Agent Technology for Timetabling

Agent technology promises possibilities for the design of distributed systems, the design of open systems and the reduction of information overload for the users. Timetabling applications must operate in heterogeneous environments at large organisations where users cannot be bothered with all the details influencing their work schedules. It is a domain where, at least at the system level, many separately developed systems must interact. In this paper, we investigate how agent technology can play a part in timetabling applications. Timetabling problems are distributed by nature but performance also contributes to the distributiveness. The performance issues are related to the complexity of the timetabling problem and the need to respond within a limited time. The problem becomes distributed when departments of a different speciality are interconnected and have to gear their timetables to one another. Timetabling systems will often function as components in larger settings. As such, they have to interact with a diversity of existing solutions and should thus best be build as open systems. Data are stored in databases; business systems may contain relevant information. The results of the timetabling operations must be fed back to these information providers because these systems ultimately control the operations. Timetabling systems need as much detail as possible while they must report on the resulting timetable to users with a limited scope. Both channels must be guarded against information overload. The user should not be prompted to specify information that is not available and he should not be confronted with information he cannot interpret. On the other hand the system should try to understand what the user really wants, present a suitable timetable and provide comments that are easily accessible to that user. The three mentioned key issues in timetabling systems are competences for which software agents are particularly appropriate. We review the available technology for distribution, interconnection and information overload and report on some experiments.