Scientific conferences play a substantial role for researchers from all over the world, as they offer a centralized platform for the exchange of experience, knowledge and news on the currently hot topics in the respective fields of research.

In general, scientific conferences consist of a collection of talks or sessions that are given by authorized speakers. The general conference planning problem seems simple: for every session one has to determine a possible time and room slot. But in addition, there are complex dependencies between the sessions such as local, temporal and clustering constraints which make a manual assignment almost impossible. Since the number of talks can grow into thousands at large conferences, the allocation problem quickly becomes a very hard challenge. Therefore, we propose a general Mixed-Integer Programming (MIP) framework for computing the schedules of large conferences. It consists of a multi-stage optimization approach, in which the entire problem is decomposed into specific subproblems that are solved consecutively.

Our optimization model was implemented and successfully applied for the International Symposium on Mathematical Programming (ISMP) 2012 in Berlin, which is the internationally largest conference in the field of applied mathematics and mathematical optimization.