Constraint Programming Approaches to Integrated Scheduling and Transportation Problems

点击翻译按钮获取中文摘要

It is common in the service industry for managers in different departments, using different tools, to solve planning problems in silos. Each of them concentrates on one piece of the puzzle, often ignoring the global picture. This behaviour is seen in large retail companies where human resources, inventory movement, and transportation are planned separately; healthcare organizations where the schedules for physicians, nurses, rooms, and patients are assembled by different people; and transportation networks where the opportunity for backhauling during operations is often lost because of myopic tactical planning. Moreover, new mobile technology is changing the business paradigm of many of these organizations. Multi-channel retailing is becoming increasingly important as more and more customers order online. Data collected in real time will soon allow many different organizations (such as public transit firms) to react instantly to changing trends and to adapt their offers accordingly. In healthcare, personalized medicine is knocking at the door.The challenges that lie ahead are thus numerous: i) Given the tremendous amounts of data that will be available, appropriate system performance measures will need to be developed and handled by optimization engines; ii) decision problems will need to be more integrated along the time line, e.g., strategic and tactical plans will need to take into account operational constraints and foresee real-time adjustments; iii) decision models will need to account for uncertainty because the available data can be incomplete or inaccurate. To tackle these challenges, the proposed research program will develop new methodology, built around rich combinatorial structures and their use within operations research methods, to solve large, integrated, and stochastic decision problems. I will target personnel scheduling, the planning and scheduling of transportation, and healthcare logistics.We have already developed methodology for complex and integrated personnel scheduling problems. These approaches will be extended by the development of new techniques that capture more complex contexts and thus allow us to integrate different decision problems. Finally, the rich internal structures will be able to capture the combined probability distribution of a series of events, and thus we will be able to use these techniques within a stochastic programming approach.We will base our methodological development on constraint programming, which we have found to be useful for many mixed integer nonlinear problems. I want to investigate this approach further to take advantage of its strong inference mechanism and to develop “learning during search” methods for difficult nonlinear problems, such as those where the objective function is obtained through data mining. The novelty of this proposal thus lies in the use of rich combinatorial structures in flexible constraint programming approaches. These new techniques will allow us to integrate different optimization problems that are part of the same global decision process, while taking into account uncertainty and real-time decision-making.