Collaborative Research: Inference, Analysis and Assessment in Simulation Optimization

该基金提出了模拟优化（SO）问题的三个主要工作领域，这些问题是优化问题，其中所涉及的目标函数和约束只能通过随机模拟来观察。首先，许多SO问题具有凹凸性或单模性等结构，如果检测到这些结构，可以改进一个SO问题。S对问题本身的理解，并在选择求解算法时加以利用。将发展数值方法来探测这种结构。其次，将开发性能度量和有效计算它们的方法，以便在测试问题上对SO算法的性能进行理论上合理的比较。第三，建立SO问题测试平台。如果获得奖励，数字检测问题结构的能力将大大提高对问题结构的理解。S问题公式，并允许更多地使用利用结构的专门算法。这还可能导致用户按照这些结构来表述问题，结果可能会产生许多新的SO问题子类。测试平台以及适当的性能度量应该有助于鼓励算法的比较和开发。到那时，我们可能能够解决比今天更大的SO问题。研究结果将在应急服务、运输物流、供应链管理、收益管理等领域得到应用，并可能在许多其他领域得到应用。

This grant proposes three main areas of work on simulation optimization (SO) problems, which are optimization problems where the objective function and constraints involved can only be observed through a stochastic simulation. First, many SO problems possess structure such as convexity or unimodality that, if detected, can improve one?s understanding of the problem itself, and be exploited in selecting solution algorithms. Numerical methods will be developed to detect such structure. Second, performance measures, and methods for efficiently computing them, will be developed to enable theoretically sound comparisons of the performance of SO algorithms on test problems. Third, a testbed of SO problems will be developed.If awarded, the ability to numerically detect problem structure will greatly improve understanding of one?s problem formulations, and allow greater use of specialized algorithms that exploit structure. This could also lead to users formulating problems to adhere to those structures, with the result that many new subclasses of SO problems might be created. The testbed, along with appropriate performance measures, should help to encourage algorithm comparisons and development. We might then be able to tackle far larger SO problems than is possible today. The results of the research will find application in areas such as emergency services, transportation logistics, supply chain management, revenue management and potentially many other fields.