Automated System for University Timetabling

Although university course timetabling is a widely studied topic, the use of automated timetabling systems is not widespread among large universities. This is particularly true in the United States, where the state of the art is typically to roll forward the last like semester’s timetable and make adjustments to room assignments. University timetabling is a hard problem because of its size and the complexity of constraints needed to satisfy the demands of students and instructors. The problem is made harder yet by the need to develop a system that is easy for everyone involved in the process to use and understand, and for them to be satisfied with the results. The system described here, and currently being implementation by Purdue University, successfully deals both with the issue of solving a large-scale problem and with addressing many of the human factors required in real applications. The size of the problem (9,000 classes, 570 rooms, and 39,000 students with 259,000 class requests) has been made more manageable by decomposing it into a large lecture problem, consisting of centrally scheduled classes serving students in many disciplines, and multiple departmental problems. This partitioning also addresses the need to give departmental timetablers ownership of the process, which is important in a complex organization. Departmental timetablers have invaluable knowledge about what is and is not important in a solution that would be extremely difficult to incorporate into a black box solver. It is important to be able to leverage this knowledge with tools that can help sort through all of the constraints and costs to find solutions that satisfy user needs. The primary design goal was therefore to assist academic timetablers with the problem of building a good timetable, not necessarily finding a true optimal solution.