MLCP: A Framework Integrating with Machine Learning and Optimization for Planning and Scheduling in Manufacturing and Services

As manufacturing and service operations have been increasingly becoming complicated, planning and scheduling which directly relate to manufacturing cost, production quality, level of services are becoming more challenging and promising. To date, a variety of planning and scheduling methods have been developed to support decision making in planning and scheduling. In most manufacturing and service operations, however, planning and scheduling are still heavily relying on high-skilled planners. According to our experiences, the reason is that, real planning and scheduling problems are partly fuzzy and dynamically changing, which make them impossible to be concretely defined by mathematical models and to be solved by conventional methods. To automating planning and scheduling, in our opinion, effectively taking advantage of high-skilled planners’ know-hows plays an important role. To this end, we develop a framework, MLCP, which integrates machine learning and optimization for real planning and scheduling problems. Since the machine learning module and optimization module in the framework are independently operable but working cooperatively, MLCP can be an useful tool for constructing large-scaled system of systems with complex planning and scheduling tasks. Through a case study, we show that, implicit know-hows of high-skilled planners such as preference for schedules can be extracted from historical schedules and imported into new schedules by employing the framework.