Seminar: ISE graduate student colloquiums
Title: Incentives and Cooperation in Game Theory Models
Presenter: Ben Chaiken
Committee: Marc Posner (advisor), Ramteen Sioshansi, Chen Chen
Abstract: Incentives and cooperation are inherent aspects of economic interactions, we consider these from two different perspectives. In market design problems, care must be taken so that policies and regulations do not inadvertently incentivize behavior that results in inefficient outcomes. We consider a proposed electricity market regulation and utilize standard game theoretic methods to show that it does create such negative incentives, undermining the regulation’s stated purpose. The methods employed in the previous work do not place any emphasis on the cooperative aspects of the economy as a social institution. In other work, we propose a new type of game theory model to study these aspects formally, with the aim of investigating connections between normative concepts of value and solutions in the model.
In a different direction, we also make some progress understanding the expected comparison between two different approaches to optimally picking orders in a warehouse. Under each approach we also address questions of optimizing physical parameters to achieve better solutions.
Title: Large Scale Transportation Network Optimization and Travel Time Prediction
Presenter: Yibo Dang
Committee: Theodore T. Allen (advisor), Marc Posner, Guzin Bayraksan
Abstract: Transportation network optimization problems continues to receive significant attention by researchers in operations research and supply chain management. Such problems consider various strategies and steps to optimize their network operational costs and efficiencies. Due to the limitations of computing power and uncertainties, many traditional methods have proved to be less practical in competitive business operations. It is critical for a decision-maker to resolve the contradictions between limited resources and the large-scale problems. This has inspired our works to help OR practitioners in transportation industry make better decisions. We focus on two types of problems in transportation: (1) The combinatorial transportation route planning problem. We are interested in solving a heterogeneous vehicle routing problem with time windows, layovers, and make-buy or “network mode” (hire third parties to deliver packages?) decisions on routes; (2) We predict the expected route travel time and evaluate its reliability by extending an optimal sparse decision tree (OSDT) algorithm on time-dependent features. The contributions of our works are: a) theoretical bounds generated for our integer programming problem, b) extensions of the interior point stabilized column generation method to our problem, c) adaption and extension of OSDT algorithm to a large scale problem.
Title: Development of an Image-Based Computational Lumbar Spine Modeling Database
Presenter: Gregory Knapik
Committee: William Marras (advisor), Eric Bourekas, Ehud Mendel
Abstract: For more than 30 years, computational spine models have been employed as research tools in an effort to better understand the spine and the causes of various low back disorders. Unfortunately, even with a long history of use and development, the vast majority of spine models in the literature suffer from a number of significant limitations that inhibit their utility for research. In the most common approach, models are developed to match cadaveric studies and not actual patients. Subsequently, these models are often purely static in nature, do not include the trunk musculature, use simplified geometry, and often reuse the same cadaveric specimen in numerous studies over a span of many years. In order to address these voids, this research seeks to develop a large normative database of advanced, highly-detailed computational spine models developed from high-resolution CT data. These dynamic models will include the impact of EMG-derived muscle forces, subject-specific motion data, and kinetics. This comprehensive database of more accurate spine models is expected to provide greater utility than existing models, serve as a basis for future spinal disorder causality research, and improve the evaluation of different treatment modalities.