Danzinger, P., Geibinger, T., Janneau, D., Mischek, F., Musliu, N., & Poschalko, C. (2022). A System for Automated Industrial Test Laboratory Scheduling. ACM Transactions on Intelligent Systems and Technology. https://doi.org/10.1145/3546871
Automated scheduling solutions are tremendously important for the efficient operation of industrial laboratories. The Test Laboratory Scheduling Problem (TLSP) is an extension of the well-known Resource Constrained Project Scheduling Problem (RCPSP) and captures the specific requirements of such laboratories. In addition to several new scheduling constraints, it features a grouping phase, where the jobs to be scheduled are assembled from smaller units. In this work, we introduce an innovative scheduling system that allows the efficient and flexible generation of schedules for TLSP. It features a new Constraint Programming model that covers both the grouping and the scheduling aspect, as well as a hybrid Very Large Neighborhood Search that internally uses the CP model. Our experimental results on generated and real-world benchmark instances show that good results can be obtained even compared to settings which have a good grouping already provided, including several new best known solutions for these instances. Our algorithms for TLSP have been successfully implemented in a real-world industrial test laboratory. We provide a detailed description of the deployed system as well as additional useful soft constraints supported by the solvers and general lessons learned. This includes a discussion of the choice of soft constraint weights, with an analysis on the impact and relation of different objectives to each other. Our experiments show that some soft constraints complement each other well, while others require explicit trade-offs via their relative weights.
Logic and Computation: 90% Information Systems Engineering: 10%