Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MC76

5 - Reinforcement Learning Applications for Managing Public Sector “Use-or-Lose” Budgets Erich Morman, Naval Postgraduate School, Monterey, CA, United States When it comes to paying for contractor work, the private and public sectors operate under different environments. Regulatory constraints at the Department of Defense (DoD) as well as predominately throughout the public sector complicate the decision of how best to pay a contractor. In the public sector, it is often required to provide full funding for a project upfront. Furthermore, the funding provided has an expiration point whereby money that is not spent within a certain timeframe is taken away by a higher treasury or comptroller office. In this study, we look to use reinforcement learning algorithms to design efficient sequential payment strategies for contracted projects that can limit the amount of funding lost by the public sector customer. n MC76 West Bldg 212C MIF Early Career Award Sponsored: Minority Issues Sponsored Session Chair: Julie Simmons Ivy, North Carolina State University, Raleigh, NC, 27695-7906, United States 1- Opportunities and Challenges in Patient Service Personalization Eduardo Perez, Texas State University, Roy F. Mitte Complex, 749 N. Comanche St., San Marcos, TX, 78666, United States Patient service personalization provides new opportunities for health care systems in their pursuit of better patient outcomes and commitment to quality and safety. Much like the recent expansion of product customization, service personalization has been expanding lately due to a variety of factors including technological advances that allow for better service delivery and communication. Much of the existing research has viewed service variability as something negative that must be controlled. However, customer variability in service needs provides an opportunity to deliver more value for patients through personalization of services. This talk will examine patient service personalization and the design of systems for service customization. Starting with a review of current work, the talk will develop a framework for patient service personalization and service design, focusing on the concept of patient variability. Then the talk will include a description on how to use this concept to extract greater value from the transaction between the patient and the health care system. Based on the framework, the talk will then identify important directions for future research from both practice and academic viewpoints. Joint Session PSOR/Practice Curated: Transportation Issues in Smart Cities Sponsored: Public Sector OR Sponsored Session Chair: Leila Hajibabai, PhD, State University of New York, Stony Brook, NY, 11794, United States 1 - Integrated Signal Timing and Traffic Metering Optimization in Connected Urban Transportation Networks Ali Hajbabaie, Washington State University, Raleigh, WA, 99164- 2910, United States, Rasool Mohebifard, S.M.A. Bin Al Islam In this paper, we proposed a distributed mathematical optimization program that dynamically optimizes the traffic signal indications at intersections and at the same time finds the optimal number of vehicles that should enter the transportation network from its boundary gates to maximize the overall network performance. The solution technique has a model distributed predictive control structure that uses the location information of connected vehicles and vehicle counts from loop detectors to estimate the system state to optimize the decision variables. The results show that the proposed algorithm outperforms several benchmark solutions and increases the network throughput by 41.8% to 43.2%. n MC77 West Bldg 213A

2 - A Consensus-based Trajectory Control Logic for Connected and Autonomous Vehicles in a Signal-Free Intersection Leila Hajibabai, State University of New York at Stony Brook, Department of Civil Engineering, 2433 Computer Science, Stony Brook, NY, 11794, United States, Amir Mirheli, Mehrdad Tajalli, Ali Hajibabai This paper presents a distributed cooperative control logic to plan conflict-free trajectories for connected and autonomous vehicles (CAVs) in signal-free intersections. The problem is formulated into CAV-level mixed-integer non-linear programs (MINLPs) that minimize each vehicle’s travel time and avoid near-crash conditions. To push CAV-level solutions towards global optimality, we develop a coordination scheme that shares vehicle states on location and speed over a prediction period and incorporates such information in each CAV’s respective MINLP. 3 - Scheduling of Heterogeneous Connected Automated Vehicles at a General Conflict Area Xiaopeng Li, University of South Florida, 4202 E. Fowler Avenue, ENG 207, Tampa, FL, 33620, United States, Saeid Soleimaniamiri A mixed integer programming (MIP) model is proposed to solve the joint optimization problem that simultaneously determine scheduling and trajectories of connected autonomous vehicles at a multi-conflict point considering heterogeneous vehicle headways and values of time. A novel dynamic programming based solution approach is proposed to obtain a near optimal solution. A set of numerical examples show that the model solutions can significantly increase the capacity of the conflict point and reduce both travel time and fuel consumption. 4 - Community-engaged Operations Research as a Tool to Support Diversity, Equity and Inclusion in the Profession OR/MS/analytics faces two important questions regarding diversity, equity and inclusion. First, what has the profession done to address important diversity- related social problems? Second, what can the profession do to meet diversity, equity and inclusion goals for INFORMS? In this talk I explain the unique role that community operational research and community-based operations research can play in enabling our profession to become more welcoming to members of diverse backgrounds, and to solving important social problems for which diversity is a critical component. n MC78 West Bldg 213B Radiation Treatment Planning Under Uncertainty Sponsored: Public Sector OR Sponsored Session Chair: Gino J. Lim, University of Houston, TX, 77204, United States Co-Chair: Azin Khabazian, University of Houston 1 - Understanding Impacts of Radiobiological Parameters in Adaptive Radiation Treatment Planning under Uncertainty Azin Khabazian, University of Houston, Houston, TX, 77057, United States, Gino J. Lim To effectively treat a cancer patient with radiotherapy, an effective treatment strategy must be in place that considers dose delivery history and the patients’ on- treatment biological changes. In this study, we seek to understand the importance of considering tumor shrinkage and proliferation during radiation treatment and how this affects the optimal prescribed dose in each fraction. We propose a stochastic sequential optimization structure under setup uncertainty of dose delivery that optimizes the dose in various fractions of an adaptive radiation therapy treatment plan by comparing the damage in tumor cells against the damage to the normal tissues volumetrically. 2 - New DVH Formulation and First-order Interior-point Method for Inverse Planning Hongcheng Liu, University of Florida, Gainesville, FL, 32611, United States In radiotherapy treatment inverse planning, a commonly used quality measure is the dose-volume histograms (DVH). The exact mathematical formulation of DVHs involves binary variables and is therefore intrinsically nonconvex. In contrast, conventional convex surrogates for inverse planning entails non-trivial differences from an exact formulation and, thus, provide limited control over the DVHs. This work presents a new formulation that combines a novel folded concave penalty-based constraint and a new kurtosis-based criterion. Tailored to the new formulation is a first-order interior-point method. Outperformance in terms of doses-at-volume was achieved through the proposed scheme. Michael P. Johnson, University of Massachusetts Boston, Department of Public Policy & Public Aff, 100 Morrissey Boulevard, Boston, MA, 02125-3393, United States

212