Industrial-size job shop scheduling with constraint programming

Abstract

The job shop scheduling problem is one of the most studied optimization problems to this day and it becomes more and more important in the light of the fourth industrial revolution (Industry 4.0) that aims at fully automated production processes. For a long time exact methods like constraint programming had problems to solve real large-scale problem instances and methods of choice were to be found in the area of (meta-) heuristics. However, developments during the last decade improved the performance of state-of-the-art constraint solvers dramatically, to the point that they can be applied also on large-scale instances. The presented work’s main target is to elaborate the performance of state-of-the-art constraint solvers with respect to industrial-size job shop scheduling problem instances. To this end, we analyze and compare the performance of two cutting-edge constraint solvers: OR-Tools, an open-source solver developed by Google and recurrent winner of the MiniZinc Challenge, and CP Optimizer, a proprietary constraint solver from IBM targeted at industrial optimization problems. In order to reflect real-world industrial scenarios with heavy workloads like found in the semi-conductor domain, we use novel benchmarks that comprise up to one million operations to be scheduled on up to one thousand machines. The comparison is based on the best makespan (i.e. completion time) achieved and the time required to solve the problem instances. We test the solvers on single-core and quad-core configurations.