A MultiAgent Architecture for Distributed Course Timetabling

The course timetabling problem consists in scheduling a set of lectures in a cyclic fixed period of time, typically a week. We consider the course timetabling problem for a set of university departments, in which each department prepares the schedule for its curricula according to private rules, constraints, and objectives, and relying on own resources. We consider the typical scenario in which resources are not common to departments, but departments could negotiate in order to share and/or exchange them for mutual benefits. For this problem, we propose a scheduling system based on a multiagent architecture that makes use of a marketplace and an artificial currency. In this framework, each department has a team of three cooperating agents, which are responsible for different tasks: searching for a local solution, negotiating resources with other departments, and managing relevant information. To prove the effectiveness of the architecture, we present an experimental analysis that shows the benefits of the system in the real situation of our institution.