Hybrid applications of constraint satisfaction and meta-heuristics to railway timetabling: a comparative study

The paper provides a discussion on the train scheduling algorithms developed to create predictive schedules for a fleet of passenger trains which travel along a primarily single-track railway with some double-track stretches. The algorithms are an object oriented constraint based heuristic and two hybrid methods, which represent combinations of the heuristic with tabu search and simulated annealing search control strategies. The data of infrastructures and working timetables of passenger trains of Iran's railway system are used to evaluate and compare the generated schedules. Three delay based performance measures, i.e., sum of weighted waiting times (SWWTs), average of unit waiting time (AUWT), and maximum ratio of waiting time to journey time (MRWJ), are used to evaluate the schedules and compare algorithms, where AUWT and MRWJ are introduced by the authors. Simulation experiments show the success of the three methods and the relative advantages and disadvantages of the algorithms are discussed.