Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WA70

2 - A Metamodel-assisted Framework for Two-stage Optimization v ia Simulation Wei Xie, Yuan Yi For the discrete two-stage optimization with the unknown response obtained from simulation, we introduce a metamodel-assisted framework that can efficiently employ the simulation resource to iteratively solve for the optimal first- and second-stage decisions. At each visited first-stage decision, we develop a local metamodel to solve a set of deterministic recourse problems simultaneously. Then, we construct a global metamodel accounting for the finite sampling error from SAA and the second-stage optimality gap. Assisted by this global-local metamodel, we propose a simulation optimization approach that can efficiently guide the search for the optimal first- and second-stage decisions. 3 - Design & Analysis of a Computer Experiment for an Aerospace Conformance Study David Edwards, Virginia Commonwealth University, 1015 Floyd Avenue, P.O. Box 843083, Richmond, VA, 23284, United States Within NASA’s Air Traffic Management Technology Demonstration, Interval Management (IM) is a flight deck tool that enables pilots to achieve or maintain a precise in-trail spacing behind a target aircraft. Previous research has shown that violations of aircraft spacing requirements can occur between an IM aircraft and its surrounding non-IM aircraft when it is following a target on a separate route. This talk focuses on the experimental design and analysis of a computer experiment which models the airspace configuration of interest in order to determine airspace/aircraft conditions leading to spacing violations during IM operation. 4 - Varying Coefficient Gaussian Processes for Computer Models with Quantitative and Qualitative Inputs Computer experiments with both qualitative and quantitative factors attracts wide attentions. Analysis of such experiments is not yet completely resolved. In this work, we propose a varying coefficient Gaussian process to model computer experiments with qualitative and quantitative factors. The proposed method considers the coefficient associated with the qualitative factor to be a varying coefficient of the quantitative factors. It embraces a flexible structure of incorporating qualitative factors in modeling the complex systems of computer experiments. The merits of the proposed method are illustrated by both numerical examples and real-data applications. 5 - Modeling and Forecast of Noisy Nonlinear Dynamics Youngdeok Hwang, Sungkyunkwan University, Department of Statistics, 25-2 Sungkyunkwanro, Seoul, 03063, Korea, Republic of, Kyongmin Yeo, EunKyung Lee Data-driven modeling of a complex physical process is of current interest due to its direct relevance to manufacturing. Although domain knowledge on the underlying physical process has been playing a key role in understanding manufacturing processes, it is often impractical or impossible to build physics models from the first principles. Here, we propose a Recurrent Neural Network (RNN) based model for the nonlinear system identification and forecast of a stochastic process with an underlying physical process. n WA70 West Bldg 106B Fault Diagnosis and Tolerant Control for System Reliability Improvement Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Yimin Zhou, Shenzhen Insitutes of Advanced Technology, Chinese Academy, 1068 Xueyuan Avenue, Shenzhen University Town, Shenzhen, Guangdong, 518055, China 1 - The Prospect of Smart Cars: The Combination of Roboticized Technology and Intelligent Vehicles Yimin Zhou, Chinese Academy, 1068 Xueyuan Avenue, Shenzhen University Town, Shenzhen, Guangdong, 518055, China, Qin Lv, Qingtian Wu In this paper, the development of the roboticized technology has been described. Since the first robot has been designed, robots have been gone through three generations from the individual robots to intelligent robots from 60’s. Potential development directions of the robotic technology is discussed. Besides, the problems encountered during the research are also analyzed, including stability control, reliability control and fault tolerant control. Furthermore, the roboticized technologies are applied in the intelligent electric vehicles to improve the driving safety in human-road-vehicle environment. The driving strategy, vehicle monitoring and human-machine interaction are reviewed in details. Xinwei Deng, Department of Statistics, Virginia Tech, 406 Hutcheson Hall, Blacksburg, VA, 24061, United States

2 - Action Detection and Abnormal Event Detection for Surveillance Monitoring Under Mobile Vehicle or Moving Camera Qingtian Wu, Shenzhen Institutes of Advanced Technology, Beijing, China In outdoor environments, many cameras have been fixed in main roads, entrances to monitor abnormal event for better security. However, in some areas such as shade or turning corner, coverage monitoring cannot be achieved due to the limited monitoring distance and angle constraints of fixed camera. In order to monitor the blind spots under the fixed camera and track the incident centroid, a rapid detection of abnormal behavior system based on the mobile robot is proposed to detect abnormal events. The system can be embedded in a patrol robot which is used in the community environment to shoot the real-time scene information by vehicle-mounted camera, then to analyze of the monitoring area through intelligent visual detection algorithm. If abnormal event occurs, alarm will be triggered to alarm the guard. 3 - Actuator Fault Tolerant Control Algorithm with Application to a Quadcopters Yimin Zhou, Shenzhen Insitutes of Advanced Technology, 1068 This paper investigates a rotor failure on the quadcopter. The goal is to detect the fault and to design a controller to make the quadcopter to perform the hover and trajectory task. The task of the controller is to make the quadcopter to perform a stable spin around its vertical axis when a rotor fault occurs. The UAV will work as a tri-copter to perform the desired tasks. The proposed controller is developed based on the feedback linearization approach and compared with PID controllers. We consider only smooth trajectory by avoiding the sharp turn’s to maintain the minimal acceleration. Simulation results show the efficacy of the proposed controller towards the UAV reliability and safety improvement. Xueyuan Avenue, Shenzhen University Town, Shenzhen, Guangdong, 518055, China, Kranthi Kumar Deveerasetty Joint Session ICS/Practice Curated: Dynamic Stochastic Optimization in Urban Transportation Sponsored: Computing Sponsored Session Chair: Qie He, University of Minnesota, Minneapolis, MN, 55455, United States 1 - Travel Time Estimation in the Age of Big Data Sebastien Martin, Massachusetts Institute of Technology, Cambridge, MA, 02139, United States, Patrick Jaillet, Dimitris Bertsimas, Arthur J. Delarue While travel time estimation has become increasingly mainstream and accurate, it is hard to build application-specific estimations that are reliable and compare with the state of the art without access to clean travel time datasets and detailed routing network information. We develop a method that tractably exploits any amount of origin-destination data and only needs a simplified routing network to build accurate estimations. Using data from the Manhattan taxis, we show that our algorithm can provide insights about urban traffic patterns on different scales and accurate travel time estimations throughout the network. Boston also used it to better route school buses and save millions of dollars. 2 - Absenteeism Prediction and Extra-board Driver Scheduling for Bus Transit Operations Xiaochen Zhang, University of Minnesota, Minneapolis, MN, 55454, United States, Qie He We develop a data-driven analytics tool to assist daily scheduling of extra-board drivers in bus transit operations. Extra-board and overtime drivers are used to cover open jobs due to the absences of regular drivers. We build a hierarchical regression model to forecast the daily absences. With the prediction, we develop a two-stage stochastic programming model to determine the daily optimal assignment of extra-board drivers and use of overtime, which also makes recommendations on the daily extra-board size. Our model’s recommended assignment based on historical data reduces the expected total operating cost significantly while the probability of losing service remains low. 3 - An Integrated Dynamic Ridesharing Dispatch and Idle Vehicle Repositioning Strategy on a Bimodal Transport Network Saeid Rasulkhani, Tai-Yu Ma, Joseph Y.J. Chow, Sylvain Klein In bimodal ridesharing, a private on-demand mobility service operator offers to drop off a passenger at a transit station, where the passenger uses the transit network to get to another transit station, and the service operator guarantees picking up the passenger (not necessarily with the same vehicle) to drop them off at the final destination. We consider dynamic bimodal ridesharing problems where real-time information is available to anticipate future demand. A new non- myopic vehicle dispatching and routing policy based on queueing-theoretical approach is proposed and integrated with a non-myopic idle vehicle repositioning strategy from the literature to solve the problem. n WA71 West Bldg 106C

436