Stochastic optimization of network design and transportation problems

运输和物流系统在现代经济和社会的运作中发挥着至关重要的作用。不断增加的人员和货物数量需要在全球范围内有效地运输，以适当地支持经济和社会活动，同时最大限度地减少这些运动造成的负面足迹，这给管理这些系统的决策者带来了复杂的规划问题。由于各种不确定因素的存在，这些不确定因素随机而急剧地改变了进行这种规划的环境，因此大大增加了这些问题的复杂性。因此，在明确考虑所有相关不确定性来源的同时，解决运输和物流问题已成为两家公司需要解决的重要问题，两家公司都在不断努力保持市场竞争力，而政府则肩负着减少与运输相关的负面影响的任务。在所有情况下，解决这些问题构成了本建议将应付的重大挑战。******两种一般类型的优化问题，这是交通和物流系统的规划和运作的核心，将被研究：动态和随机车辆路线问题和随机网络设计问题。我的短期目标有两个：1)通过结合机器学习技术，并行计算和数学规划策略，创建新的随机优化方法，以有效地解决所研究的问题；2)设计分析工具，能够系统地评估各种不确定性来源对决策过程的总体影响。******因此，在本提案中进行的各种研究将以新的和有效的决策支持工具的形式为一系列重要的优化问题带来显著的好处，这些问题是在规划和运营运输和物流系统的背景下出现的不确定性。从科学的角度来看，这一提议将通过连接机器学习和随机优化领域来突破界限。最终，该研究计划将产生一个新的方法框架来评估和研究不确定性对优化问题的影响，并使这些问题能够有效地解决实际大小的实例（即，例如它们出现在实际设置中）。

Transportation and logistics systems play a paramount role in how modern economies and societies function. The ever-increasing volumes of people and freight that need to be efficiently transported worldwide to properly support economic and social activities while, at the same time, minimize the negative footprint caused by these movements, pose complex planning problems for the decision makers managing these systems. The complexity of such problems is considerably increased by the presence of various sources of uncertainty that randomly and drastically change the contexts in which such planning is conducted. Therefore, solving transportation and logistics problems while explicitly considering all relevant sources of uncertainty has become an important issue to address by both companies, who are continuously striving to remain competitive on the markets, and governments, who are tasked with reducing the negative effects associated with transportation. In all cases, solving such problems poses significant challenges that the present proposal will meet.******Two general types of optimization problems, which are central to the planning and operation of transportation and logistics systems, will be studied: dynamic and stochastic vehicle routing problems and stochastic network design problems. My short-term objectives are twofold: 1) to create novel stochastic optimization methods, by combining machine learning techniques, parallel computation and mathematical programming strategies, to efficiently solve the studied problems and 2) design analytical tools capable of systematically assessing the overall effects of the various sources of uncertainty on the decision-making processes.******Therefore, the various studies that will be performed in the present proposal will bring significant benefits in the form of new and efficient decision support tools for a series of important optimization problems under uncertainty that arise in the context of planning and operating transportation and logistics systems. From a scientific perspective, this proposal will push boundaries by bridging the fields of machine learning and stochastic optimization. Ultimately, the research program will produce a novel methodological framework to assess and study the impacts of uncertainty on optimization problems and enable such problems to be efficiently solved for realistically sized instances (i.e., such as they appear in practical settings).