DEVELOPMENT OF AUTOMATIC TRAIN REGULATOR FOR ON-LINE RESCHEDULING OF TRAIN DWELL TIME AND DISPATCH FREQUENCY USING TABU-FUZZY OPTIMIZATION TECHNIQUES

Abstract

The purpose of this thesis is to outline the development of a software simulation package named the 'Automatic Train Regulator' or 'ATR' that is used to optimize the dwell time and dispatch frequency of an automated railway system under both steady state and disturbed conditions. Simulations were conducted on the 'ATR' package to assess the performance of mass transit systems both with and without automatic train regulation to highlight the operational advantages of introducing automatic train dispatch to maximize energy savings and passenger comfort without excessive deviations from the trains' pre-established schedule. In this project, an attempt is being made to design an Automatic Train Regulator (ATR) using two controller schemes, namely the Fuzzy and Tabu-Fuzzy Optimization schemes. The on-line rescheduling strategy is to optimize passenger service while maintaining a low operational cost. The 'Automatic Train Regulator' simulation package thus employs a dwell time and dispatch frequency controller which makes use of fuzzy decision-making coupled with Tabu-search techniques to determine the optimal schedules for train dwell times and dispatch intervals based on the criteria of regularity of service, energy consumption, train congestion as well as platform congestion levels. State of the art techniques like object oriented programming and discrete event based modeling have been applied to make the 'ATR' program as versatile and powerful a simulation package as possible. Results obtained through the simulation clearly shows that the Automatic Train Regulator is able to effectively improve the Mass Rapid Transit (MRT) system performance for both normal and disturbed train operation. It does this based on its fuzzy decision making mechanism which takes into account train regularity, energy consumption and congestion considerations. The inclusion of tabu search helps in tuning the weighting coefficients of the various performance indices, which in turn leads to a more justifiable and closer-to-human-intuition decision by balancing the levels of attainment of the different performance indices. Future work may be focused on trying to develop a conflict-based dispatching system for use in multi-track railway networks involving branching and looping. Such a system should be able to assign the right of way at the traffic junctions to achieve orderly and safe passage of the trains, and more importantly to optimize the performance criteria of regularity, congestion and energy considerations based on the decision.