Constraints Based Effective railway Rescheduling System

Delays of trains occur owing to accidents, bad weather conditions, vehicle troubles, etc. Trains run according to the departure order of trains at each station (called "SOT") in the scheduled plan. Therefore, a delay of a train delays its succeeding train, which may result in a chain reaction of delays and confusion arises to the whole network finally. The rescheduling needs to adjust SOT and the order of using resources (called "ROT"). Rescheduling involves many factors intricately related. One of major factors is related to resources such as line sections, arrivaldeparture tracks, cars, drivers, etc. To use these resources, there are many constraints such as safety requirements, performance of cars, labor regulations, etc. Some constraints may break out dynamically among a group of trains which are interrelated by SOT. A group of trains are also related each other by car rostering. In view of each car, a trajectory of its movement is folded over at each terminal station of trains operated by this car. As about 5 to 15 trains are operated in an hour from each direction, these trajectories exhibit a chaotic nature. As a result, a decision on a train at a station may greatly change the future situations of succeeding trains. At present a train dispatcher carries out a reschedule on the basis of his expertise with the help of a traffic simulator. However, if a large scale disorder occurs, it is difficult for a train dispatcher to predict the train diagram of the future and to suitably reschedule them. For example, a one hour delay may influence over 6 hours in the Shinkansen Network of East Japan Railway Company (JREast). To support the decision making process of a train dispatcher in these rescheduling environments, we have been studying the architecture of this decision support system and adopted the constraints-based approach. Implementation of the prototype system CIBERS is based on constraint programming packages provided by ILOG. Case studies have been carried out with real data from the Shinkansen Network of JR-East.