Use of multi-agent system to switch driving strategy in rail transit and procedure simulation

The present paper proposes a multi-agent control system for rail transit. The multi-agent system (MAS) consists of multiple train agents, station agents and a central agent. The implementation of the proposed system is based on distributed optimal control. In the proposed system, each train agent can directly obtain information from the neighboring train agents. Each train agent consists of five subsystems, which are train data set, safety inspection system, timetable inspection system, energy optimization system and trajectory generate system. The train agent can optimize the speed trajectory according to the running state of adjacent trains with the cooperation of the five subsystems. The built-in algorithm of the proposed system can infer the driving strategy (such as time-priority or distance-priority) that should be adopted based on specific situations, which provides a good anti-disturbance ability. Furthermore, the distributed optimization method enables the system to perform a multi-objective optimization in a short time when the system is disturbed.