2016 INFORMS Annual Meeting Program

WE65

INFORMS Nashville – 2016

WE65 Mockingbird 1- Omni Advanced Monitoring Techniques for Complex Data Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Mohammed Saeed Shafae, Virginia Polytechnic Institute and State University, TBD, Blacksburg, VA, 00000, United States, shafae1@vt.edu Co-Chair: Lee Wells, Western Michigan University, 1903 W Michigan Ave, Kalamazoo, MI, 49008, United States, lee.wells@wmich.edu 1 - Statistical Process Monitoring Of Multimode Shape Profiles Kai Wang, Hong Kong University of Science and Technology, kwangai@connect.ust.hk Traditional shape profile monitoring focuses on one mode of shapes. Little attention has been paid to multimode shape profiles, where different types of shapes appear in a sample of objects. In this work, we exploit the process monitoring of multimode shape profiles. First, we develop a two-step feature extraction approach, where different shape modes can be separated into several clusters. This enables us to build a finite Gaussian mixture model for the extracted features. In Phase II, a control chart is built for detecting shifts in the proportions and shape features of multimode shape profiles. Numerical simulations and a real example demonstrate the effectiveness of our proposed framework. 2 - A Bayesian Self-starting Control Chart For Count Data Baosheng He, University of Iowa, Iowa City, IA, United States, baosheng-he@uiowa.edu, Yong Chen In this work we propose a Bayesian framework to detect a random but sustained shift in count data, including Poisson and Binomial data. The in-control and out- of-control states are both unknown and modeled by corresponding priors, and so as the shift probability, if necessary. The decision is based on the posterior probability that the shift occurs at each time. The monitoring performance is evaluated by the average run lengths. The effectiveness of the method is demonstrated via simulation and real data. 3 - Functional Regression Based Monitoring Of Service Systems Devashish Das, Mayo Clinic, 2015, 41st Street NW, F47, Rochester, In this research, we focus on building a statistical monitoring method for service systems that experience time varying arrival rates. The goal of the proposed method is to build a functional regression model based on customer arrival and departure time data collected from an in-control system. It is then used to find discrepancy between expected departure rates and observed departure rates. Deviations from expected departures greater than a threshold are used to signal a deterioration ins quality of service. MN, 55901, United States, das.devashish@mayo.edu, Kalyan Pasupathy, Curtis B. Storlie, Mustafa Y Sir Service Robotization: Building a Collaborative research Agenda for Interactive Service Robots Sponsored: Service Science Sponsored Session Chair: Thorsten Gruber, Loughborough University, Centre for Service Management, Sir Richard Morris Building, Loughborough, LE11 3TU, United Kingdom, T.Gruber@lboro.ac.uk Co-Chair: Willy Barnett, The University of Manchester, Manchester Business School, Manchester, M13 9QS, United Kingdom, willy.barnett@postgrad.mbs.ac.uk 1 - Human-robot Interaction For Real-world Situations Julie Adams, Professor/Computer Science & Computer Engineering, Vanderbilt University, Nashville, TN, 37212, United States, julie.a.adams@vanderbilt.edu Robots (generally any semi-autonomous vehicle) are increasingly being used by individuals in their daily activities, be such activities personal or professional. However, robots have traditionally been used by highly trained personnel in highly controlled environments and settings. Real-world environments tend to be highly dynamic with large amounts of uncertainty. Further, the human may not have extensive training and cannot be a dedicated robot controller or supervisor, but must also be responsible for other activities and actions in said environment. Thus, traditional interaction mechanisms are difficult to use and place too many demands on the humans. The question is how can the human-robot interaction become more natural for the human in order to support the collective goals? The WE66 Mockingbird 2- Omni

answer is multi-dimensional. From one perspective, the human needs to easily interact with the robot and be able to develop a reliable understanding of the robot’s expected actions. Further, the robots need to easily perceive the human’s state, predict what the human will do, and develop a plan to maximize the likelihood of achieving the goal, while minimizing the demands placed on the human. These and related questions are still open research questions within human-robot interaction. 2 - Developing an Innovative Research Methods Toolkit To Explore User Perceptions Of HRI: Part 1- The Laddering +ZMET Method Thorsten Gruber, Loughborough University, Sir Richard Morris Building, Loughborough, United Kingdom, T.Gruber@lboro.ac.uk, Kathy Keeling User acceptance is a critical issue in the domain of Human-Robot Interaction (HRI). Many studies address user acceptance through methods such as needs analysis, which allow users to interact with robots and identify the pros/cons of use. Such methods are effective in uncovering superficial needs but fail to obtain deeper meanings of use. We argue that HRI could benefit from innovative research designs that help researchers obtain a deeper understanding of what users value in HRI. We present a toolkit consisting of three well-established methods adapted to HRI and present research examples. 3 - Developing an Innovative Research Methods Toolkit To Explore User Perceptions Of HRI: Part 2 – The Netnography Method Kathy Keeling, University of Manchester, Booth Street West, Manchester, United Kingdom, kathy.keeling@manchester.ac.uk User acceptance is a critical issue in the domain of Human-Robot Interaction (HRI). Many studies address user acceptance through methods such as needs analysis, which allow users to interact with robots and identify the pros/cons of use. Such methods are effective in uncovering superficial needs but fail to obtain deeper meanings of use. We argue that HRI could benefit from innovative research designs that help researchers obtain a deeper understanding of what users value in HRI. As a follow-up to the first section on Innovative Research Methods, this section will introduce the method known as Netnography. The presentation concludes with some examples previous studies. 4 - Older Consumers Value Perceptions Of Service Robots: Exploring The Intersection Of Marketing, Robotics, And Design Willy Barnett, The University of Manchester, Manchester Business School, Manchester, M13 9QS, United Kingdom, willy.barnett@postgrad.mbs.ac.uk This study explores the relationship between two major phenomena facing developed worlds today: robotics and global aging. It attempts to examine human-robot interaction through a marketing lens by exploring the nature of robot value perceptions, their relationship to robot design and user acceptance. To address this goal, a multi-method, qualitative study of a service-dominant network consisting of older adults and their care providers is conducted. Results show that robot acceptance can be better understood and communicated in terms of high level user values associated with robot use. WE67 Mockingbird 3- Omni High Dimensional Statistical Process Monitoring and Diagnosis Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Yan Jin, University of Washington, Seattle, WA, United States, yanjin@uw.edu Co-Chair: Shuai Huang, University of Washington, shuaih@uw.edu 1 - Parametric Uncertainty Propagation In Potassium Channel Model Of Mouse Ventricular Myocytes Dongping Du, Texas Tech University, dongping.du@ttu.edu Cardiac potassium (K+) channel plays an important role in cardiac electrical signaling. Mathematical models have been widely used for investigating the effects of K+ channels on cardiac functions. However, K+ channel models involve parametric uncertainties. It is critical to assess the parameter uncertainties to provide more reliable predictions. In this study, a generalized polynomial chaos expansion is used to propagate the uncertainties onto the modeled predictions of steady state activation and steady state inactivation of the K+ channel. As compared with the Monte Carlo simulations, the proposed method shows superior performance in terms of computational efficiency and accuracy.

504