Driving Environmental Recognition Framework and Deep Learning Algorithm for Autonomous Train Control System

The ground-based environmental recognition system transmits information to trains through sensors, signal control room and Automatic Train Supervision (ATS) with human judgment / intervention. These cause limitations to active and intelligent train operation. We propose a framework for a vehicle-based environment recognition system for an Autonomous Train Control System (ATCS). The vehicle-based environment recognition system directly transmits information from the sensor to the train, and is responsible for all steps of collecting, recognizing, classifying, determining, and controlling information. We also analyze artificial intelligence technology to process multiple types of sensing information in the vehicle-based environment recognition system of ATCS. The Multi-Point Long Short Term Memory model, which is suitable for processing time-series sensing information, is used for deep learning. The deep learning model is studied using currently available Busan Metro Line 1 metro data. Driving data is learned through data selection, processing, and parameter optimization. Based on this, the speed of the train is predicted and the performance of the deep learning model is verified by comparing it with the actual speed data.