Hybrid constraint generation approaches for industrial scheduling and logistics

货物运输（即物流）和制造过程的调度构成了加拿大经济的两个关键要素。在实践中，在研究文献中，运筹学技术，如混合整数规划、约束规划和启发式，被应用于这类问题。在过去的15年中，PI对复杂组合问题（包括物流和调度）的混合约束生成方法的研究和发展做出了贡献。这种技术依赖于问题分解和子问题之间的约束交换，以确保找到最优解或解质量保证。文献中的证据表明，与传统方法相比，这些技术通常可以实现一到两个数量级的问题解决能力提升。该项目将进一步开发混合约束生成技术，以解决目前由工业合作伙伴Visual Thinking International Ltd通过传统方法解决的四个物流和调度问题：大湖库存路线，电镀线排序，线圈装饰优化和批量优化。这些是与许多制造和物流公司相关的核心问题的现实世界版本，并且已经在文献中进行了研究，但尚未开发约束生成技术。该项目将(1)将PI在学术文献中参与开发的技术转让给一家加拿大软件公司，(2)通过来自工业合作伙伴的数据解决工业规模问题的挑战，进一步开发约束生成方法。

The transportation of goods (i.e., logistics) and the scheduling of manufacturing processes form two key elements in the Canadian economy. In practice, and in the research literature, Operations Research techniques, such as mixed integer programming, constraint programming, and heuristics, are applied to such problems. Over the past 15 years, the PI has contributed to the research and development of hybrid constraint generation approaches to hard combinatorial problems, including logistics and scheduling. Such techniques rely on a problem decomposition and the exchange of constraints amongst the sub-problems to ensure that optimal solutions or solution quality guarantees can be found. Evidence in the literature points to such techniques often achieving one or two orders-of-magnitude increase in problem solving power compared to traditional approaches.This project will further develop hybrid constraint generation techniques for four logistics and scheduling problems that are currently solved by traditional approaches by the industrial partner, Visual Thinking International Ltd: Great Lakes Inventory Routing, Sequencing for Plating Lines, Coil Trim Optimization, and Batch Optimization. These are real-world versions of core problems that are relevant to many manufacturing and logistics companies and have been studied in the literature, but for which no constraint generation techniques have been developed.The project will (1) transfer the technology that the PI has been part of developing in the academic literature to a Canadian software company and (2) further develop the constraint generation approach through the challenge of addressing the industrial-scale problems via data from the industrial partner.