Download Adobe Reader

Seminar Series

Each semester the Department of Industrial Engineering hosts a seminar series to educate our students, faculty, alumni, and partners. The department invites guest presenters from prominent industrial engineering programs as well as features the research of our own faculty and students.

Seminar Series - Spring 2018

Industrial Engineering Distinguished Researcher Seminar Series
Date Time Location Speaker
March 30 1:25 - 2:25 pm Freeman Auditorium

Dr. Xin (Sophie) Liu
Assistant Professor
Dept. Mathematical Sciences

Clemson University

Title: Optimal controls for production systems with perishable inventory and time-varying demand

Abstract: We study production systems with non-stationary stochastic demand, perishable inventory, and abandonment of backorders.  In addition to inventory-related (holding, perishment) and demand-related (waiting, abandonment) costs, we consider a cost that penalizes rapid fluctuations of production rates.  We formulate a finite-time production rate planning problem to minimize the overall costs.  A crucial challenge from the non-stationary demand is how to effectively capture both the time variability and stochastic variability of the systems into the optimal decision process.  In response, we develop a novel two-stage optimal control method that takes advantage of the fluid control problem's capability of capturing the system's time variability and the diffusion of control problem's capability of capturing stochastic variability under the associated FCP solutions.  The proposed method expands the scope of existing stochastic control methods from the Brownian control problems, because the critically loaded condition can be bypassed.  Such a feature is crucial in non-stationary systems because the system inevitably experiences significant periods of overloading and underloading.

Date  Time Location Speaker
March 16 1:25 - 2:15 pm Freeman Auditorium Arie and Sabrina

Title: Optimized Appointment Scheduling in General Practitioner Practices

Abstract: In this talk, we consider the appointment scheduling in general practitioner (GP) practices. The topic becomes increasingly important since long waiting times are the most common reasons for patients' complaints. Still, an appointment scheduling policy focusing solely on short waiting times is disadvantageous for practice owners since they strive for a more productive and effective workflow which requires high utilization of GPs. To obtain a satisfactory trade-off for both patients and GPs, we develop a Mixed-Integer-Program which divides the GP's working time for chronic and regular patients with appointments as well as walk-ins. Furthermore, we introduce a concept of masks which provides templates to support decisions about the number of reserved appointments. Finally, we evaluate our methods by means of simulation.

Date  Time Location Speaker
March 2 1:25 - 2:15 pm Freeman Auditorium Dr. Jorge Sefair
Assistant Professor
School of Computing, Informatics, and Decision Systems Engineering

Arizona State University

Title: An exact model and algorithm for the multi-vehicle route selection and scheduling problem with trajectory coordination.

Abstract: We study the problem of finding optimal routes for multiple vehicles traveling in a directed network. Vehicles may have different origins and destinations, and must coordinate their routes in such a way that no two vehicles can be closer to each other than a given distance at any time. For this purpose, we not only need to determine a route, but also a schedule for each vehicle, which possibly includes waiting times in some nodes. Typical applications of this problem include the transportation of hazardous materials, autonomous vehicle routing, and ground operations subject to geographic failures (e.g., natural disasters, malicious attacks, etc.). We discuss the hardness of this problem and present an exact formulation to solve this problem. We also discuss a solution approach based on a network decomposition that relies on the sparsity of the optimal solution. We then illustrate the performance of our methods over a real network from the city of Berlin and discuss possible variants to this problem. This is a joint project with Navid Matin-Moghaddam, a PhD student at Arizona State University.

Date  Time Location Speaker
February 23 1:25 - 2:15 pm Freeman Auditorium Dr. Dotan Shvorin
Department of Industrial Engineering

Clemson University

Dr. Phillip Moschella
Clinical Associate Professor

University of South Carolina School of Medicine-Greenville

Abstract: In this seminar we will be discussing the work of Industrial engineering the in the Medical field. Dr. Shvorin and Dr. Moschella will be sharing some of the work they have done and discussing the direction and role Inustrial Engineers play in the Health Care Sector. This semiar will provide insight into what being an IE in the health care sector means, dicuss research projects in Emergency Medicine Departments, and allow students to ask questions about the work being done in health care.

Date  Time Location Speaker
February 16 1:25 - 2:15 pm Freeman Auditorium Dr. Shuchisnigdha Deb
Postdoctoral Associate
Department of Center for Advanced Vehicular Systems

Mississippi State University

Title: Pedestrian perception of fully autonomous vehicles

Abstract: With the innovation and implementation of autonomous vehicles, many questions have been raised regarding their acceptance by different road users and their compatibility with existing road and signal infrastructures. Researchers have conducted studies for interface designs based on in-vehicle road-users (e.g., drivers and passengers) as well as on out-of-vehicle road-users (e.g., pedestrians). This study aims to further the research on the interaction between pedestrians and autonomous vehicles. Since FAVs are operated by software and hardware, with no human driver required, interactions between other road-users and FAVs must be understood, and potential risks must be addressed. This is especially true for pedestrians, who often exhibit unpredictable behavior and are one of the most vulnerable road-user groups. Most of the past studies have considered the interaction between pedestrians and autonomous vehicles without paying attention to the presence and/or absence of a human inside. An additional issue with pedestrian-autonomous vehicle interaction occurs when there is a human sitting on the left side of the front-row seat, the conventional driver seat for a human-driven vehicle. The presence of a passenger in the driver seat of a self-driving vehicle will lead pedestrians to assume that the passenger is operating the vehicle manually, thus causing them to look to the passenger for cues when instead they should be looking for cues from the self-driving vehicle. Currently there is a wide range of automated driving features (e.g. autopilot system by Tesla Motors) that allow drivers to engage and disengage the automated self-driving system as desired. This ability to rapidly switch between self-driving modes of operation and manual operation could potentially further complicate pedestrian-FAV interaction by sending contradictory cues to pedestrians as to whether they should interact with the vehicle or the passenger within the vehicle. For example, the passenger may look at the pedestrian which would indicate s/he is aware of the pedestrian, but at the same time the car indicates an unsafe condition for the pedestrian to cross. The objective of this research was to investigate pedestrians’ perspective of autonomous vehicles based on the interaction effect between operator status (three levels: no operator, operator paying attention, or distracted operator) and external features (e.g., no feature, image, text message, verbal message, music). The results describe pedestrians’ behavior based on their head and foot movements along with their street crossing behavior and self-reported ratings for features when presented with these different scenarios.

Date  Time Location Speaker
February 9 1:25 - 2:15 pm Freeman Auditorium Dr. Joseph Scott
Assistant Professor
Department of Chemical & Biomolecular Engineering

Clemson University

Title: Efficient Solution of Mixed-Integer Multistage Stochastic Programs for the Integrated Design and Operation of Smart Manufacturing Systems using "Differentiable-in-Expectation" Decision Rules

Abstract: The ability to optimally adapt to dynamic and uncertain operating environments is central to the smart manufacturing paradigm. Such adaptability has tremendous potential to reduce costs and increase efficiency by exploiting real-time markets, leveraging variable renewable resources, and accommodating process contingencies. However, highly adaptive manufacturing poses critical challenges because it couples decision-making across the conventional hierarchy from design to planning, scheduling, and control. In this context, this talk considers the challenging problem of integrating long-term investment decisions, such as design and expansion planning, with mid-term operational decisions such as adaptive scheduling, real-time optimization, and supervisory control. Technically, the integrated design and operation of such systems leads to challenging multistage stochastic programs (MSPs) with mixed-integer recourse decisions (e.g., adaptive scheduling, unit commitment, etc.) and very many stages (e.g., hundreds). This talk will present recent progress towards a new approach for efficiently solving such MSPs using “differentiable-in-expectation” decision rules. Decision rules determine (suboptimal) recourse decision in each stage either as (i) fixed functions of the random variables (RVs) and system states, or (ii) via embedded optimization problems representing, e.g., a supervisory controller. Using decision rules leads to a single-stage ‘simulation-optimization’ problem with many fewer decisions. Unfortunately, this problem is potentially discontinuous because the embedded rules must make binary decisions. However, our recent work in (Hakizimana et al., JOTA, 173, 2017) provides sufficient conditions under which the expected cost can still be a smooth function of the first-stage decisions. When these conditions hold, the original MSP is reduced to a single-stage, smooth simulation-optimization problem that can be solved very efficiently using, e.g., stochastic gradient descent. This raises a key research question: Can accurate mixed-integer decision rules be constructed so as to satisfy these differentiability conditions by design, thereby yielding a efficiently solvable problems in general? We will discuss recent progress towards answering this question, as well as numerical results showing the efficient solution of challenging MSPs arising in smart chemical manufacturing and distributed power generation, which have mixed-integer recourse decisions in thousands of stages.

Date  Time Location Speaker
February 2 1:25 - 2:15 pm Freeman Auditorium Rebecca Albers and Megan Byham

Title: BM/MS Program Information Session

Abstract: In this seminar we will be discussing the BS/MS Program that is provided by the IE Department at Clemson. The BS/MS is a program in which students are able to finish their BS degree while also working towards their MS degree. The seminar is provided to give insight to the program, answer FAQ, and allow some time for students to ask any questions they may have regarding the program as well.

Date  Time Location Speaker
January 26 1:25 - 2:15 pm Freeman Auditorium Ruiwei Jiang
Assistant Professor
Department of Industrial & Operations Engineering

University of Michigan

Title: Distributionally Robust Contingency-Constrained Unit Commitment

Abstract: In this talk, we propose a distributionally robust optimization approach for the contingency-constrained unit commitment problem. In our approach, we consider a case where the true probability distribution of contingencies is ambiguous, i.e., difficult to accurately estimate. Instead of assigning a (fixed) probability estimate for each contingency scenario, we consider a set of contingency probability distributions (termed the ambiguity set) based on the N-k security criterion and moment information. Our approach considers all possible distributions in the ambiguity set, and is hence distributionally robust. Meanwhile, as this approach utilizes moment information, it can benefit from available data and become less conservative than the robust optimization approaches. We derive an equivalent reformulation and study a Benders' decomposition algorithm for solving the model. The case studies on a 6-Bus system and the IEEE 118-Bus system demonstrate that the proposed approach provides less conservative unit commitment decisions as compared with the robust optimization approach. This is a joint work with Chaoyue Zhao (Oklahoma State University), and is supported in part by the National Science Foundation via grants CMMI-1555983 and CMMI-1662774.