Real-time railway transit management based on multi-agent system (MAS)

Because of environmental concerns, increasing attention has been given to developing energy-efficient technologies. As an essential public transportation method, metro transit is facing increasing pressure. Disturbances may cause the offline-optimized timetable and speed trajectory to be invalid. The present paper proposes a multi-agent system (MAS) train control method, in which each train is controlled by an agent. Each train agent has a simulation platform and an optimization platform. The simulation platform collects information such as track slop, speed limitation, speed trajectories of neighboring trains from a few neighboring agents. The simulation platform sends a signal with the collected information to the optimization platform when it detects a disturbance. Afterward, the optimization platform performs energy-aimed optimization to the timetable and corresponding speed trajectory based on a combination of a trained Neuro Network and a Mixed Integer Linear Programming (MILP) model. The test result shows an encouraging balance in optimization time and accuracy. The case study result proves that the proposed approach could provide a more energy-efficient control strategy when disturbances occur.