AUTOMATIC TRAIN OPERATION (ATO) CONTROL SYSTEM USING ROBUST MODEL PREDICTIVE CONTROL ON 2-CARS AGT TRAIN

• Main Author: Hermanto
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/36475
• Institution: Institut Teknologi Bandung

Nowadays, the development of signalling system technology in Indonesia is entering a new era with the start of the construction of rail-based transportation infrastructure for urban areas or known as urban transit such as Mass Rapid Transit (MRT) in Jakarta, Light Rail Transit (LRT) in Jakarta-Bogor-Depok-Bekasi (Jabodebek) and Palembang, and Automated People Movers (APM) at Soekarno-Hatta International Airport, Tangerang. Signalling technology that will be applied to the urban transit is a new signalling technology that had never before existed in Indonesia, namely Communication Based Train Control (CBTC) and European level 1 Train Communication System (ETCS). In the CBTC signalling system with a high degree of automation, the train regulation function is no longer carried out by the operator but is controlled by the system automatically. The subsystem to handle this function in the CBTC signalling system is the Automatic Train Operation (ATO) subsystem. In addition to carrying out the train control function, the ATO subsystem also automates opening and closing the train doors and arranging train departures based on the train dwelling time at the station that has been set. To be able to perform this automatic control function, ATO is given a train speed profile that must be followed by ATO. This speed profile is formed between one station and the next closest station based on the travel time that must be reached by train and the travel distance between the two stations. To form a speed profile, the Pontryagin principle method is used in which the travel time limit and the distance between stations are boundary conditions (initial conditions and final conditions) of the system. This speed profile will then be the setpoint for the designed control system. The control system is designed using Model Predictive Control (MPC). The MPC controller is a model-based controller, so the predicted output is calculated against the model of the train system to be controlled. Predictions are carried out by interval sampling on a finite prediction horizon and the control signal input given to the train system is the result of optimization with objective functions minimizing changes in control signals and errors. In real conditions, there is uncertainty in the mass of the train caused by passenger flows during certain operating hours, especially during peak hours. To deal with this uncertainty, the control system is designed to have a breakdown in the changes in the mass of the train. In this research, the Robust Counterpart method was used to deal with these problems by including the MPC problem formulation with the Uncertainty Set selected in the form of ellipsoidal. The simulation results show that the ATO control system design has been successfully carried out both on the nominal mass value and the mass uncertainty range value. The ATO system can perform the train speed control function against the speed profile setpoint with the RMSE value of 0,1865 and IAE of 53,7276 at its nominal value. While considering mass uncertainty, the RMSE results were 0,3351 and IAE 114,6638.