52 ISE Magazine | www.iise.org/ISEmagazine
Inside IISE Journals
research
Tackling the next level of uncertainty
in medical decision-making
In the words of Sir William Osler, the Father of Modern
Medicine: “Medicine is a science of uncertainty and an art
of probability.” Medical decision-making is a process that
involves considering uncertainty in the consequences and
trade-offs of decisions. For instance, when managing a pa-
tients cardiovascular disease, doctors must consider uncer-
tainty about whether the patient will have a heart attack or
stroke in the future. While some medications may reduce
the risk of a heart attack and a stroke, these medications cost
money and can have side effects.
Fortunately, mathematical models like the Markov deci-
sions process (MDP) can be used to synthesize all of this
information to guide treatment decisions that are made se-
quentially over time. Given an estimate of the risk (i.e.,
probability) that the patient will have a cardiovascular
event, MDPs can weigh the beneﬁts and harms of many
possible treatments to inform a strategy of whether and
when to treat that makes the most sense for the patient.
But what should a modeler do when there are multiple,
conicting estimates of a patient’s risk? In the case of car-
diovascular disease, there are multiple, well-established risk
calculators that use a patients information to estimate their
risk of a cardiovascular event, but different risk calculators
can give conicting estimates for the same patient prole.
MDPs that rely on these conicting estimates in turn give
different treatment recommendations, leading to ambiguity
in how the doctor should proceed to manage the patients
condition.
In their article “Multi-model Markov Decision Processes,
Georgia Tech assistant professor Lauren N. Steimle, University
of Michigan professor Brian Denton and University of Mich-
igan-Dearborn assistant professor David Kaufman propose a
new approach to mitigate this ambiguity. Their approach ﬁnds
treatment strategies that work well across multiple plausible
models of the probabilities while maintaining a similar level of
complexity of strategies that are based on one model.
Using a case study of car-
diovascular disease, the au-
thors show that their approach
recommends treatment strate-
gies with lower expected re-
gret compared to those based
on a single model alone, as
is typically done in practice.
The study suggests that in-
corporating multiple plausible
models into solution methods
This month we highlight two articles in IISE Transactions. The ﬁrst article contemplates what a medical decision-
maker should do when there are multiple, conﬂicting estimates of a patient’s risk. The authors propose a multimodel
Markov decision process method to mitigate the ambiguity in medical treatments that could be caused by the
conﬂicting estimates. The authors show a positive impact of the new approach, in terms of lower expected regret,
in a case study of cardiovascular disease treatment. The second article looks into the issue of how to efciently
operate and coordinate thousands of overhead hoist transporters in a semiconductor fabrication. For this, the
authors propose a cooperative zone-based rebalancing algorithm, which is to zone the semiconductor fabrication
and then derive a decentralized rebalancing policy for each zone. They demonstrate a remarkably reduced retrieval
time when the new research is applied to a semiconductor fabrication facility at SYNUS Tech, South Korea. These
articles will appear in the October 2021 issue of IISE Transactions (Volume 53, No. 10).
Lauren N. Steimle
Brian Denton David Kaufman
September 2021 | ISE Magazine 53
may be a promising way to mitigate the effects of ambiguity
in medical decision-making and sequential decision-making
In short, Steimle, Kaufman, and Denton propose a new
take on Markov Decision Processes that that pivot conict-
ing models from a dilemma to a feature, with applications to
medicine and beyond.
CONTACT: Lauren Steimle; steimle@gatech.edu; (404) 894-4659; Geor-
gia Tech H. Milton Stewart School of Industrial and Systems Engineering,
755 Ferst Drive, NW, Atlanta, GA 30332.
Operating multiple robot taxis to transport
wafers in a semiconductor fabrication
Large-scale complex engineering systems are often composed
of subsystems interacting with each other. Examples include
wind turbines in a farm, robots or machines in a factory and
energy systems in a smart grid. As the system becomes more
complex, it becomes challenging to control such subsystems
cooperatively to achieve system-level performance.
Multi-agent reinforcement learning, a multiagent exten-
sion of reinforcement learning, is an effective learning-based
method that can learn the cooperative and decentralized con-
trol policy controlling subsystems to achieve system-level per-
formance.
In a semiconductor fabrication (FAB), thousands of over-
head hoist transporters automatically transport a bundle of
wafers from one machine to another while moving along a
The authors of a paper on cooperative zone-based rebalancing
Manufacturing Intelligence Innovation Center at KAIST.
Pictured from left are professor Jinkyoo Park and doctoral
student Kyuree Ahn.
Call for special issue submissions
Author submissions are being sought for two upcoming special
issues of IISE Transactions.
Analytical methods for detecting, disrupt-
ing and dismantling illicit operations
This special issue will highlight analytical approaches that can
help detect, disrupt and ultimately dismantle illicit operations.
Its goal is to showcase the role of analytical methods in the ﬁght
against illicit operations by bringing together those within the
industrial and systems engineering community and domain
experts in distinct areas of ﬁghting illicit operations. Of special
interest are papers that integrate domain expertise and/or
stakeholder engagement in their model formulation, analysis
and validation.
The guest editors for the issue are Thomas Sharkey of
Clemson University, Renata Konrad of Worcester Polytechnic
Institute, Burcu Keskin of the University of Alabama and Maria
Mayorga of North Carolina State University.
The deadline for abstract submissions is Oct. 31, 2021, with
provide a submission, visit think.taylorandfrancis.com/special_
issues/analytical-methods-illicit-operations.
Modeling and optimization of
supply chain resilience to pandemics
and long-term crises
This issue aims to attract novel research dealing with supply
chain resilience modeling and optimization in context of long-
term crises motivated by the COVID-19 pandemic. The authors
seek methodically rigor and practically relevant papers dealing
with the settings when recovery/adaptation should be planned
and deployed in the presence of disruptions and bouncing
back to “old normal” is impossible or difﬁcult on a short-term
scale, and the only way to survive is to adapt. They encourage
submissions that explicitly account for the context of epistemic
and deep uncertainty (i.e., unknown-unknown) and supply chain
viability.
The special issue editors are David Coit and Weiwei Chen,
both of Rutgers University; Dmitry Ivanov of the Berlin School
of Economics and Law; and Nezih Altay of DePaul University.
and provide a submission, visit think.taylorandfrancis.com/
special_issues/supply-chain-resilience-pandemics.
54 ISE Magazine | www.iise.org/ISEmagazine
one-way track installed under a ceiling. The operation of the
transporters in a FAB is similar to the operation of robot taxis
in a city; a taxi should be optimally allocated to a customer
while considering the trafﬁc conditions of the road network
and the geographical distributions of taxis and customers to
maximize the service rate. As the semiconductor FAB become
larger and, thus, more transporters need to be operated, it is
imperative to operate transporters in a FAB intelligently to
increase the productivity of automatic material handling.
In the work “Cooperative Zone-based Rebalancing of Idle
Overhead Hoist Transportations Using Multi-Agent Rein-
forcement Learning with Graph Representation Learning,
doctoral student Kyuree Ahn and professor Jinkyoo Park from
the Korea Advanced Institute of Science and Technology
(KAIST) propose a cooperative zone-based rebalancing algo-
rithm to allocate idle overhead hoist transporters in a semicon-
ductor FAB. They discretize the FAB into a number of zones
and derive a decentralized rebalancing policy for each zone
by applying multiagent reinforcement learning with graph-
centric state representation. The graph representation module
effectively extracts and processes the geographical distribution
of transporters (taxis) and the waiting wafers (customers), and
uses them to determine the optimum rebalancing action.
The authors then demonstrated that the proposed meth-
od can signiﬁcantly reduce the average retrieval time while
reducing the transporter utilization ratio. In addition, they
show that the rebalancing policy trained with various FAB
conditions can be used under unseen dynamic scenarios
without further training, hence validating the transferabil-
ity of the proposed method. This research was funded by
SYNUS Tech.
CONTACT: Jinkyoo Park; Jinkyoo.park@kaist.ac.kr; +82 (42) 350-3133;
Department of Industrial & Systems Engineering, Korean Advanced Insti-
tute of Science and Technology (KAIST), Daejeon, South Korea
Yu Ding is the Mike and Sugar Barnes Professor of Industrial
and Systems Engineering at Texas A&M University and Associ-
ate Director for Research Engagement at the Texas A&M Institute
of Data Science. He is editor-in-chief of IISE Transactions and
a fellow of IISE.
research
IISE Transactions (link.iise.org/iisetransactions) is IISE’s ﬂagship
research journal and is published monthly. It aims to foster exchange
among researchers and practitioners in the industrial engineering
community by publishing papers that are grounded in science and
mathematics and motivated by engineering applications.
To subscribe, call (800) 494-0460.
The IISE Foundation and the IISE Scholarship Fund
offer you the chance to donate and help build
future generations of ISEs.
From the prestigious Frank & Lillian Gilbreth Award
to the \$65,000-plus in scholarships bestowed in
2021, your major bequest, annual contribution,
one-time gift – or a combo of all three – are
tax-deductible opportunities to give a helping hand
to the ISEs who will continue your legacy.
Visit iise.org/plannedgiving
For confidential assistance concerning donations
and scholarships, contact Donna Calvert at
dcalvert@iise.org or (770) 349-1108.
BUILD YOUR
LEGACY
IISE, AES, SHS Partners get membership discounts of 15-50%
• Corporate Partners can affordably develop staff and get them the training required to optimize the enterprise.
• IISE University Partners boost their faculty’s opportunities to network and publish research.
• Discounts of 40% on ISE magazine and Career Center advertising.
SPIRIT AERO FLIES HIGH BY PARTNERING WITH IISE
Welcome new and renewing
Corporate and University Partners
Consumer Energy • Clemson University • Herren Associates
Contact Doug Long, Global
Director of Strategic Alliances, at
dlong@iise.org or (770) 349-1109
Our corporate membership provided many
unexpected benefits. We have made contacts via
IISE that have provided input to simulation and
other tools. We have strengthened our relationship
with the local university through joint meetings, tours
and partnerships. The process of combining
memberships also helped bring our members and
diverse skills together and provided broader exposure
to how our company is applying their expertise within
our enterprise.
- David Logback, industrial engineer, Spirit AeroSystems Inc.
Spirit AeroSystems Inc., the world’s largest first-tier
aerostructures manufacturer, found that an IISE
Corporate Partnership brought together employees with
diverse skills, such as lean, simulation, quality, systems
engineering, shop support and ergonomics. Spirit
employees use IISE’s online resources, read ISE
magazine, attend conferences and network to stay up
to date with the latest applications and discover how