An improved hybrid MILP/CP algorithm framework for the job-shop scheduling

This paper studies an existing hybrid MILP/CP algorithm framework for a kind of job-shop scheduling problem, which can obtain the optimum solution but often wastes too much computing time. Based on observations on a series of cuts dynamically generated in the computing process, we introduce the concept of critical job assignment and critical cut and then propose two strategies to improve the framework. The strategies can generate many critical cuts instead of a commonplace cut for each infeasibility and thus speed up the algorithm. The experiments show that the proposed strategies are effective and efficient, which can actually extend application of the framework to more complex problems.