EVALUATING MRT SCHEDULING IN DHAKA: A DYNAMIC, DATA-DRIVEN APPROACH USING GENETIC ALGORITHM AND DEEP LEARNING METHODS

In rapidly expanding cities such as Dhaka, effective metro scheduling is essential to enhance servicequality and reduce operational inefficiencies. Currently, it operates on a fixed schedule. This fixedtimetable is ineffective as it does not adjust according to the number of individuals present at the stationat any given time, particularly during peak hours. This may result in increased overcrowding, extendedwait times for passengers, and diminished resource efficiency. This paper presents a dynamic schedulingframework designed to enhance the efficiency of Dhaka's MRT operations. The architecture integratesdeep learning for demand forecasting, real-time passenger monitoring, and heuristic optimization. AnLSTM model, trained on historical data, is employed to forecast the number of passengers expected toarrive in the near future. Simultaneously, real-time inflow assessments from RFID and CCTV systemsare continuously monitored to identify any anomalous demand trends. An optimizer employs a GeneticAlgorithm (GA) to identify the optimal scheduling system and allocates the subsequent headway for themetro trains. The objective is to reduce inefficiencies for both passengers and operators. The proposeddynamic schedule is evaluated against the existing fixed timetable using synthetic demand scenarios. Thecost of passenger wait times, operational expenses, and service reliability are all critical performancemetrics. Preliminary findings indicate that the dynamic timetable has improved conditions for passengersand operational efficiency. The optimized system reduces wasted work hours and positively impacts theenvironment by consuming less energy. The proposed framework is a feasible and scalable method toenhance urban rail operations and facilitates the development of intelligent transportation systems inDhaka.