Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TE06

along with DP framework to optimally manage traffic along diamond interchange corridor. MIDAS proactive control is extended to autonomous vehicular traffic, which optimally schedules vehicle movements and manages platoons(by controlling leader and followers). 3 - Comprehensive and Quantitative Analysis of the Coordination Between Urban Railway and City Yong Yin, Southwest Jiaotong University, Chengdu, China National United Engineering Laboratory of Integrated and Intelligent Transportation, Chengdu, China, Jie Liu, Qiyuan Peng, Xu Yan, Anjun Li Evaluating the coordination between urban railway and the city correctly and comprehensively is of great significance for urban railway construction and city development. Based on the fractal theory, the coordination index of urban railway network and urban road network and the coordination index of urban railway station and urban traffic demand were constructed from aspects of multi radius and multi direction. Then, the comprehensive coordination index of urban railway and city was established based on fractal dimension consistency and vector similarity. The research has a certain significance in guiding urban railway planning and improving the coordination between urban railway and city. 4 - Evolvement of Public Charging Infrastructure in a Competitive and Stochastic Market Zhaomiao Guo, University of Central Florida, 6566 Tealwood Drive, Orlando, FL, United States, Julio Deride, Yueyue Fan, Yueyue Fan This paper presents a network-based multi-agent optimization model for strategic planning of charging facilities in a competitive and stochastic market. We provide a solution method based on alternating direction method of multipliers (ADMM). 5 - Transit Network Design with Congested Common Lines David Z.W. Wang, Associate Professor, Nanyang Technological University, 50 Nanyang Avenue, Singapore, 639798, Singapore This study focuses on a continuous transit network design problem with explicit consideration of congested common-lines. A tri-level programming model is presented to formulate the problem. Basically, the upper-level program optimizes the transit service frequencies to achieve the objective of both operators and transit users; the middle-level problem describes the passengers’ routing choices, which is indeed an equilibrium transit assignment problem; the lower-level program formulates the congested common-line problem. The tri-level model is reduced into an equivalent single level program to be solved.The global optimal solution of the problem is to be obtained. 6 - Inspection Based Predictive Maintenance for Railways Ayca Altay, Rutgers University, 640 Barthalomew Rd, Piscataway, NJ, 08854, United States, Pedro Cesar Lopes Gerum, Melike Baykal-Gursoy Maintenance activities are essential to preserve safety and cost-effectiveness in railways. The related literature evaluates preventive and corrective maintenance conditions. However, the maintenance activities involve a structured policy of inspections, whose outcomes shape the replacement decisions. This study provides a holistic approach by integrating the prediction of rail and geometric defects, together with the scheduling of inspection-driven maintenance activities. Results indicate a high accuracy rate in prediction and an efficient scheduling structure. n TE08 North Bldg 124A Stochastic First-order Methods Sponsored: Optimization/Nonlinear Programming Sponsored Session Chair: Farzad Yousefian, Oklahoma State Univeristy, Stillwater, OK, 74074, United States 1 - Distributed Algorithms for Stochastic Nash Games Jinlong Lei, Pennsylvania State University, 310 Leonhard Building, University Park, PA, 16803, United States, Uday Shanbhag We consider the development of distributed variable sample-size gradient- response and best-response schemes for structured nonsmooth stochastic Nash games. In each instance, we show that when batch-sizes grow at suitable rates and with sufficient number of consensus rounds, the schemes display linear rates of convergence. Additionally, we quantify the iteration complexity of each scheme.

n TE06 North Bldg 122C

Joint Session OPT/Practice Curated: Network Optimization in Applications Sponsored: Optimization/Global Optimization Sponsored Session Chair: Golshan Madraki, PhD, Clarkson University, Potsdam, NY, 13699, United States 1 - Most Closeness Central Clique Problem Farzaneh Nasirian, University of Massachusetts-Boston, 100 William T. Morrissey Blvd, Boston, MA, 02125, United States, Foad Mahdavi Pajouh This talk addresses the most closeness-central clique problem in which we are interested in detecting a most accessible clique in a graph. We use two metrics of maximum and total distance to a clique for measuring its accessibility resulting in two variants of the most closeness-central clique problem. For each of these two problems, we address the computational complexity, develop a new mixed 0-1 integer programming formulation, and propose the first combinatorial branch- and-bound algorithm. The computational performance of these exact algorithms is studied on a test-bed of real-life instances. 2 - Information Based Drone Assisted Parcel Delivery in Urban Environments Cesar N. Yahia, The University of Texas at Austin, Austin, TX, 78705, United States, Can Gokalp, Prashanth Venkatraman, Stephen D. Boyles We investigate the problem of using unmanned aerial vehicles alongside a truck for last-mile parcel delivery in an urban environment. The objective is to determine the route that the truck should traverse as well as the locations where the drone should be deployed to minimize total truck travel time. We propose real-time algorithms that exploit the travel time estimation capabilities of the drone. 3 - On the Structure of Potential Driven Networks Gerrit Slevogt, Universit t Duisburg-Essen, Ruediger Schultz, Sabrina Nitsche Potential driven networks such as water, gas and power are core utilities of today’s world. They are governed by specific (non-linear) constraints such as derivatives of the Euler equations in gas and water and Kirchhoff’s circuit laws in power networks. Especially in power networks the rise of renewable energies is driving the expansion and meshing of networks. Thus, the problem of finding operational bounds on the supported input-output nominations is getting more complex. Finding optimal controls or flows on a case by case basis is operationally feasible but unsatisfying. An analysis of the structure of such networks and arising properties can lead to a more comprehensive view of such networks. 4 - Accelerating the Scheduling Improvement Heuristics by Finding the Longest Path in the Perturbed Graph Golshan Madraki, Clarkson University, Potsdam, NY, USA Scheduling improvement heuristics iterate over trial schedules to determine a satisfactory schedule. During each iteration, a performance measure (e.g., makespan) is calculated. This research presents an efficient algorithm, Structural Perturbation Algorithm (SPA), that accelerates the calculation of makespan. This means all scheduling improvement heuristics using SPA to calculate makespan for each trial schedule will run faster. We model the manufacturing by a Directed Acyclic Graph (DAG). Schedule trials are represented by perturbed DAGs where multiple edges are added and deleted. SPA can handle multiple edge deletions/additions through a single pass which improves the time complexity in comparison with current approaches. n TE07 North Bldg 123 Joint Session OPT/Practice Curated: Network Optimization Models for Transportation Sponsored: Optimization/Network Optimization Sponsored Session Chair: Ayca Altay, Rutgers University, 110 Washington Rd, Princeton, NJ, 08540, United States 1 - Midas Proactive Traffic Control; Autonomous Intersection & Diamond Interchange Viswanath Potluri, Research Associate, Arizona State University, Tempe, AZ, 85281, United States MIDAS proactive traffic control uses forward recursion Dynamic Programming (DP) approach with efficient data structures, over a finite-time horizon that rolls forward and, then uses a backward recursion to retrieve the optimal decision sequence. MIDAS architecture uses vehicle GPS data, queue estimation models

382