How to solve a timetabling problem by negotiation

Problems involving resource allocation such as appointment taking or timetable creation have usually been tackled as constraint satisfaction problems. For the two aforementioned applications, this forces the users to reveal their agenda in order to give their constraints. But one doesn’t always want to reveal his private events. Thus a solution is for the users to keep their agenda private and to use negotiation to take appointments or courses. In this article, we present a negotiation based approach for timetable creation. Our aim here is not to present an approach that gives the best results but an original one that is more flexible than others which must restart from scratch at each change in the environment. Using agent based negotiation enables us to dynamically add or remove constraints or agents. We show how to transform the timetabling problem into a negotiation problem between agents. Note that an agent can represent a person as well as a thing such as a room. To illustrate our purpose, we use a university timetable creation application. A benchmark has been proposed by the group Asa GDR-I3 [1]. This problem needs, in order to provide a solution, to be able to adapt itself in response to dynamic changes in the environment. This problem needs a collective search of the solution, and isn’t a simulation problem: the aim isn’t to recreate virtually the behaviour of an existing organism, but to furnish an expertise. In order to develop this application, we use a generic negotiation API: GeNCA [2]. This API provides the whole management of negotiation processes and only needs communication configuration and strategy definition. For the application we want to develop, there’ll be no work for communication configuration as we’ll use the provided MAS communication (either Magique or Madkit). Default strategies are also provided with GeNCA which are well-suited for our application. In this paper, we first present the negotiation approach for resource allocation problems and the university timetabling creation application that serves our purpose. Then, we present the negotiation toolkit used to build this application, the way to develop it and the results of this approach.