Parallel Crew Scheduling on Workstation Networks Using PVM* Christos Goumopoulos 1, Efthymios Housos 1 and Olle Liljenzin2 1Department of Electrical & Computer Engineering, University of Patras, GR-26500, Greece 2 Carmen Systems AB, S-41103 Gothenburg, Sweden Tel.: +30 61 997304, Fax: +30 61 997316 E-mail: {goumop, housos }@ee.upatras.gr, olle@carmen.se Abstract. In this paper the ability to efficiently solve large crew scheduling problems on a network of workstations (NOW) is presented. Large crew scheduling problems from the Lufthansa set of problems have been solved with a near linear speedup on the generator component of the problem. The generator is the most time consuming component of the solution process, which implies that a significant improvement of the overall solution process is possible. This paper presents the first tangible results of the HPCN Esprit project PAROS, where the complete crew scheduling procedure on a NOW is parallelized and extended. 1 Introduction and Motivation The long term planning and the scheduling of resources in the transportation industry is a very complex and time consuming process. This process can be usually seen as a sequence of phases which proceed sequentially and have minor feedback links. The following figure illustrates the typical planning process in the airline industry. Timetable ] ~ Aircraft Rotations t j Crew Pairings ~ Actual Crew t Construction Construction I r[ Construction Assignments Fig. 1. The resource planning process in the airline industry The main problem every airline must solve, after the construction of its aircraft rotations for a given planning horizon, involves the optimal construction of crew pairings and their assignment to its crew members [1]. Given a set of flights (legs) that an airline must fly, the crew scheduling problem is to find a set of legal round-trips (pairings) to satisfy all the contractual and safety rules of the airline while covering each flight in the timetable at a minimum cost. The crew pairings can only start and end at pre-defined crew bases. The solution of the crew scheduling problem for a large airline is a very time consuming and deadline driven procedure. All airlines would prefer to schedule their crews as late as possible, in order to minimize possible fleet changes and also give the marketing department time to better gage the market needs. In recent years, many of the European airline companies have invested in automatic tools for crew planning and scheduling. Carmen Systems has developed an automatic crew scheduling product [2], that is currently in production use by many of the major European airlines. However, the runtime required for the solution of large * This work is partly funded by the European Commission, within the HPCN domain of the Esprit programme (project PAROS, No. 20.115)