ORSIS 2015 Program Abstracts - Faculty of Industrial Engineering

Transcription

‫האגודה הישראלית לחקר ביצועים‬
‫)איל"ב(‬
‫הכינוס השנתי‬
‫‪ 10-11‬במאי ‪2015‬‬
‫בארגון‪:‬‬
‫הטכניון — מכון טכנולוגי לישראל‬
‫הפקולטה להנדסה תעשייה וניהול על שם ויליאם דוידסון‬
‫הפקולטה להנדסת חשמל‬
‫בתמיכת‪:‬‬
‫הוועדה המארגנת‪:‬‬
‫נחום שימקין )יו"ר(‬
‫מיכל פן‬
‫גייל גלבוע‪-‬פרידמן‬
ORSIS 2015: PROGRAM-AT-A GLANCE
Haifa, 10-11 May 2015
Leonardo Hotel – Almog Building
Sunday
09:00-09:30
09:30-09:40
09:40-10:30
10:30-10:50
10:50-11:10
11:10-12:25
‫פנינה‬
‫סוויטה‬
‫אלמוג‬
‫אקוומרין‬
12:25-13:45
13:45-14:35
‫אלמוג‬
14:35-15:25
‫פנינה‬
15:25-15:45
15:45-17:25
‫פנינה‬
‫אלמוג‬
‫סוויטה‬
17:45-19:45
19:45-21:15
‫לובי‬
Registration
‫אלמוג‬
Opening Session
‫אלמוג‬
S1 Naor Plenary Lecture:
Nimrod Megiddo
‫אלמוג‬
Awards Ceremony
Coffee Break
S2 Parallel Sessions:
a. Queues and their Applications
b. Game Theory 1
c. Supply Chain Management 1
d. Combinatorial Optimization 1
‫אקוואמרין‬
Lunch
& Orsis General Assembly (13:25)
S3 Plenary Lecture
Yishay Mansour
S4 Semi-Plenary Tutorials:
a. Rami Atar
b. Eran Hanany
Coffee Break
a. Topics in Continuous Optim.
b. Supply Chain Management 2
c. OR in practice
d. Health Care Management
Evening Program:
National Maritime Museum
Evening Program:
Dinner, with Guest Lecture
Monday
09:00-09:30
09:30-11:10
‫אלמוג‬
‫פנינה‬
‫סוויטה‬
11:10-11:30
11:30-12:20
‫פנינה‬
12:20-13:20
13:20-14:10
‫אלמוג‬
14:15-15:55
‫אלמוג‬
‫פנינה‬
‫סוויטה‬
15:55-16:10
16:10-17:25
‫אלמוג‬
‫פנינה‬
17:25-17:30
Registration
‫לובי‬
M1 Parallel Sessions:
a. Combinatorial Optimization 2
b. Queues and Stochastic Systems 1
c. Game Theory 2
d. Police Applications
Coffee Break
M2 Semi-plenary Tutorials:
a. Assaf Zeevi
b. Iddo Eliazar
Lunch
M3 Plenary Lecture:
Aharon Ben-Tal
a. Prize Winners Session
b. Scheduling
c. Optimization
d. Queues and Stochastic Systems 2
Coffee Break
a. Transportation
b. Water Management
c. Strategic Behavior in Queues
Farewell
.10 ‫ רח' דוד אלעזר‬,(‫ הרחב יותר‬,‫ בניין "אלמוג" )הבניין הצפוני‬,‫ מלון לאונרדו חיפה‬:‫מקום הכנס‬
."‫ סמוך לתחנת הרכבת "חוף הכרמל‬,‫המלון ממוקם על חוף הים בכניסה הדרומית לחיפה‬
‫ כולל יציאה וחזרה‬,17:45-19:45 ‫ נפתח בסיור מודרך במוזיאון הימי הלאומי )בשעות‬:‫ יום א' בערב‬:‫תוכנית חברתית‬
‫ ולאחריו ארוחת ערב מלווה בהרצאתו של ד"ר ג'ק סילברמן בנושא "השלכות סביבתיות של חיפושי גז‬,(‫למלון הכנס‬
."‫ונפט בים העמוק באזור המים הכלכליים של מדינת ישראל‬
‫ המרחק למלון הוא כקילומטר אחד בהליכה נוחה צפונה לאורך‬,"‫ יש לרדת בתחנת "חוף הכרמל‬-‫ ברכבת‬:‫תחבורה‬
.‫ שרות מוניות זמין ביציאה המערבית מהרכבת‬,‫ לחילופין‬.‫חוף הים‬
(‫ הכניסה מצפון דרך צומת דרך הים )מסעדת מקסים‬,(‫ יש להיכנס לרחוב דוד אלעזר )כביש פנימי לאורך החוף‬- ‫במכונית‬
.‫ הכניסה אליו מצפון לבניין אלמוג‬,‫ ניתן לחנות בחניון הפנימי של הבניין‬.(‫או מדרום דרך צומת מת"ם )מול דרך פרויד‬
Supported by
The organizing Committee:
Nahum Shimkin (Chair)
Michal Penn
Gail Gilboa-Friedman
1
Detailed Program ‐ Sunday, 10.5
09:00-09:30 Registration
09:30-09:40
S1 Opening Session
09:40-10:30
Naor Plenary Lecture
‫אלמוג‬
Nimrod Megiddo, IBM Almaden
The State-of-the-Art of the Theory of Linear Programming
Chair and Opener: Aharon Ben-Tal
10:30-10:50
Awards Ceremony
‫אלמוג‬
‫אלמוג‬
10:50-11:10 Coffee Break
11:10-12:25
S2a
Queues and Their
Applications
Chair: Uri Yechiali
‫פנינה‬
S2b
Game Theory 1
Chair: Eran Hanani
Amir Elalouf, Yael Perlman,
Uri Yechiali
Who Will Receive the Kidney? A Queueing Model for
Live Organ Allocation
Gabi Chanukov, T.
Avinadav, T. Chernonog,
U. Spiegel, U.Yechiali
Utilization of Server’s Idle Time for Greater Efficiency
in Queueing Systems
Nir Perel, Uri Yechiali
Noam Goldberg
An Unlimited Batch-Service Multi-Queue System
with Uniform and with Geometric Group-Joining
Policies
Nonzero-sum Nonlinear Network Interdiction
Gal Cohensius, Ella Segev
Sequential First Price Auction
Shoshana Anily
Total Balancedness of Regular Cooperative Games
Reut Noham, Michal Tzur
Network Design in Humanitarian Supply Chains: Pre
and Post Disaster Decision Making
The Two-Phase Stochastic Lotsizing Problem with
Optimal Timing of Additional Replenishment
‫סוויטה‬
S2c
Supply Chain
Management 1
Chair: Michal Tzur
Dina Smirnov, Yale T.
Herer, Retsef Levi, Assaf
Avrahami
Ohad Eisenhandler,
Michal Tzur
The Humanitarian Pickup and Distribution Problem
S2d
Combinatorial
Optimization 1
Niv Buchbinder,
Moran Feldman, Seffi
Naor, Roy Schwartz
Approximation Algorithms for Submodular
Maximization
Chair: Seffi Naor
Michael Dreyfuss, Yahel
Giat
Fill Rate Window as a Criterion for Spares Allocation
Yaarit M. Cohen,
Liron Yedidsion
The Traveling Repairman Problem on a Single Line
with Release Times
‫אלמוג‬
2
12:25-13:45 Lunch, & Orsis General Assembly (13:25)
13:45-14:35
S3 Plenary Lecture
‫אלמוג‬
Yishay Mansour, Tel Aviv University
Robust Inference and Local Algorithms
Chair: Nahum Shimkin
14:35-15:25
S4 Semi-Plenary Tutorials
Rami Atar, Technion
‫פנינה‬
Moderate Deviations and Heavy Traffic
Chair: Haya Kaspi
Eran Hanany
Dynamic Decisions under Ambiguity
Chair: Gail Gilboa Freedman
15:25-15:45 Coffee Break
15:45-17:25
S5a: Topics in
Continuous
Optimization
Chair:
Shoham Sabach
‫פנינה‬
S5b
Supply Chain
Management 2
Chair: Yael Perlman
‫אלמוג‬
Shimrit Shtern, Amir Beck
Shoham Sabach, Amir Beck
Marc Teboulle
Linearly Convergent Conditional Gradient Variants
for Non-strongly Convex Functions
An Alternating Semi-Proximal Method for Nonconvex
Problems
Jérôme Bolte, Edouard
Pauwels
Nonsmooth Optimization with Semi-algebraic Data:
Convergence Beyond the Proximal Setting
Nir Halman
Yigal Gerchak
Approximation Schemes for Stochastic Dynamic
Programs with Continuous State and Action Spaces
Consignment Contract for Mobile Apps Between a
Single Retailer and Competitive Developers with
Different Risk Attitudes
Strategic Inventory with Demand Uncertainty
Yale Herer,
Avinoam Tzimerman
3rd Party Logistics Coordinator Mechanism for
Transshipments in a Decentralized System
Tal Avinadav, Tatyana
Chernonog, Yael Perlman
The Effect of Risk Sensitivity on Supply Chain
Management under a Consignment Contract with
Revenue Sharing and Quality Investment
Tal Avinadav, Tatyana
Chernonog, Yael Perlman
(S5 continues on the next page)
3
S5c
OR in Practice
Vladimir Lipetz
Chair:
Michael Masin
Evgeny Shindin, Odellia
Boni, Michael Masin
S5d
Health Care
Management
Chair: Amir Elalouf
‫סוויטה‬
Closing the Gap between OR Models and Real
Applications
Robust Optimization of System Design
David Amid, Ateret AnabyTavor
Watson Tradeoff Analytics - Decision Making Made
Accessible for Everyone
Zilla Sinuany-Stern
OR/MS in the Higher Education
Developing an Optimal Appointment Scheduling for
Healthcare Systems with Non-flexible Uptake Time
Under Pre-determined Service Levels
‫מודל החלטה דינמי לקבלת כליה מן המת – בהתאם‬
‫ מיכאל בנדרסקי‬,‫ישראל דויד‬
‫לפונקציית הכשל והתפלגות החיים של המועמד להשתלה‬
Using a "Floating Patients" Approach for Improving
Guy Wachte, Amir Elalouf
Patients-Flow in Hospitals and Emergency
Departments
Challenges in Medical Quality Improvement
Eliaz Miller, MD
Illana Bendavid, Yariv N.
Marmor, Boris Shnits
Evening Program (Sunday)
17:45-19:45 National Maritime Museum (bus leaves at 17:45)
19:45-21:15 Dinner, with Guest Lecture:
‫ השלכות סביבתיות של חיפושי גז ונפט בים העמוק באזור‬.‫ המכון הלאומי לאוקיאנוגרפיה‬,‫ד"ר ג'ק סילברמן‬
.‫המים הכלכליים של ישראל‬
4
Detailed Program ‐ Monday, 11.5
09:00-09:30 Registration
09:30-11:10
M1a
Combinatorial
Optimization 2:
In Memory of
Uri Yovel
Chair: Rafi Hassin
‫אלמוג‬
M1b: Queues and
Stochastic Systems
1
Chair:
Galit Yom-Tov
‫פנינה‬
M1c
Game Theory 2
Chair: Gail
Gilboa Freedman
M1d
OR Applications in
the Israeli Police
Chair: Mali Sher
‫סוויטה‬
Rafi Hassin, Uri Yovel
Sequential Scheduling on Identical Machines
Tal Raviv, Uri Yovel,
Tom Cherchy
New Integer Programming Formulation of Routing
Problems in Dense Networks
Dorit S. Hochbaum, Asaf
Levin
Weighted Matching with Pair Restrictions
Enrique Gerstl, Gur
Mosheiov
Single Machine Scheduling to Minimize Total
Earliness-Tardiness with Unavailability Period
Nitzan Carmeli, Haya Kaspi,
Avishai Mandelbaum
Modeling and Analyzing IVR Systems, as a Special
Case of Self-services
Zhenghua Long, Nahum
Shimkin, Jiheng Zhang
Optimal Priority Control for Multiclass Many-server
Queues with General Patience Distributions
Galit B. Yom-Tov, Jing Dong, The Impact of Delay Announcements on Hospital
Network Coordination and Waiting Times
Elad Yom-Tov
Eitan Bachmat
Geometric Queueing Theory
Shiran Rachmilevitch
Egalitarian-Utilitarian Bounds in Nash's Bargaining
Problem
Payoff Externalities and Social Learning
Itai Arieli
Gail Gilboa Freedman,
Rann Smorodinsky
Axiomatization of Privacy
Gadi Fibich, Arieh Gavious
Revenue Equivalence of Large Asymmetric Auctions
,‫ תהילה הירש‬,‫חיננית אפרתי‬
‫ מלי שר‬,‫אירית נוביק‬
‫ייעול עבודת הסיירים ביחידת המקומות הקדושים של‬
‫ישראל‬-‫משטרת‬
‫ אורטל‬,‫שירלי ארקוסין‬
‫ מלי‬,‫ איריס פורמה‬,‫סעידיאן‬
‫שר‬
‫השמת עובדים אופטימאלית‬
,‫ ינון סבתו‬,‫אלי אורבך‬
‫ מלי שר‬,‫מיכאל דרייפוס‬
‫ישראל‬-‫ייעול פריסת מצלמות האכיפה במשטרת‬
Mali Sher, Nicole Adler,
Shalom Hakkert
A Traffic Enforcement Camera Operational Model
11:10-11:30 Break
5
11:30-12:20
M2 Semi-Plenary Tutorials
Assaf Zeevi, Columbia University
‫פנינה‬
Chasing Bandits: Exploration and Exploitation in a
Non-stationary Environment
Chair: Michal Penn
Iddo Eliazar, Intel
From Inequality to Big Data: The Sociogeometry of Sizes
Chair: Uri Yechiali
12:20-13:20 Lunch
13:20-14:10
M3 Plenary Lecture
‫אלמוג‬
Aharon Ben-Tal
Some Remedies for Some Intractable Optimization Problems
Chair: Avishai Mandelbaum
14:15-15:55
M4a: Opimization
and Stochastics
Amir Beck, Nadav Hallak
(Rothblum Award)
On the Minimization Over Sparse Symmetric Sets:
Projections, Optimality Conditions and Algorithms
Josh Reed, Yair Shaki
A Fair Policy for the G/GI/N Queue with Multiple
Server Pools
Prize Winners
Session
Chair: Moshe Haviv
‫אלמוג‬
(Rothblum Award)
Liron Ravner
(Mehrza Award)
Equilibrium Arrival Times to a Queue with Order
Penalties
Arik Sadeh
An Asymptotically Optimal Online Algorithm to
Choosing Binary Factors Affecting Purchasing in eCommerce
Public Transportation with Flexible Timetable
M4b
Scheduling
Tal Grinshpoun, Elad
Shufan, Hagai Ilani
Chair: Dvir Shabtay
Enrique Gerstl, Gur
Mosheiov
Scheduling with Two Competing Agents to Minimize
Total Weighted Earliness
Baruch Mor, Gur Mosheiov
Minimizing Maximum Earliness and Minimizing the
Number of Early Jobs on a Proportionate Flowshop
An Asymptotically Optimal Online Algorithm to
Minimize the Total Completion Time on Two
Multipurpose Machines with Unit Processing Times
‫פנינה‬
Dvir Shabtay, Shlomo Karhi
(M4 continues on the next page)
6
M4c
Optimization
Chair: Gideon Weiss
‫סוויטה‬
Saleh Soltan, Mihalis
Yannakakis, Gil Zussman
Renata Poznanski, Refael
Hassin
Rafiq Mansour, Yair Censor
Evgeny Shindin, Gideon
Weiss
M4d: Queues and
Stochastic Systems
2
Chair: Yoav Kerner
Yoav Kerner, Opher Baron
Ruth Sagron, Gad
Rabinowitz, Israel Tirkel
Barron Yonit
Tal Avinadav,
Tatyana Chernonog, Yael
Lahav, Uriel Spiegel
Joint Cyber and Physical Attacks on Power Grids:
Graph Theoretical Approaches for Information
Recovery
Optimal Multi-Period Network Flows with Coupling
Constraints
New Douglas-Rachford Algorithmic Structures and
Their Convergence Analyses
A simplex-type Algorithm for Continuous Linear
Programming
Queueing Model for Safety Stock Inventory Model
with Perishable Items and General Distribution
Hybrid Simulation-Regression Approximation for
Tandem Queues with Downtime Events
Clearing Control Policies for MAP Inventory Process
with Partially Satisfied Demand
Dynamic Pricing and Promotion Expenditures in an
EOQ Model of Perishable Items
15:55-16:10 Break
16:10-17:25
M5a
Transportation
Mor Kaspi, Tal Raviv,
Michal Tzur
Regulating One-Way Vehicle Sharing Systems
through Parking Reservation Policies
Chair: Tal Raviv
Sharon Datner, Tal Raviv,
Michal Tzur
Setting Inventory Levels in Bike-Sharing Networks
Hila Hindi-Ling, Hillel BarGera, Arie Sachish
The Effect of a Quayside Cranes Buffer on Ships
Unloading Process
‫אלמוג‬
Modeling Combined Technological, Environmental
M5b
Amos Bick, Ioannis K.
Water Management Kalavrouziotis, Gideon Oron and Economic Considerations in Domestic Sludge
Chair: Amos Bick
‫פנינה‬
M5c: Strategic
Behavior in Queues
Chair: Moshe Haviv
David Raz, Ariel Daliot
Beni Lew, Olga Tarnapolski,
Vladimir Yudachev, Amos
Bick
Moshe Haviv, Binyamin Oz
Moshe Haviv, Liron Ravner
Nahum Shimkin
Reuse via the Analytic Hierarchy Process (AHP)
A Generic Modeling Language for Water Supply
Systems Optimization
Membrane Treatment of Brackish Groundwater for
Unrestricted Use for Irrigation and Sustainable
Agricultural Production: Decision Analysis via The
Hasse Diagram Technique (HDT)
Self-Regulation of a Queue via Random Priorities
Accumulating Priority Queue with Strategic
Customers
What to (Truthfully) Tell Customers to Make Them
Join a Queue
17:25-17:30 Fairwell
7
ABSTRACTS
Plenary Lectures
S1: Naor Plenary Lecture. Chair and Opener: Aharon Bental
Speaker: Nimrod Megiddo, IBM Almaden Research Center
The State-of-the-Art of the Theory of Linear Programming
The talk will survey the following topics: (i) interior-point methods, (ii) worst-case, probabilistic
and smoothed analysis of the simplex method, (iii) strongly polynomial bounds, and (iv) the
diameter of a polytope.
S3: Plenary Lecture. Chair: Nahum Shimkin
Yishay Mansour, Tel Aviv University
Robust Inference and Local Algorithms
Robust inference is an extension of probabilistic inference, where some of the observations may
be adversarially corrupted. We limit the adversarial corruption to a finite set of modification rules.
We model robust inference as a zero-sum game between an adversary, who selects a modification
rule, and a predictor, who wants to accurately predict the state of nature. There are two variants of
the model, one where the adversary needs to pick the modification rule in advance and one where
the adversary can select the modification rule after observing the realized uncorrupted input. For
both settings we derive efficient near optimal policy runs in polynomial time. Our efficient
algorithms are based on methodologies for developing local computation algorithms. We also
consider a learning setting where the predictor receives a set of uncorrupted inputs and their
classification. The predictor needs to select a hypothesis, from a known set of hypotheses, and is
tested on inputs which the adversary corrupts. We show how to utilize an ERM oracle to derive a
near optimal predictor strategy, namely, picking a hypothesis that minimizes the error on the
corrupted test inputs.
Based on joint works with Uriel Feige, Aviad Rubinstein, Robert Schapira, Moshe Tennenholtz,
Shai Vardi.
8
M3: Plenary Lecture. Chair: Avishai Mandelbaum
Aharon Ben-Tal, Technion
Some Remedies for Some Intractable Optimization Problems
The need to solve real-life optimization problems poses frequently a severe challenge, as the
underlying mathematical programs threaten to be intractable. The intractability can be attributed
to any of the following properties: large dimensionality of the design dimension; lack of convexity;
parameters affected by uncertainty. In problems of designing optimal mechanical structures (truss
topology design, shape design, free material optimization), the mathematical programs typically
has a large dimensional Semi Definite Program. Some Signal Processing and Estimation problems
may result in nonconvex formulations. In the wide area of optimization under uncertainty, some
classical approaches, such as chance (probabilistic) constraints, give rise to nonconvex NP-hard
problems.
In all the above applications, we explain how the difficulties were resolved. In some cases this
was achieved by mathematical analysis (notoriously duality theory) which converted the problems
(or its dual) to a tractable convex program. In the Robust Control example, a reparameterization
scheme is developed under which the problem is converted to a tractable deterministic convex
program.
Semi-Plenary Tutorials
Semi-Plenary Session S4
Rami Atar, Technion.
Chair: Haya Kaspi
Moderate Deviations and Heavy Traffic
There is a vast literature on heavy traffic analysis of controlled queueing models at the diffusion
scale. We will motivate the study of these models at the moderate deviations scale, and describe
recent developments in this direction.
Eran Hanany, Tel Aviv University. Chair: Gail Gilboa-Freedman
Dynamic Decisions under Ambiguity
Preferences that allow for decision makers to care about ambiguity have drawn increasing interest
in recent years. Dynamic consistency is the fundamental requirement that contingent plans made
at an initial time should remain optimal at later times. This requirement leads to Bayesian updating
under ambiguity neutrality, but to different updating rules under ambiguity aversion. I will present
recent advances in this field, including a theory of belief polarization as an optimal response to
ambiguity, and equilibrium notions for incomplete information games involving players who
perceive ambiguity about the types of others.
9
Semi-Plenary Session M2
Assaf Zeevi, Columbia University.
Chair: Michal Penn
Chasing Bandits: Exploration and Exploitation in a Non-stationary Environment
In a stochastic multi-armed bandit (MAB) problem a gambler needs to choose at each round of
play one of K arms, each characterized by an unknown reward distribution. Reward realizations
are only observed when an arm is selected, and the gambler¹s objective is to maximize his
cumulative expected earnings over some given horizon of play T. To do this, the gambler needs to
acquire information about arms (exploration) while simultaneously optimizing immediate rewards
(exploitation). Bandit problems have been studied extensively since their inception. The bulk of
this literature is concerned with settings where the reward distributions are stationary, namely, the
statistical properties of the arms do not change over time. In this talk we will consider a formulation
that relaxes this restriction, highlight some of the key theoretical findings that characterize this
setting, and discuss an emerging application domain in the space of online services that motivates
our work.
Iddo Eliazar, Intel Corporation. Chair: Uri Yechiali
From Inequality to Big Data: The Sociogeometry of Sizes
In the era of big data we are inundated with large datasets of sizes – non-negative numerical values
representing count, score, length, area, volume, duration, mass, energy, etc. These datasets display
numerous types of statistical variability, or intrinsic randomness, commonly quantified either by
standard deviation, or by entropy. The standard deviation measures the sizes' Euclidean divergence
from their mean, entropy measures the sizes' divergence from the benchmark of pure determinism,
and both these gauges are one-dimensional. In this talk we overview a sociogeometric framework
of quantifying the randomness of datasets of sizes. The framework follows a socioeconomic
approach of measuring the sizes' inequality – their divergence from the benchmark of pure
egalitarianism – and yields a rich and multidimensional methodology of gauging the sizes'
statistical variability via sociogeometric inequality indices. The aim of this overview is to serve
both researchers and practitioners as a crash-intro to the sociogeometry of sizes, and as a crashmanual to the implementation of this methodology.
References:
I. Eliazar, The sociogeometry of inequality: Part I, Physica A (2015) in press.
DOI: 10.1016/j.physa.2014.12.021
I. Eliazar, The sociogeometry of inequality: Part II, Physica A (2015) in press.
DOI: 10.1016/j.physa.2015.01.016.
10
Parallel Sessions (in Chronological Order)
Sunday, 10.5
S2a: Queues and Their Applications. Chair and Organizer: Uri Yechiali
Amir Elalouf, Yael Perlman, Uri Yechiali
Who Will Receive the Kidney? A Queueing Model for Live Organ Allocation
We consider models of dynamically allocating randomly arriving kidneys to candidates waiting for
transplant. Two main models are investigated: (i) a kidney is allocated only to a candidate with the same
blood type, and (ii) a kidney with blood type O is allocated to either O or A type of candidate, while type
A kidney is allocated only to type A candidate. The allocation is based on the measure Expected Value of
Transplantation (EVT) which takes into consideration Human Leukocyte Antigen (HLA) matching. We
also study the case where a fraction of the available kidneys is kept for future candidates. Finally, we treat
the case of exponential life time of a kidney and an exponential residual life time of candidates.
Gabi Chanukov, T. Avinadav, T. Chernonog, U. Spiegel, U.Yechiali
Utilization of Server’s Idle Time for Greater Eficiency in Queueing Systems
We consider an M/M/1-type queueing system in which the server utilizes his/her idle time to partially
prepare future services off-line in order to reduce the on-line service duration. Such a routine is applicable
for service systems in which part of the service (termed “start”) can be prepared without the presence of the
customer. The “starts” are produced and stored during periods of time when there are no customers in the
system. When a customer arrives, the server stops producing new “starts” and completes an inventoried
“start” (if any) in a faster processing rate. When all inventoried “starts” are used, the service rate returns to
normal until the next time the server is idle and returns producing new “starts” again. The decision variable
is the maximal number of inventoried "starts" in order to optimize an economical measure (e.g., minimal
total cost rate or average waiting time of customers). Five model variants are formulated and analyzed: (i)
The duration of a full service and of each of its two parts is exponentially distributed; (ii) The duration of
each service part is exponentially distributed, and the duration of a full service is a simple summation of its
two parts; (iii) The customers know when inventoried “starts” are available and increase their arrival rate;
(iv) Some of the inventoried “starts” suffer from decay, thus losses, since they cannot be used; and (v) The
server faces two types of customers: one that is willing to pay more for a fast service and another that pays
less for a normal service rate.
Nir Perel, Uri Yechiali
An Unlimited Batch-Service Multi-Queue System with Uniform and with Geometric
Group-Joining Policies
We study a multi-queue single-server system with unlimited-size batch service under either Uniform or
Geometric group-joining policy, where the next queue to be served is the one with the most senior customer.
We derive a set of performance measures related to sojourn times, group sizes and busy periods, and present
a corresponding set of numerical results. The effects of the system's parameters on the performance
measures are investigated.
11
S2b: Game Theory 1. Chair: Eran Hanani
Noam Goldberg
Nonzero-sum Nonlinear Network Interdiction
A novel nonzero-sum game is presented for a variant of a classical network interdiction problem. In this
model an interdictor (e.g. an enforcement agent) decides how much of an inspection resource to spend along
each arc in the network in order to capture the evader (e.g. a smuggler). The evader selects a probability
distribution on paths from source nodes to destinations. The evasion probabilities nonlinearly decrease in
the inspection resources spent by the interdictor. We show that for logarithmically convex functions, Nash
equilibria of this game can be efficiently computed. A special case of exponential functions is further
analyzed and presented with examples.
Gal Cohensius, Ella Segev
Sequential First Price Auction
We study asymmetric first price auctions in which bidders place their bids sequentially, one after the other
and only once. We show that with a strong bidder and a weak bidder (in terms of first order stochastic
dominance of their valuations distribution functions), already with small asymmetry between the bidders,
the expected revenue in the sequential bidding first price auction (when the strong bidder bids first) is higher
than in the simultaneous bidding first price auction. Moreover it is higher than the expected revenue in the
second price auction. The expected payoff of the weak bidder is also higher in the sequential first price
auction. Therefore a seller interested in increasing revenue facing asymmetric bidders may find it beneficial
to order them and let them bid sequentially instead of simultaneously. In terms of efficiency, both the
simultaneous first price auction and the sequential first price auction cannot guarantee full efficiency (as
opposed to a second price auction). The sequential bidding auction when the stronger bidder bids first
achieves lower efficiency than the simultaneous auction. However, when the order is reversed and the
asymmetry is large enough the sequential first price auction achieves higher efficiency than the
simultaneous auction.
Shoshana Anily
Total Balancedness of Regular Cooperative Games
The paper "Cooperation in service systems" by Anily and Haviv (Operations Research 2010) attracted the
attention of the research community as it proves that the most basic model of cooperation in queueing,
where a number of M/M/1 systems cooperate by forming a single M/M/1 system whose arrival (service)
rate is the sum of the individual respective rates, is totally balanced. The interest in the paper was also a
result of the original proof's technique: the authors define an auxiliary game that is monotone and whose
core is contained in the core of the game and proved that the auxiliary game is concave (submodular),
implying that its core is totally balanced, and therefore also the original game is totally balanced. The idea
of the auxiliary game can easily be applied to any non-monotone game in order to generate a monotone
new game whose core is a subset of the core of the primary game. However, the proof that the characteristic
function of the auxiliary game is concave is custom-made to the specific given set function. Suchlike proofs
are usually tedious. Several researchers have raised the question whether the idea behind the auxiliary
game could be further generalized beyond that special game as a tool in proving total balancedness of other
cooperative games.
In this paper we consider regular games where each player is associated with a real vector, and the
characteristic function of a coalition depends only on the collection of the real vectors that are
associated with the players. Under a few conditions the characteristic function of regular games,
which is a set function, can be replaced by a mapping of real vectors into the real numbers. We
prove that if this mapping satisfies the law of diminishing returns then the auxiliary game is
concave, and therefore the game itself is totally balanced. We demonstrate the simplicity of the
technique on the above described game, as well as on other games.
12
S2c: Supply Chain Management 1
Chair: Michal Tzur
Reut Noham, Michal Tzur
Network Design in Humanitarian Supply Chains: Pre and Post Disaster Decision Making
Humanitarian logistics is an emerging field that addresses the major challenges in providing humanitarian
relief operations for both natural and man-made disasters. Strategies to overcome these challenges include
disaster preparedness and response, under high uncertainty and limited availability of resources and
infrastructure to address needs. Existing models in the academic literature address network design and
resource allocation challenges that are relevant to pre- and post-disaster situations, respectively. However,
they adopt a global optimization point of view, which may not be attainable, due to the actual decision
making process. The latter is based mostly on practitioners' knowledge and experience, simple rules of
thumb, and the local population behavior. In our work, we develop new mathematical models that represent
practical considerations such as those mentioned above. For small/medium instances of the considered
problem we present an efficient optimal solution method while for large instances we use a heuristic
algorithm which is based on Tabu search. Our preliminary results demonstrate that the effectiveness of
network design decisions (made at the pre-disaster phase) is sensitive to post-disaster decisions, and
therefore, to the extent possible, it is critical to accurately model/predict post-disaster decisions during the
pre-disaster phase.
Dina Smirnov, Yale T. Herer, Retsef Levi, Assaf Avrahami
The Two-Phase Stochastic Lotsizing Problem with Optimal Timing of Additional
Replenishment
Recent advances in Information Technology have provided decision makers in the supply chain with
extensive, often real time and accurate, data. Wisely used data can assist in improving the performance of
inventory systems. In particular, for single-period systems it has become possible to receive actual and
accurate sales data more frequently than once in the sales period. This has enabled decision makers to
execute an additional replenishment based on early sales information. We are given a single sales period
and a single possibility to review the inventory level after the start of sales and to execute an additional
replenishment if necessary. The items from the additional replenishment are intended to be sold during that
same sales period. It is desirable to find the optimal quantity to replenish before the start of sales, the optimal
moment to perform the additional review, and the optimal quantity to replenish at the moment of additional
review. This problem has not been addressed with respect to this set of policy decisions as of yet. In this
work we use analytical tools and propose an exact and tractable algorithm for simultaneously determining
these policy decisions. Our model is applicable to a wide range of businesses such as bakeries, retail of
seasonal, perishable or technology-related goods, and print industry.
Ohad Eisenhandler, Michal Tzur
The Humanitarian Pickup and Distribution Problem
Food rescue, i.e., the collection of perishable products from food suppliers who are willing to make
donations and their distribution to welfare agencies that serve individuals in need, has become increasingly
widespread in recent years. This is due to economic crises that have increased the demand for nutritional
aid, and the benefit to donors who can avoid in this way the costs of destroying excess production while
reflecting a social-aware image. The problem we study focuses on the logistic challenges of a food bank
coordinating this operation on a daily basis, using vehicles with limited capacity whose travel time cannot
exceed an imposed maximal duration (defined by the driver's working hour regulations). We model this
problem as a routing–allocation problem, with the aim of maintaining equitable allocations to the different
agencies in each period, while delivering as much as possible in total. We discuss an appropriate objective
function that promotes effectiveness and equity. We show how these two measures can be combined in a
way that satisfies desired properties of the allocation, that is easy to compute and implement within a
mathematical formulation, and that balances effectiveness.
13
S2d: Combinatorial Optimization 1 Chair: Seffi Naor
Niv Buchbinder, Moran Feldman, Seffi Naor, Roy Schwartz
Approximation algorithms for Submodular Maximization
The study of combinatorial problems with submodular objective functions has recently attracted much
attention, and is motivated by the principle of economy of scale, prevalent in real world applications.
Submodular functions are also commonly used as utility functions in economics and algorithmic game
theory. In combinatorial optimization submodular functions and submodular maximization play a major
role as several well-known examples of submodular functions include cuts in graphs and hypergraphs, rank
functions of matroids and covering functions. In this talk I will present several recent results for the problem
of maximizing a general submodular function.
Michael Dreyfuss, Yahel Giat.
Fill Rate Window as a Criterion for Spares Allocation
The biggest problem for the successful adoption of electric cars is the frequent need to recharge the battery
and the waiting time associated with it. One of the suggestions to overcome this problem is that carmakers
retain ownership of batteries and provide service stations in which customers replace their depleted batteries
with recharged batteries in lieu of waiting for their battery to recharge. Motivated by this approach, we
consider spare allocations in an exchangeable-item multi-location repair system with Poisson arrivals and
ample servers with general repair time distribution. Customers expect to be served within a certain time
window and penalize the service provider if they have to wait more than that window of time. Accordingly,
instead of minimizing average waiting time, we suggest that firms should consider maximizing the fill rate
window, i.e. the probability that customers wait less than a predetermined time window. We derive the
entire system’s fill rate window for any time window, and characterize its functional form. For each location
the fill rate window can be either concave or S-shaped. For a sufficiently large time window, we show that
the system’s problem is concave in which case the “biggest bang for the buck” approach can be used to
solve the optimal spares allocation. When the window is not large enough, at least some of the locations
have an S-shaped fill rate window. In this case we define a concave covering function. Since the covering
function is concave we can efficiently derive its optimal solution. To find the optimal solution of the original
problem we need to execute a correction procedure, which is provided. We provide a numerical example
motivated by the recent unsuccessful attempt to introduce electric cars into Israel.
Yaarit M. Cohen, Liron Yedidsion
The Traveling Repairman Problem on a Single Line with Release Times
The Traveling Repairman problem (TRP) is a well-known NP-hard problem (also referred to in the
literature as The Delivery Man Problem and The Minimum Latency Problem). In the classical TRP, a
repairman has to visit each of n stationary targets exactly once, in order to minimize the sum of the targets'
flow time. Where the flow time of target i (Fi) is the time since the target appears till the time it is
intercepted. In this research we consider a special case of TRP where all targets are confined to a single
line. However, some targets might not be available at time zero and appear later on. We denote this problem
as the Single Line TRP (SL-TRP). Algorithms for SL-TRP have real-life applications that can help civil
and military needs. Consider a line of machines each one requires a delivery of raw material delivered by
a robotic arm moving along a production line; or a border patrol unit that has to reach numerous check
points along the border based on alerts of invasion. Note that the border doesn't have to be strait to be
considered a "strait line" as long as the patrol unit travels alongside the border line. In both of these problems
the robot or the border patrol unit has to visit each point exactly once and should complete its tour and
minimize the sum of the targets' flow time.
We proved this problem to be NP-hard by using a reduction from a special case of the partition problem.
Next, we suggest several heuristics as well as exact Branch and Bound (B&B) algorithm for solving the
SL-TRP. We use computational experiments in order to evaluate and compare their performances as well
as the efficiency of the B&B algorithm when using large instances.
14
S5a: Topics in Continuous Optimization
Chair: Shoham Sabach, Organizer: Amir Beck
Shimrit Shtern, Amir Beck
Linearly Convergent Conditional Gradient Variants for Non-strongly Convex Functions
The conditional gradient method was presented by Frank and Wolfe in 1956. Its aim is to minimize a smooth
function over a compact convex set, where each iteration of the method consists of minimizing a linear
function over the feasible set. Generally, the method’s rate of convergence is 1/k, and is linear only in
specific cases, where the function optimized is strongly convex and the optimal solution lies in the relative
interior of the feasible set, or when the set is uniformly (strongly) convex and the gradient of the objective
function is bounded away from zero. Lately, two variants of the conditional gradient algorithm -- the away
step conditional gradient and the local conditional gradient -- were proven to converge linearly for
minimizing strongly convex functions over polyhedral sets. We extend these results, and prove that these
algorithms also admit a linear rate of convergence for well-structured functions, which are not strongly
convex. Moreover, for the version that incorporates away steps, we provide a new convergence rate with
computable constants that also enables the comparison between the two algorithms.
Shoham Sabach, Amir Beck, Marc Teboulle
An Alternating Semi-Proximal Method for Nonconvex Problems
We consider a broad class of regularized structured total-least squares problems (RSTLS) encompassing
many scenarios in image processing. This class of problems results in a nonconvex and often nonsmooth
model in large dimension. To tackle this difficult class of problems we introduce a novel algorithm that
blends proximal and alternating minimization methods by benefi cially exploiting data information and
structures inherently present in RSTLS. The proposed algorithm, which can also be applied to more general
problems, is proven to globally converge to critical points, and is amenable to e_fficient and simple
computational steps.
Jérôme Bolte, Edouard Pauwels
Nonsmooth Optimization with Semi-algebraic Data: Convergence Beyond the Proximal
Setting
We focus on convergence of iterative schemes for non-smooth non-convex optimization in finite
dimension. Most of current results are given for "prox-friendly" data: the nonsmooth part can be handled
through efficiently computable operators. Many methods and applications do not fit this setting. We focus
on Sequential Quadratic Programming ideas for general Nonlinear Programs. Despite their large usage,
these methods lack satisfactory convergence analysis. This work constitutes a step toward the obtention of
such theoretical guaranties. We combine properties of local tangent majorizing models with results from
algebraic geometry to analyse the asymptotic properties of two recent methods for solving general
Nonlinear Programming problems.
Nir Halman
Approximation Schemes for Stochastic Dynamic Programs with Continuous State and
Action Spaces
Dynamic optimization problems are generally NP-hard to solve, thus approximated solutions are
of interest. The main approaches to tackle dynamic optimization problems are optimal control
(Pontryagin's minimum principle) and dynamic programming (Bellman's optimality equation).
Taking the latter approach, I'll consider a broad subfamily of dynamic programs (DP's) that is still
NP-hard. DP's of this class with discrete state and action spaces are known to admit relative-error
approximation schemes. I'll show that such DPs with *continuous* state and action spaces are
harder to approximate than their discrete counterparts. In particular, they do not necessarily admit
constant factor approximations. Despite this discouraging result, I'll be successful at designing for
them approximation schemes by using a novel measure of error
15
S5b: Supply Chain Management 2 Chair and Organizer: Yael Perlman
Tal Avinadav, Tatyana Chernonog, Yael Perlman
Consignment Contract for Mobile Apps Between a Single Retailer and Competitive
Developers with Different Risk Attitudes
Mobile applications (apps) are software programs designed to run on smartphones and tablets. They are
commonly downloaded through application distribution platforms, such as the Apple (iTunes) App Store,
Google Play, the Windows Phone Store and BlackBerry App World. As suggested by Apple’s central
marketing message—“there’s an app for that”—the market for apps is crowded and diverse (BBC Trust,
2010). At the same time, there is intense competition among companies marketing similar apps. For
example, the iTunes App Store offers at least twelve device finder apps, similar to “Find My iPhone”; these
apps compete with one another in terms of both price and quality (Myers, 2012). Clearly, the question of
how to manage brand competition and channel competition is important both for app developers (the
suppliers) and the platform distributor (i.e., the app retailer).
Yigal Gerchak
Strategic Inventory with Demand Uncertainty
We consider a decentralized two-period problem with uncertain demands. The manufacturer determines
the second period's wholesale price based on the amount of inventory left at the retailer from the first period.
Research with deterministic demands has shown that the retailer will over-order in the first period, so that
what is not sold then will cause the manufacturer to set a low wholesale price, hence the name Strategic
Inventory. We examine the veracity of this phenomenon in a setting with random demands. We do so in a
setting with fixed retail prices, as well as in one with price-sensitive demand.
Yale Herer, Avinoam Tzimerman
3rd Party Logistics Coordinator Mechanism for Transshipments in a Decentralized System
We investigate a multi-retailer single period stochastic lot sizing problem. The retailers are independent
and experience stochastic demand for a single item. Moreover, the retailers act separately, doing what is in
their own best interest, i.e., decentralized control. We add to this constellation the option of transshipments.
Transshipments have been widely studied with central control, but much less so in a decentralized setting.
We introduce transshipments by proposing a mechanism that involves the presence of a third party whom
we call 3PLC, 3rd Party Logistics Coordinator. 3PLC incentivizes the retailers to take part in the mechanism
by providing each retailer with payments for their holding and shortage costs (the incentive), and in return,
3PLC is allowed to reap the benefits of transshipments. Using 3PLC we are able to implement
transshipments and coordinate the supply chain. The 3PLC mechanism answers this challenge without
sharing private information such as cost and demand data among the retailers. We also show how the
mechanism can be used to split supply chain profits among the players in any way desired. Most
importantly, this mechanism both retains the independent nature of the entities and completely exploits the
benefits of the centralized system.
Tal Avinadav, Tatyana Chernonog, Yael Perlman
The Effect of Risk Sensitivity on Supply Chain Management under a Consignment
Contract with Revenue Sharing and Quality Investment
We analyze pricing and quality investment strategies in a two-echelon supply chain of mobile applications
(apps) under a consignment contract with revenue sharing. Specifically, we focus on how risk-sensitive
behavior of supply chain members affects chain performance. The platform provider sets the level of
revenue sharing, and the app developer determines the investment in quality and the selling price of the
app. The demand for an app, which depends on both price and quality investment, is assumed to be
uncertain, so the risk attitude of the supply chain members has to be considered. The members′ equilibrium
strategies are analyzed under different attitudes toward risk: averse, neutral and seeking. We show that the
retailer's utility function has no effect on the equilibrium strategies, and suggest schemes to identify these
strategies for any utility function of the developer. We find that (i) the revenue sharing contract circumvents
the double marginalization effect associated with vertical competition and therefore yields the best selling
16
price for the customer; (ii) a decentralized supply chain sometimes performs better than a centralized one;
and (iii) a risk-seeking developer may obtain a higher expected profit than does a risk-neutral developer.
S5c: OR in Practice
Chair and Organizer: Michal Massin
Vladimir Lipetz
Closing the Gap Between OR Models and Real Applications
Optimization can help and provide productivity boost in many domains,however, its practical usage is
pretty limited. In this presentation we focus on features and tools that help to make optimization usable by
end users. The users are typically domain experts, but usually have no optimization and programming
background. We focus on performance, usability, and maintainability aspects and how these aspects can
change the underlying optimization models. Finally, we demonstrate the process of transferring scheduling
models into reusable tools in two real-life examples. The first one is an integrated part of IBM tool (Maximo
Asset Manager) for technicians scheduling during maintenance and emergency activities. The second
example is a stand-alone enterprise level application for El Al staff scheduling
Evgeny Shindin, Odellia Boni, Michael Masin
Robust Optimization of System Design
The data of real-world optimization problems are usually uncertain, that is especially true for early stages
of system design. Data uncertainty can significantly affect the quality of the nominal solution. Robust
Optimization (RO) methodology uses chance and robust constraints to generate a robust solution
immunized against the effect of data uncertainty. RO methodology can be applied to any generic
optimization problem where one can separate uncertain numerical data from the problem's structure. Since
2000, the RO area is witnessing a burst of research activity in both theory and applications. However, RO
could lead to over-conservative requirements, resulting in typical-case bad solutions or even empty solution
spaces. This drawback of the classical RO methodology can be overcome by distinguishing between real
decision variables and so-called /state/ variables. While the first type should satisfy the chance or robust
constraints and their value cannot depend on a specific realization of the uncertain data, the state variables
are adjustable (i.e., their value can depend on the specific realization of the uncertain data), since most of
the constraints defining state variables merely “calculate” their exact value, and hence are always satisfied.
In this paper we summarize how adjustable RO approach can be applied to a general uncertain linear
optimization problem. Then, using an allocation example we demonstrate how this approach can be
integrated in the design optimization process and its impact on the optimal system design.
David Amid, Ateret Anaby-Tavor
Watson Tradeoff Analytics - Decision Making Made Accessible for Everyone
The Watson Tradeoff Analytics service helps people optimize their decisions while striking a balance
between multiple, often conflicting, objectives. The service can be used to help make complex decisions
like what mortgage to take or which laptop to purchase. Tradeoff Analytics uses Pareto filtering techniques
to identify the optimal alternatives across multiple criteria. It then uses various analytical and visual
approaches to help the decision maker explore the pros and cons of their alternatives.
Zilla Sinuany-Stern
OR/MS in the Higher Education
Much has been written on Operations Research (OR) and Management Science (MS) in many other areas
such as health care and energy. I see the Higher Education (HE) industry/sector as a positive leading catalyst
for economic development and change. We are witnessing the rise of tuition and the oversupply of college
graduates - a new era of e-learning and new accreditation of online institutions, globalization of HE, and
new outlook on learning outputs and proficiencies required for the workplace. The HE sector is changing
fast and requires more efficiency, more agile and lean planning, and new strategies, and quality assurance.
Namely, there is a need for more applications of OR/MS, and IE methodologies, and business approaches,
which are already occurring in the HE sector.
17
The old type leaders of HE are not always equipped to foresee these processes. They need the help of our
trade and tools in general and quantitative methods in particular to survive. As many of the OR/MS experts
are members of HE institutions. Many of us wish to improve our institutions. Research students in our fields
should be more encouraged to apply OR/MS and IE methodologies in HE.
Areas in HE where OR can be used include: facility planning and scheduling, faculty outputs and
compensation, budgeting, quality assurance, international comparisons, ranking universities, students
choice of institution, students admission. OR methodologies used for HE are: optimization models,
scheduling, forecasting, simulation, Data Envelopment Analysis, game theory, multi-criteria decision
analysis, etc. We will present several examples: 1. Measuring the differentiability of faculty salaries in
Israeli universities by rank, by institution, and by faculty outputs, using statistical measures 2. Budget
allocation in HE institution using quadratic optimization model, and linear programming. 3. Measuring the
efficiency of academic departments using Data Envelopment Analysis (DEA).
S5d: Health Care Management
Chair and Organizer: Amir Elalouf
Illana Bendavid, Yariv N. Marmor, Boris Shnits
Developing an Optimal Appointment Scheduling for Healthcare Systems with Non-flexible
Uptake Time under Pre-determined Service Levels
One of the critical steps in patient care path is diagnosis. The demand for advance imaging tests, such as
CT, MRI and PET, increased dramatically in the past 15 years. Since imaging equipment remains relatively
expensive, in order to fit the demand, the imaging resources must be managed effectively. In most
healthcare systems, where examination length is uncertain (stochastic), the goal of the appointment
scheduling need to balance between resource utilization and patient waiting times.
In some imaging scans, such as PET, a radiopharmaceutical (radioactive substance) is injected to the
patients in order to perform the diagnosis. In these systems, the time between the substance injection and
the scan is non-flexible (for example, due to short half-life duration). This constraint makes the patient
appointment scheduling more challenging, because, on the one hand, there is a predetermined time required
between the injection of the radiopharmaceutical and the scan – the uptake time (time that it takes to the
substance injected to be absorbed into the body), while on the other hand, if at the end of the expected
uptake time the scanner is not available, the quality of the scan is jeopardized. Of course, the availability of
the scanner is a consequence of appointments and durations of prior scans.
Therefore, the aim of this work is to develop a method for determining a patient appointment scheduling in
a system with non-flexible uptake time in order to minimize the end of day and increase resource utilization
while keeping minimal pre-determined service levels.
To this end, we consider the following setting: a given sequence of patients is to be scheduled on one
scanner machine; the durations of scans are normally distributed with various expectations and variances;
a minimal probability for each appointment to start on time is required (service level).
In order to solve this stochastic problem, we formulate its equivalent deterministic problem, based on
simulated data, as a mixed-integer linear programming. To overcome the dimensionality limitations, we
also develop a simulation-based sequential model. We found that a constant slot per scan, as a benchmark,
is inferior to our method both in achieving stable service level and reducing the end of day.
‫ מיכאל בנדרסקי‬,‫ישראל דויד‬
‫מודל החלטה דינמי לקבלת כליה מן המת – בהתאם לפונקציית הכשל והתפלגות החיים של המועמד להשתלה‬
‫ ידועה בעיית הפער החמור בין ביקוש‬.‫ספיקת כליות סופית‬-‫השתלת כלייה חיה מהווה את הפיתרון הטוב ביותר למצב של אי‬
.‫ דווקא על רקע זה חשוב לגבש מדיניות הקצאה אופטימלית‬.‫להיצע במערכת ההשתלות המנוהלת ע"י מקבל ההחלטות הציבורי‬
‫ שכן מגיעים אליו היצעים‬,("‫ עומדת בעיית החלטה גם בפני "המועמד הבודד" )לצורך המחשה – זה שב"ראש התור‬,‫למעשה‬
.‫ עם חלוף הזמן חלה הרעה במצבו ועליו להיות פחות בררן‬.(‫ כתהליך בזמן )מעין בעיית מזכירה‬,‫בדרגות התאמה שונות‬
‫ שהיא כמובן קלה יותר לפיתרון אנליטי מבעיית‬,‫במאמר זה אנו מסתכלים על הבעייה אכן מנקודת מבטו של המועמד הבודד‬
‫ מנקודת הראות של מקבל ההחלטות הציבורי‬.‫ אנו אף מציעים בהתאם כלי תומך החלטה מבוסס אקסל‬.‫ההקצאה והעיתוי הכללית‬
‫ וכמו כן ככלי מחקרי‬,‫יוריסטית לניהול תור ההשתלות המלא‬-‫כלי זה הוא בעל חשיבות כ"אבן לגו" בבניין מדיניות אופטימלית‬
‫ התפתחה בשנים האחרונות‬,‫ במקביל‬.‫ לגנטיקה של האוכלוסיות המעורבות ועוד‬,‫ לעוצמת תהליך התרומה‬:‫לבדיקות רגישות שונות‬
18
‫ לפיה החולה נוטל חלק פעיל בהחלטות הנוגעות לו )ולפעמים שוכר לשם כך מומחה‬,"Patient Choice" ‫פרקטיקה רפואית של‬
.‫ מבוססת נתוני אמת ומודל אנליטי הסתברותי‬,‫ הכלי הנוכחי תומך בקבלת החלטות מושכלת שכזו‬.(‫חיצוני‬
‫ מבוסס תיכנון‬,‫ ואת מודל ההחלטה המתמטי הבסיסי‬,‫נציג את הגורמים הפיסיולוגיים והסטטיסטיים הרלבנטים להצלחת השתלה‬
‫ וכמובן את הפעלת‬,‫ נסקור את מקורות הנתונים ששימשו אותנו בדוגמאות ואשר מגולמים בחוברת העבודה של האקסל‬.‫דינמי‬
‫ מרכיב חדשני מן המאמר האחרון בנושא עוסק במידול ההסתברותי הדינמי של הדטריורציה של‬.(‫גיליון האקסל עצמו )המימשק‬
‫ דבר המאפשר גמישות‬,‫ משפחה זו מאופיינת בשני פרמטרים‬.‫ באמצעות משפחת התפלגויות החיים גאמה‬,(‫החולה )תחת דיאליזה‬
.‫בהתאמתה לנתונים סטטיסטיים לפי מאפיינים שונים ואוכלוסיות‬
:‫מבוסס על המאמרים‬
M. Bendersky and I. David, "The Full-Information Best-Choice Problem with Uniform or Gamma
Horizons", submitted to Optimization.
M. Bendersky and I. David, "Deciding Kidney-Offer Admissibility Dependent on Patients' Lifetime Failure
Rate", submitted to The European Journal of Operations Research.
Guy Wachtel, Amir Elalouf
Using a "Floating Patients" Approach for Improving Patients-Flow in Hospitals and
Emergency Departments
Overcrowding in hospitals along with long length of stay, high arrival rates, budget constraints, and
increasing demand for high service quality create challenges for the workflow and patient flow of hospital
emergency departments (EDs). From a managerial perspective, overcrowding can cause substantial profit
loss to the ED and the other departments. In order to prevent this profit loss, we assume that the hospital
management determines a maximal (fixed or dynamic) value for patients’ length of stay and for crowding
levels in the various departments, and that patients who cannot be evaluated in the ED in a timely fashion
are redirected for treatment in other hospital departments. The latter approach (referred to as the "floating
patient” method) has been practiced, for example, in Israel. This paper proposes two algorithmic approaches
that are designed to enable ED decision makers (specifically, in the triage) to optimally schedule evaluations
for patients who are waiting for treatment in the ED. The algorithms have been developed gradually and
embedded in a simulation model. To build the algorithms, we first solve a problem in which the triage
decision maker has full information on patients’ conditions and on how long their evaluations expect to
take. We then extend this problem to incorporate uncertainty as in real life scenarios: The triage decision
maker (physician) needs to carry out initial examinations to obtain information on the patient's situation
and, at each point in time, decides whether to continue to examine patients or to stop the process (halting
rule) and ”float” the remaining patients to other departments. Next, the physician determines the optimal
schedule for the full ED evaluations of the examined patients. We embed the algorithms into simulation
procedures and run simulations using empirical data collected from Bnei-Zion hospital in Israel. We
checked the method of "floating patients" when applied for different rates of crowding in the ED and the
other departments (i.e. the ED management can decide to use this method only in above 90% crowding in
the ED). The simulation took into consideration also the crowding situation in the other departments in
order to see the marginal effect of the "floating patients" method of the patients-flow in the hospital as a
whole system. Implementation of the "floating patient" method is shown to reduce patients' length of stay,
queues for beds in departments and the ED, and cumulative treatment time in the ED. These improvements
reflect a better balance of work-rate and crowding between the ED and the other departments.
19
Monday, 11.5
M1a: Combinatorial Optimization 2 Chair and Organizer: Rafi Hassin
Rafi Hassin, Uri Yovel
Sequential Scheduling on Identical Machines
We study a sequential version of the well-known KP-model: Each of $n$ agents has a job that needs to be
processed on any of $m$ machines. Agents sequentially select a machine for processing their jobs. The
goal of each agent is minimize the finish time of his machine. We study the corresponding sequential price
of anarchy for m identical machines under arbitrary and LPT orders, and suggest insights into the case of
two unrelated machines.
Keywords: sequential price of anarchy, machine scheduling, congestion games, load balancing, subgameperfect equilibrium, makespan minimization.
Tal Raviv, Uri Yovel, Tom Cherchy
New Integer Programming Formulation of Routing Problems in Dense Networks
A typical road network can be represented by a relatively sparse graph. However, in the arc based
formulations that are currently in use for many vehicle routing problems the underlining graph is extended
into a complete graph. Thus, one binary variable is defined for each pair of nodes. In this study, we
introduce a new modeling approach that exploits the sparsity of the networks. In the proposed models the
number of integer decision variables can be significantly reduced since the variables correspond only to
direct links between nodes. On the other hand the assumptions that each node should be visited exactly
once and each arc can be traversed at most once no longer hold. Our solution method thus starts by finding
some Eulaerian subsets of the graph that span the required set of nodes and may include an arbitrary number
of copies of each arc. The Eulerian circles in these sub-graphs are then reduced into a set of Hamiltonian
circles to derive an exact optimal solution. A similar approach can be used to obtain approximated solutions
where the accuracy of the solution is controlled by the number of unattractive arcs (respectively, integer
decision variables) that are deleted from the network (respectively, the model). The effectiveness of our
formulation when solved by a branch and cut algorithm is demonstrated by a numerical experiment. The
talk will be concluded with some adaptations of the approach for classic vehicle routing problems.
*We started this research project together with Uri short time before he got ill.
Dorit S. Hochbaum, Asaf Levin
Weighted Matching with Pair Restrictions
The weighted matroid parity problems for the matching matroid and gammoids are among the very few
cases for which the weighted matroid parity problem is polynomial time solvable. In this work we extend
these problems to a general revenue function for each pair, and show that the resulting problem is still
solvable in polynomial time via a standard weighted matching algorithm. We show that in many other
directions, extending our results further is impossible (unless P=NP). One consequence of the new
polynomial time algorithm is that it demonstrates, for the first time, that a prize-collecting assignment
problem with ``pair restriction" is solved in polynomial time. The prize collecting assignment problem is a
relaxation of the prize-collecting traveling salesman problem which requires, for any prescribed pair of
nodes, either both nodes of the pair are both matched or none of them are.
20
Enrique Gerstl, Gur Mosheiov
Single Machine Scheduling to Minimize Total Earliness-Tardiness with Unavailability
Period
We study several versions of a single-machine scheduling problem, where the machine is unavailable for
processing for a pre-specified time period. In the basic problem, a common due-date for all the jobs is
assumed, and the objective function is minimizing total earliness-tardiness. We consider first the setting
that no idle times are allowed. We then extend the problem to general earliness and tardiness cost functions,
to the case of job-dependent weights, and to the setting that idle times are allowed. All these problems are
known to be NP-hard. We introduce in all cases efficient pseudo-polynomial dynamic programming
algorithms.
M1b: Queues and Stochastic Systems 1 Chair: Galit Yom-Tov
Nitzan Carmeli, Haya Kapi, Avishai Mandelbaum
Modeling and Analyzing IVR Systems, as a Special Case of Self-services
Call centers play a very important role in today's economy, serving as the main customer contact channel
in many different enterprises. Call centers are highly labor intensive. Typically, 60%-70% of the overall
operating expenses of call centers are derived from agents’ employment costs. Reducing the number of
agents handling calls, without degrading service level, is thus of interest and importance. Enabling
customers self-service is one of the basic means for doing so, with Interactive Voice Response (IVR)
systems being one of the main self-service channels.
The goal of our research is to improve and enhance IVR systems, as a special case of self-service systems.
To do so, we model and analyze customer flow within an IVR system. The model features were established
and inspired by an Exploratory Data Analysis (EDA) of real IVR transactions in a call center of a large
Israeli bank, based on more than one year of data, including millions of calls.
An IVR system usually offers several services. Customers enter the IVR and then follow a series of menus
in order to reach a desired service or services. We represent the IVR system as a rooted tree and model
customer flow within it as a stochastic search. We model the search of group of customers, which is
commonly characterized by its perceived rewards and costs, its success probabilities, its service time
distribution, and its patience distribution. The goal of our search is to find the optimal path on the IVR tree,
which will result in maximal expected discounted revenue for customers within each group. We show that,
at each stage, an index can be assigned to each feasible option, and the optimal policy is to choose the
option with the highest index at each stage.
One of the main observations derived from the EDA was that some customers leave the IVR system without
getting any relevant information. These customers may either leave the call center or join the agents queue
(opt-out) to receive the desired service. In both cases, we say that these customers abandon the IVR service.
When customers are self-served, finding whether their service was successful or not is not an easy task.
The subject of identifying abandonments from self-service systems, such as IVR, is thus of interest and we
are addressing this issue in our work. We also discovered that there is a learning process, which means that
as customers gain more experience within the system, their response time is getting shorter. This fact was
incorporated into our model.
Our model enables the comparison between alternative IVR designs, both from the customer point of view
and from the enterprise point of view, thus supplementing existing research from other fields such as
Human-Factor-Engineering and Telecommunication Engineering. Although this research focuses on IVR
systems, we believe that both the theoretical model and some of the methods presented in our EDA can be
easily implemented to other self-service systems, which now become highly relevant, such as Internet
websites.
21
Zhenghua Long, Nahum Shimkin, Jiheng Zhang
Optimal Priority Control for Multiclass Many-server Queues with General Patience
Distributions
We consider the problem of server scheduling in an overloaded multiclass queueing system with multiple
homogeneous servers and customer abandonment. In the case of exponential reneging, an indexed priority
policy, called the cμ/θ rule, is known to be asymptotically optimal in the many-server heavy traffic regime,
where the arrival rates and number of servers increase proportionally. For general patience distributions,
we aim to find an asymptotically optimal control policy among the class of fixed priority rules. We first
describe a fluid model, which is known to be the limit of the stochastic system under priority rules, and
show its convergence to an equilibrium state in the case of exponential service. We then formulate an
optimization problem in terms of these equilibrium states, which leads to a nonlinear program of a certain
type which we term as the Fractional 0-1 Knapsack Problem. A dynamic programming algorithm is
developed to efficiently solve this new type of knapsack-like problem
Galit B. Yom-Tov, Jing Dong, Elad Yom-Tov
The Impact of Delay Announcements on Hospital Network Coordination and Waiting
Times
We investigate the impact of delay announcements on the coordination within hospital networks using a
combination of empirical observations and numerical experiments. We show that patients take delay
information into account when choosing emergency service providers and that such information can help
increase coordination in the network, leading to improvements in performance of the network, as measured
by Emergency Department wait times. Our numerical results indicate that the level of coordination that can
be achieved is limited by the patients’ sensitivity to waiting, the load of the system, the heterogeneity among
hospitals, and, importantly, the method hospital use to estimate delays. We show that delay estimators that
are based on historical average may cause oscillation in the system and lead to higher average waiting times
when patients are sensitive to delay. We provide empirical evidence that suggests that such oscillations
occurs in hospital networks in the US.
Eitan Bachmat
Geometric Queueing Theory
Queueing theory and PERT/CPM project management are two of the pillars of OR. In the talk we will
present an approximate analogue of the Pollaczek-Khinchine formula from queueing theory in the theory
of PERT/CPM, when the precedence relation have a geometric interpretation. This provides a vast
generalization of the context to which formulas of the Pollaczek-Khinchine type can be applied and leads
to a wealth of interesting new examples and problems. It also shows that PERT/CPM problems with a
geometric interpretation have an approximate modular symmetry. Our motivating example for the theory
is an analysis of the following airplane boarding policy which was recently implemented by a few airlines:
Passengers with no luggage for the overhead bins have boarding priority over passengers who do. We will
analyze such policies, show why they are not very efficient and how they can be made more efficient. The
key is to "cloak" slow passengers by constructing, thin focal lenses in space-time geometry.
M1: Game Theory 2 Chair and Organizer: Gail Gilboa Freedman
Shiran Rachmilevitch
Egalitarian-Utilitarian Bounds in Nash's Bargaining Problem
For every 2-person bargaining problem, the Nash bargaining solution selects a point that is “between” the
relative (or normalized) utilitarian point and the relative egalitarian (i.e., Kalai-Smorodinsky) point. Also,
it is “between” the (non-normalized) utilitarian and egalitarian points. I improve these bounds. I also derive
a new characterization of the Nash solution which combines a bounds-property together with strong
individual rationality and an axiom which is new to Nash's bargaining model, the sandwich axiom. The
sandwich axiom is a weakening of Nash's IIA.
22
Itai Arieli
Payoff Externalities and Social Learning
We consider a social learning model with payoff externalities in which one of two stage games is chosen at
random and then played repeatedly by a different group of agents in every period. The current ``generation''
of every period is informed of the history of actions chosen by the preceding generations and receives
conditionally independent private signals about the realized game. We show that with probability one, the
play converges to the set of equilibria of an appropriate convex combination of the two underlying stage
games. We identify a range of private signals, as a function of the stage games, for which asymptotic
learning holds. We show that in some cases, unlike the classical model with no payoff externalities,
asymptotic learning holds for a wide range of bounded private signals.
Gail Gilboa Freedman, Rann Smorodinsky
Axiomatization of Privacy
In many cases, conveying data of the individual agents is useful for achieving a particular objective. For
example, the participating of individuals in a clinical data base may be useful for clinical research, while
this data is sensitive from the individuals' perspectives. Thus, an inherent trade-off between effectiveness
and privacy prevails.
How much privacy is lost during a process? Can two systems be rigorously compared on the privacy loss
aspect? Can one propose a benchmark for privacy loss, possibly to be adopted as a standard?
The absence of unequivocal answers to those questions is conspicuous. Especially, since privacy-enhancing
technologies are intensively developed without a rigorous method for approximating their necessity nor
their marginal contribution.
A remarkable example for a privacy benchmark for handling statistical databases is the notion of
differential-privacy. In our study, we formalize the conceptual notion of privacy, and suggest an alternative
measure.
We view the privacy-preserving problem as a function, form the set of component-level-inputs into the set
of possible-outcomes/consequences. The inputs and outputs are often referred to as `secrets’ and `signals’,
respectively. By construction, each outcome is correlated with the individuals' inputs. As a result, it
jeopardizes privacy to some level. We aim to quantify this level, formally.
Our measure is inspired by the f-divergence, which is an information-theoretic quantity, associated with a
pair of probability distributions. We study a set of necessary and sufficient behavioral axioms which
uniquely define our measure and so we show that it is a natural construction.The applicability of our
research is for having a rigorous methodology for prioritizing systems by the levels of preserving their
privacy.
Gadi Fibich, Arieh Gavious
Revenue Equivalence of Large Asymmetric Auctions
Using asymptotic analysis, we calculate the seller's expected revenue in large asymmetric firstprice and second-price auctions, as well as in optimal auctions, with risk-neutral players. These
calculations show that the revenue difference between asymmetric first-price and second-price
auctions scales as ε2/n3, where n is the number of players and ε is the level of asymmetry
(heterogeneity) among the cumulative distribution functions of players' valuations. This scaling
law explains previous numerical findings that the revenue difference between first-price and
second-price auctions is extremely small even with as few as n=6 bidders, and shows that bidders'
asymmetry has a negligible effect on revenue ranking of large auctions. Furthermore, our
asymptotic calculations show that the revenue differences between asymmetric first- or secondprices auctions and the optimal mechanism also scale as ε2/n3. Hence, asymmetric first-price and
second-price are asymptotically optimal.
23
‫‪M1: OR Applications in the Israeli Police Chair and Organizer: Mali Sher‬‬
‫חיננית אפרתי‪ ,‬תהילה הירש‪ ,‬אירית נוביק‪ ,‬מלי שר‪.‬‬
‫ייעול עבודת הסיירים ביחידת המקומות הקדושים של משטרת‪-‬ישראל‬
‫משטרת ישראל מופקדת על אכיפת החוק‪ ,‬שמירת הסדר הציבורי וביטחון הפנים במדינת ישראל‪ .‬פרויקט זה התמקד בייעול עבודת‬
‫הסיירים ביחידת המקומות הקדושים‪ ,‬במרחב דוד במחוז ירושלים‪ .‬יחידת המקומות הקדושים אחראית‪ ,‬בין היתר‪ ,‬על אבטחת הר‬
‫הבית ושמירה על השקט בתחומו‪ .‬העבודה בהר הבית הינה קשה‪ ,24/7 ,‬וכוללת זקיפות בשערי הכניסה למתחם‪ ,‬לבוש הגנה מתאים‬
‫ונשיאת נשק ארוך לאורך כל שעות המשמרת‪ .‬העבודה הינה מורכבת ודרוכה‪ ,‬כהגנה על אזור רגיש ונפיץ‪ .‬כתוצאה מהעבודה‬
‫המורכבת והקשה מחד והקושי באיוש תפקידים אלו מאידך הוחלט על תגמול מיוחד לשוטרים המופקדים על משימה זו וכן על‬
‫הפעלת שיטת משמרות ייחודית‪ .‬האיוש כיום מתבסס בעיקר על שוטרים המתגוררים בצפון הארץ‪ .‬עובדה זו מקשה מאוד על שיבוץ‬
‫המשמרות ביחידה‪ .‬כך נוצר מצב ייחודי ביחידה בו משמרת רגילה של העובדים הרחוקים )מהצפון( הינה על בסיס ‪ 24‬שעות‬
‫למשמרת אחת ולאחריה ‪ 48‬שעות מנוחה‪ .‬יש לציין כי ביחידה משרתים מעט סיירים המתגוררים באזור ירושלים והמרכז‪ ,‬המועסקים‬
‫במשמרות רגילות‪ .‬הפרויקט בחן מספר היבטים‪ (1) :‬תכנון המשמרות השבועי ושיבוץ השוטרים למשמרות‪ (2) ,‬בחינת העומס על‬
‫השוטרים מבחינת סך שעות עבודת השוטרים ביחידה ו‪ (3) -‬בחינת מאפיינים אישיים של שוטרים‪ .‬בכך לתת למשטרה מידע מקדים‬
‫על איכות עבודת השוטר הצפויה לצורך קבלת החלטה מושכלת בנוגע לקבלתו לעבודה‪ .‬התוצאות שהתקבלו הנן )‪ (1‬מודל תכנון‬
‫לינארי בשלמים המשבץ מיטבית את השוטרים למשמרות עפ"י אילוצי "קו אדום" למשמרת‪ ,‬מגורי השוטרים ונהלי העבודה‬
‫במשטרת‪-‬ישראל‪ .‬פתרון המודל הביא לשיפור של ‪ 70%‬מהמצב כיום‪ (2) .‬שעות עבודת השוטר ביחידה נבחנו מול משטרות אחרות‬
‫בעולם ונמצא כי שוטרי היחידה עובדים מספר רב של שעות‪ ,‬בפער של עד כ‪ 20 -‬שעות שבועיות ממשטרות אחרות‪ .‬לפיכך‪ ,‬הומלץ‬
‫למשטרה לבחון מחדש את שעות עבודת השוטרים ביחידה‪ (3) .‬נמצא כי הנתונים האישיים המשפיעים על איכות עבודת השוטר הם‬
‫גיל‪ ,‬מצב משפחתי וותק במשטרה‪.‬‬
‫שירלי ארקוסין‪ ,‬אורטל סעידיאן‪ ,‬איריס פורמה‪ ,‬מלי שר‪.‬‬
‫השמת עובדים אופטימאלית‬
‫פרויקט זה מבוצע עבור אגף משאבי אנוש של המשטרה בשיתוף עם אגף התנועה הארצי במטרה לבצע השמה אופטימלית של‬
‫עובדים‪ .‬אגף משאבי אנוש של משטרת ישראל פועל להעצמת המשאב האנושי ‪ -‬איכותו‪ ,‬מקצועיותו ורווחתו‪ ,‬תוך התאמתו ליעדים‬
‫ולצרכים של המשטרה‪ .‬משטרת ישראל שמה לעצמה כמטרה לפעול להעצמת השוטרים והמשרתים בארגון והגברת הזדהותם עם‬
‫הארגון‪ .‬כיום תהליך שיבוץ השוטרים במשטרה נעשה ע"י אנשים – שוטרים באגף משאבי אנוש של המשטרה שאמונים על שיבוץ‬
‫עובדים לתפקידים השונים ע"פ המשרות הפנויות‪ .‬אופן השיבוץ הינו מלאכת מחשבת אנושית שמתייחסת להתניות‪ ,‬הכשרות‪ ,‬נתונים‬
‫ורצונות ע"פ שיקול דעת בריא‪ .‬מטרת הפרויקט הינה ייעול ושיפור תהליך שיבוץ השוטרים על ידי מידול הבעיה ומציאת פתרון‬
‫אופטימלי בהתאם לפרמטרים המוגדרים ותוך התחשבות בצרכי הארגון‪ ,‬רצון המועמד והתאמתו לתפקיד‪ .‬במסגרת הפרויקט בוצע‬
‫ניתוח של התפקידים המוצעים ע"י המשטרה‪ ,‬תוך זיהוי של התפקידים בעלי אופי דומה שיאפשרו העברה של שוטרים מתפקיד‬
‫לתפקיד‪ .‬בהתאם לכך נבנו שני מודלים לשיבוץ אופטימלי של העובדים לתפקידים בין תחנות שונות‪ :‬האחד ‪ -‬מודל תכנון ליניארי‬
‫ הפועל לשיבוץ השוטר לתפקיד תוך צמצום המרחק בין מקום מגורי העובד לתחנה בה הוא משרת וצמצום הפער בין התקן הארצי‬‫של התפקיד )מוגדר ע"י המשטרה( לבין תקן המצבה )מספר השוטרים המשובצים לתפקיד בפועל(‪ ,‬נכתב בתוכנת ‪IBM ILOG‬‬
‫‪ .CPLEX‬והשני ‪ -‬מודל שידוך יציב ‪ -‬הפועל לשיבוץ תוך התייחסות לרצון‪/‬העדפת העובד )לפי שרירות ליבו( ורצון‪/‬העדפת‬
‫המשטרה )שמיוצגת ע"י לוגיקה לצמצום המרחקים בין תחנת המשטרה למקום מגורי העובד(‪ .‬המודל מיישם את אלגוריתם גייל‪-‬‬
‫שפלי לשידוך יציב )‪ ,(Matching Problem‬נכתב בתוכנת ‪ .MATLAB‬עיקרי הפתרון המוצע כוללים ניתוח של הנתונים‬
‫שהתקבלו ממשטרת ישראל‪ ,‬שימוש בנתונים אלו לצורך הפעלת המודלים‪ ,‬וניתוח פלטי המודלים בכדי להעריך יעילות ואפקטיביות‬
‫השיבוץ החדש בהשוואה לשיבוץ שמתבצע כיום במשטרה מה שתורם להתייעלות הארגון‪.‬‬
‫אלי אורבך‪ ,‬ינון סבתו‪ ,‬מיכאל דרייפוס‪ ,‬מלי שר‪.‬‬
‫ייעול פריסת מצלמות האכיפה במשטרת‪-‬ישראל‬
‫כחלק מהחלטות הממשלה למאבק בתאונות הדרכים‪ ,‬תוקצב פרויקט א‪) 3‬אכיפה אלקטרונית אוטומטית( במשטרת ישראל‪ ,‬באגף‬
‫התנועה‪ .‬פרויקט זה החל לפעול מבצעית בשנת ‪ 2012‬וכולל כיום כ‪ 150 -‬עמדות לאכיפת מהירות מעל המותר ומעבר צומת באור‬
‫אדום‪ .‬העמדות ממוקמות בתחום העירוני והבינעירוני‪ ,‬בצמתים ובקטעי הדרך‪ .‬כיום מספר המצלמות קטן ממספר העמדות והמצלמות‬
‫מנוידות בין העמדות באופן אקראי‪ .‬מחקר זה מטרתו לפתח מודל לשיבוץ מצלמות האכיפה במיקום ובזמן המיטביים לצמצום מרבי‬
‫של תאונות הדרכים‪ .‬המודל מתבסס על נתוני תאונות הדרכים בכל רחבי הארץ בשנים האחרונות‪ .‬בשלב ראשון של המחקר בוצע‬
‫ניתוח נתוני כל תאונות הדרכים ב ‪ 5‬שנים האחרונות ומיקומם ביחס למיקום עמדות א‪ .3‬בשלב השני של המחקר בוצע חיזוי מספר‪,‬‬
‫חומרה ומיקום של תאונות הדרכים לתקופת התכנון‪ .‬בשלב השלישי‪ ,‬בהינתן התחזית ונתינת "ציונים" לכל תאונה עפ"י חומרתה‪,‬‬
‫נקבע מיקום המצלמות האופטימלי‪ ,‬ע"פ המודל תוך התחשבות במספר אילוצים‪.‬‬
‫‪24‬‬
Mali Sher, Nicole Adler, Shalom Hakkert.
A Traffic Enforcement Camera Operational Model.
Police enforcement resources impact safety levels by changing driver behavior. The existence of an
enforcement camera reduces the number and severity of offences which in turn reduces the number of road
accidents and serious injuries. After measuring the impact of a set of cameras over the last year in Israel,
we note that traffic parameters were decreased on average. Based on the enforcement camera recordings,
tickets are issued for red-light and speed offences. The owner of the vehicle is either (i) sent a fine, (ii) a
fine with points or (ii) a court summons, according to the severity of the offence and the available
police/court resources. The time halo effect causes a camera’s productivity to be reduced over time, once
the tickets issued have been served, and it is therefore worthwhile moving the camera after a period of time.
This research investigates the traffic police enforcement policy with respect to the use of semi-fixed
cameras on a road network. In the first stage, a public committee choose the location of a set of fixed camera
poles. Subsequently, the police force decided on a monthly basis which poles will contain active cameras.
In the final stage, the decision is made as to the most appropriate operational policy such that a specific
speed threshold determines the issuing of tickets. An integer linear program model was developed to
determine the lowest enforcement speed per camera site over the planning period such that ticket issuances
are maximized according to an analysis of the most important traffic parameter (average speed, variance
etc.) at each site. The constraints are budget constraints as well as limitations on the processing capabilities
of the police back office and the judicial system. There are three administrative level restrictions: the police
back-office producing the tickets, a separate unit handling requests from the car owners and the courts that
are limited by the number of judges. Finally, a municipal limitation restricts the number of tickets issued
within a specific geographic region. The results of a year-long study show that the number of offences was
reduced by up to 50% wherever cameras operated, highlighting the importance of this tool in the traffic
enforcement field. In addition, we found that the time halo effect exists for approximately two months
which will impact future decisions and the frequency with which cameras will be relocated.
M4a: Optimization and Stochastics, Prize Winners Session
Chair:Moshe Haviv
Amir Beck, Nadav Hallak: Uriel Rothblum Award for 2015
On the Minimization Over Sparse Symmetric Sets: Projections, Optimality Conditions and
Algorithms
In this paper we consider the problem of minimizing a general continuously differentiable function
over symmetric sets under sparsity constraints. These type of problems are generally hard to solve as the
sparsity constraint induces a combinatorial constraint into the problem, rendering the feasible set to be
nonconvex. We begin with a study of the properties of the orthogonal projection operator onto sparse
symmetric sets. Based on this study, we derive efficient methods for computing sparse projections under
various symmetry assumptions. We then introduce and study three types of optimality conditions: basic
feasibility, L-stationarity and coordinate-wise optimality. A hierarchy between the optimality conditions is
established by using the results derived on the orthogonal projection operator. Methods for generating
points satisfying the various optimality conditions are presented, analyzed, and finally tested on specific
applications.
Yair Shaki (joint work with Josh Reed): Uriel Rothblum Award for 2015
A Fair Policy for the Servers in the G/GI/N Queue
We consider the G/GI/N queue with multiple server pools, each possessing a pool-specific service time
distribution. We then consider the class of non-idling routing policies referred to as u-greedy policies. These
policies route incoming customers to the server pool with the longest weighted cumulative idle time. Our
first set of results demonstrate that asymptotically in the Halfin-Whitt regime and under any u-greedy
policy, the diffusion scaled cumulative idle time processes of each of the server pools are held in fixed
proportion to one another. We next move on to providing a heavy-traffic limit theorem for the process
25
keeping tracking of the total number of customers in the system. Our limit may be characterized as the
solution to a stochastic convolution equation.
Liron Ravner: Abraham Mehrez Award for 2015
Equilibrium Arrival Times to a Queue with Order Penalties
Suppose customers need to choose when to arrive to a congested queue with some desired service at the
end, provided by a single server that operates only during a certain time interval. We study a model where
the customers incur not only congestion (waiting) costs but also penalties for their index of arrival.
Arriving before other customers is desirable when the value of service decreases with every admitted
customer. This may be the case for example when arriving at a concert or a bus with unmarked seats or
going to lunch in a busy cafeteria. We provide game theoretic analysis of such queueing systems with a
given number of customers, specifically we characterize the arrival process which constitutes a symmetric
Nash equilibrium.
Arik Sadeh
Choosing Binary Factors Affecting Purchasing in e-Commerce
Purchase intention of potential buyers in e-commerce may be affected by numerous factors. The business
and ethical conduct of a given e-store has an important role on purchase intention. In this study, binary
factors are used to describe the e-store characteristics. Naturally, not all possible combinations of k factors
(2k) can be included in one questionnaire. An algorithm was developed to help choosing relevant
combinations of level of factors to be included in questionnaires. In this study, success is defined as a
potential consumer intends to purchase a given product. Having results from the survey, a second algorithm
is used to determine what values of factors that are included in a combination would lead to better success
rate. This suggested mechanism can be applied in various applications in engineering, economics and
management.
M4: Scheduling Chair and Organizer: Dvir Shabtay
Tal Grinshpoun, Elad Shufan, Hagai Ilani
Public Transportation with Flexible Timetable
In a regular bus service, every bus line has a given route and a published timetable. Passengers who know
in advance the stations along the bus's route and the expected schedule plan their ride correspondingly.
Another model of public transportation is the DARP (Dial A Ride Problem) model where, in opposition,
the passengers' requests for traveling are known in advance and the bus's route and schedule are planned
accordingly. We hereby present a transportation model, which is a mix of the two mentioned models. The
route of the bus is known in advance but the timetable is set according to the passengers' requests. For a
given operation cost, the aim is to maximize user satisfaction, by minimizing the sum of passengers' waiting
times. We introduce algorithms for solving two variants of the fixed route DARP -- one for a fleet of infinite
capacity vehicles, and one for the more general case of vehicles with heterogeneous capacities. Contrary to
general DARP, which is NP-Hard, the presented algorithms are polynomial in the number of ride requests.
Enrique Gerstl, Gur Mosheiov
Scheduling with Two Competing Agents to Minimize Total Weighted Earliness
We study a single machine scheduling problem with two competing agents and earliness measures. Given
a common deadline for all the jobs of both agents, the objective function is minimizing the total weighted
earliness of the first agent, subject to an upper bound on the maximum earliness of the jobs of the second
agent. This problem was recently proved to be NP-hard, leaving the question of the complexity class open.
We introduce a pseudo-polynomial dynamic programming algorithm, implying that the problem is NP-hard
in the ordinary sense. An extensive numerical study indicates that the dynamic programming is very
effective for solving medium size instances. We also propose an efficient heuristic, which is shown
numerically to produce very close-to-optimal schedules. The dynamic programming algorithm is extended
26
to any (given) number of agents, proving NP-hardness in the ordinary sense of the general multi-agent
setting. Finally, we study the inverse problem of minimizing the maximum earliness of one agent subject
to an upper bound on the maximum weighted earliness of the second agent. We introduce a pseudopolynomial dynamic programming algorithm, a simple greedy-type heuristic and a lower bound. Our
numerical tests verify that the heuristic produces very small optimality gaps.
Baruch Mor, Gur Mosheiov
Minimizing Maximum Earliness and Minimizing the Number of Early Jobs on a
Proportionate Flowshop
A proportionate flowshop is a special case of the classical flowshop, where the job processing times are
machine-independent. Most classical scheduling objective functions have been studied in the context of a
proportionate flowshop. In most cases, the solution was shown to be identical to that of the single machine
version. We introduce two rare cases where the extension to a proportionate flowshop leads to different
solutions. Specifically, we study the problems of minimizing maximum earliness and minimizing the
number of early jobs. We show that the problems remain polynomially solvable and introduce algorithms
that guarantee an optimal solution in O(n ) and O(n ) time, respectively, where n is the number of jobs.
Dvir Shabtay, Shlomo Karhi
An Asymptotically Optimal Online Algorithm to Minimize the Total Completion Time on
Two Multipurpose Machines with Unit Processing Times
In the majority of works on online scheduling on multipurpose machines the objective is to minimize the
makespan. We, in contrast, consider the objective of minimizing the total completion time. For this purpose,
we analyze an online-list scheduling problem of n jobs with unit processing times on a set of two machines
working in parallel. Each job belongs to one of two sets of job types. Jobs belonging to the first set can be
processed on any of the two machines while jobs belonging to the second set can only be processed on the
second machine. We present an online algorithm with a competitive ratio of ρ(LB) + O
, where ρ(LB)
is a lower bound on the competitive ratio of any online algorithm and is equal to 1 +
where α = + 116 − 6√78
asymptotically optimal.
/
+
√
× /
√
,
/
≈ 1.918. This result implies that our online algorithm is
M4: Optimization Chair: Gideon Weiss
Saleh Soltan, Mihalis Yannakakis, Gil Zussman.
Joint Cyber and Physical Attacks on Power Grids: Graph Theoretical Approaches for
Information Recovery
Recent events demonstrated the vulnerability of power grids to cyber and physical attacks. Therefore, we
focus on joint cyber and physical attacks and develop methods to retrieve the grid state information
following such an attack. We consider a model in which an adversary attacks a zone by physically
disconnecting some of its power lines and blocking the information flow from the zone to the grids' control
center. We use tools from linear algebra and graph theory and leverage the properties of the power flow
DC approximation to develop methods for information recovery. Using information observed outside the
attacked zone, these methods recover information about the disconnected lines and the phase angles at the
buses. We identify sufficient conditions on the zone structure and constraints on the attack characteristics
such that these methods can recover the information. We also show that it is NP-hard to find an approximate
solution to the problem of partitioning the power grid into the minimum number of attack-resilient zones.
However, since power grids can often be represented by planar graphs, we develop a constant
approximation partitioning algorithm for these graphs. Finally, we numerically study the relationships
between the grid's resilience and its structural properties, and demonstrate the partitioning algorithm on real
power grids. The results can provide insights into the design of a secure control network for the smart grid.
27
Renata Poznanski, Refael Hassin.
Optimal Multi-Period Network Flows with Coupling Constraints
Suppose we have two networks defined on duplicates of the same directed graph, with given source and
sink nodes, but with different edge capacities. Our problem is to compute feasible flows such that the sum
of flows is maximized subject to coupling constraints that force identical flows on duplicate copies of the
same edge for a subset of edges. When the subset consists of a single edge, we prove an existence of an
integral optimal solution and provide an efficient algorithm. For multiple edges we prove that there exists
an integral solution for all possible capacity functions if and only if the graph is series parallel. We also
consider variations to other minimum cost flow problems.
Rafiq Mansour, Yair Censor.
New Douglas-Rachford Algorithmic Structures and Their Convergence Analyses.
We study new algorithmic structures for the Douglas-Rachford (DR) algorithm to solve convex feasibility
problems. For a finite family of closed convex sets in a Hilbert space, with nonempty intersection, the
convex feasibility problem (CFP) is to find an element in the nonempty intersection of all the sets. There
are many algorithms in the literature for solving CFPs, in particular, two algorithmic structures that
encompass many specific feasibility-seeking algorithms are the String-Averaging Projections (SAP)
method and the Block-Iterative Projection (BIP) method. We do two things: (i) create new algorithmic
structures with the 2-sets-DR algorithmic operator, and (ii) define and study an “m -sets-DR operator”. First
we employ the two-sets-DR algorithmic operator and embed it in the SAP and BIP algorithmic structures.
In doing so we obtain two new families of DR algorithms, of which the two-sets-DR original algorithm and
the recent cyclic-DR algorithm are special cases. Finally, we propose and investigate a generalization of
the DR algorithmic operator itself. We propose to allow the algorithmic operator to perform a finite number
of consecutive reflections into the sets and only then take the midpoint between the current iterate and the
end-point of the consecutive reflections as the next iterate. We show how this “m -sets-DR operator” works
algorithmically. We study the convergence of all three algorithmic schemes by using properties of strongly
quasi-nonexpansive operators and firmly nonexpansive operators.
Evgeny Shindin, Gideon Weiss
A simplex-type Algorithm for Continuous Linear Programming
We consider continuous linear programs over a continuous finite time horizon T, with a constant coefficient
matrix, linear right hand side functions and linear cost coefficient functions, where we search for optimal
solutions in the space of measures or of functions of bounded variation. These models generalize the
separated continuous linear programming models and their various duals, as formulated in the past by
Anderson, by Pullan, and by Weiss. In previous papers we have shown that these problems possess optimal
strongly dual solutions. We also have presented a detailed escription of optimal solutions and have defined
a combinatorial analogue to basic solutions of standard LP. In this paper we present an algorithm which
solves this class of problems in a finite bounded number of steps, using an analogue of the simplex method,
in the space of measures.
M4 Queues and Stochastic Systems 2
Chair and Organizer: Yoav Kerner
Yoav Kerner, Opher Baron
Queueing Model for Safety Stock Inventory Model with Perishable Items and General
Distribution
We consider and inventory model in which the supply always regenerates the inventory to a given level.
The lead time of the supply is random and hence an order is placed whenever the inventory goes below a
predefined level. In addition, the inventory has a perish time, independent on everything else. We analyze
the steady state distribution of the inventory, the queue, and the remaining lead time for the case where the
perish times and/or the lead times follow a general distribution.
28
Ruth Sagron, Gad Rabinowitz, Israel Tirkel
Hybrid Simulation-Regression Approximation for Tandem Queues with Downtime Events
In this work, we introduce the Hybrid Simulation-Regression method for approximating the class-departure
variability in tandem queues with downtimes. Existing decomposition methods reach partial success, due
to the non-renewal process caused by the downtime events. Analytic approximations lack accuracy, and
one that combines simulation tools as well, requires very high computation efforts. The proposed hybrid
method paves a new approach for both reducing efforts and improving accuracy, by integrating existing
decomposition methods with two variability functions. One depends on downstream traffic intensity for the
within-class effect, and another on a departure from a single queue and downstream queue's waiting-time
for the between-class effect. This method enables modeling different policies of downtimes (e.g. FCFS,
Priority). Numerical experiments demonstrate relative errors about four times smaller than existing analytic
procedures, and two times smaller than another method that combines analytic and simulation tools.
Yonit Barron
Clearing Control Policies for MAP Inventory Process with Partially Satisfied Demand
We consider a production/clearing process in a random environment where a single machine produces a
certain product into a buffer continuously. The demands arrive according to a Markov Additive Process
(MAP) governed by a continuous-time Markov chain, and their sizes are independent and have phase-type
distributions depending on the type of arrival. Negative inventory is not allowed, thus, the demand is
partially satisfied. The production process switches between predetermined rates which depend on the state
of the environment. In addition, the system is totally cleared at stationary renewal times and staring anew
at level zero immediately. Several clearing policies are considered; clearing at random times, clearing at
crossings of a specified level and a combination of the above policies. We assume the total cost includes a
fixed clearing cost and variable holding and lost demand costs. By applying the regenerative theory, we use
tools from exit-time theorem for fluid process and martingales to obtain cost functionals under both the
discounted and average criterion. Finally, illustrative examples and a comparative study are provided.
Tal Avinadav, Tatyana Chernonog, Yael Lahav, Uriel Spiegel
Dynamic Pricing and Promotion Expenditures in an EOQ Model of Perishable Items
This study considers dynamic decisions of a retailer with regard to the selling price and promotion
expenditures associated with a perishable item. We propose an EOQ model in which the retailer faces a
general demand function that is separable into multiplicative components of selling price, items' age and
promotion expenditure. We find analytical expressions for the optimal price and promotion trajectories;
and we show that the former increases and the latter decreases in the items' age, and that both are
independent of the cycle length. Moreover, we show that the selling price is independent of the promotion
expenditure, but not vice versa. It is also proved that under these optimal trajectories, the profit rate is
strictly pseudo-concave in the cycle length. A comparison between dynamic and stationary strategies is
given.
M5a: Transportation
Chair and Organizer: Tal Raviv
Mor Kaspi, Tal Raviv, Michal Tzur
Regulating One-Way Vehicle Sharing Systems through Parking Reservation Policies
One-way vehicle sharing systems allow users to rent vehicles in one of many unmanned stations scattered
in the city, use them for a short ride and return them at any station. The demand processes for vehicles and
parking spaces are typically unbalanced. Consequently, shortages in vehicles or parking spaces may occur
in some of the stations along the day. We propose implementing passive regulations as means for redirecting
the demand so as to improve the performance of the system. In particular, we focus on parking reservation
policies. Under such policies, the users may be required, upon renting a vehicle, to reserve a parking space
at their destination and these spaces are kept for them. We measure the performance of the vehicle sharing
system in terms of the total excess travel time users spend due to shortages of vehicles or parking spaces.
We formulate a Markovian model of the system and use it to compare two extreme policies: a complete
29
parking reservation (CPR) policy under which all users are required to reserve a parking space and a
baseline policy entitled no-reservation (NR). Through this model, we prove that under realistic demand
rates, the CPR policy outperforms the NR policy.
We also devise mathematical programming based bounds on the total excess travel time under any passive
regulation and in particular under any parking space reservation policy. These bounds are compared to the
performance of: the above policies, several partial parking reservation policies and a utopian parking space
overbooking policy. A detailed user behavior model for each policy is presented and a discrete event
simulation is used to evaluate the performance of the system under various settings. The analysis of two
case studies of real-world systems demonstrates that: (1) a significant improvement of what theoretically
can be achieved is obtained by the CPR policy. (2) The performance of the proposed partial reservation
policies monotonically improve as more reservations are required. (3) Parking overbooking is not likely to
be beneficial. In conclusion, our results demonstrate the effectiveness of the simple CPR policy and suggest
that parking space reservations should be used in practice, even if only a small share of the users are required
to place reservations
Sharon Datner, Tal Raviv, Michal Tzur
Setting Inventory Levels in Bike-Sharing Networks
Bike-sharing systems allow people to rent a bicycle at one of many automatic rental stations scattered
around a city, use them for a short journey, and return them at any other station in that city. A crucial factor
in the success of such a system is its ability to meet the fluctuating demand for both bicycles and vacant
lockers at each station. Setting the inventory levels of each station is a complicated task, due to the nature
of users' behavior. If bicycles are not available at the desired origin of a user's journey, the user may either
abandon the system, possibly use other means of transportation, or she may look for available bicycles in a
neighboring station. If, on the other hand, a locker is not available at the destination, the user is obliged to
find a station with available space in order to return the bicycle to the system. In this study we introduce a
method to determine the inventory levels, who considers the interaction described between neighboring
stations in the bike sharing network. Using a simulation based guided local search, we set inventory levels
that would improve the systems' quality of service.
Hila Hindi-Ling, Hillel Bar-Gera, Arie Sachish
The effect of a quayside cranes buffer on ships unloading process
Container terminals are essential intermodal interfaces in the global transportation network. Efficient
container handling at terminals is important for reducing transportation costs and keeping shipping
schedules. The temporary storage of the inbound and outbound containers is one of the most important
services at the container terminal. The storage area in the terminal is divided into the several blocks of
containers. Each block consists of a number of side by side lanes. Each lane is formed by a row of container
stacks. Every stack may hold up to 4–5 tiers of containers (one on top of the other). The fast storage and
retrieval of containers at the blocks is essential for the economic performance of container terminals and
also shipping companies. These issues affect directly the traffic of the handling equipment and consequently
the dwell and turnaround time of vessels. In the unloading phase, containers are unloaded and transported
from the vessel to the storage yard. The equipment involved generally includes quay cranes (QC) at berths,
yard cranes at storage yards, and terminal trucks. The QCs are in charge of lifting and moving containers
from the vessels to the trucks. A number of trucks travel in a dedicated closed loop to pick up containers at
the berth and drop them at the storage yard. Each truck usually handles one container premises at a time.
Once the container arrives at the yard, a yard crane lifts it and stores it inside the yard. Delays can occur if
trucks are queued at the berth and/or the yard, depending on the number of available cranes and the arrival
rate of the trucks. This presentation will examine the potential of a container buffer between the QC and
the terminal trucks. We will present a quantitative analysis of the effect of buffer size, fleet size, and
operation rate parameters on queues, delays, overall vessel dwell time, and total operation costs.
30
M5b: Water Management
Chair: Amos Bick
Amos Bick, Ioannis K. Kalavrouziotis, Gideon Oron
Modeling Combined Technological, Environmental and Economic Considerations in
Domestic Sludge Reuse via the Analytic Hierarchy Process (AHP)
An optional method of using recycling and reuse efficiently of sludge is presented. Sludge can be originated
from a number of sources, although the one obtained from wastewater is the most common one. It composes
around 30% percent of the wastewater treated by activated sludge or other similar methods. It always creates
problems related to the environmental. A multi-objective function is defined. It consists of treating the
sludge by four alternatives: (i) disposal of the sludge untreated to the sea; (ii) incineration of sludge for
energy generation; (iii) treatment of the sludge via anaerobic digestion, and; (iv) using the treated sludge as
a soil amendment, primarily for agriculture. Comparisons of the treatment and disposal methods were made.
Comparison is based on the Analytic Hierarchy Process (AHP) model and Total Order Ranking Methods
(Absolute Reference, Dominance Functions, and Hasse Average Ranking). The comparison criteria
included similarly also five criteria. These criteria included the cleanness of the environment
(environmental criteria), the cost of treatment according to the final product, benefit to society in terms of
energy generation, benefits to agriculture in terms of food production and public acceptance. There is no
doubt that the solutions will be location dependent however, they demonstrate the options in each region.
David Raz, Ariel Daliot
A Generic Modeling Language for Water Supply Systems Optimization
We look at modeling Water Supply Systems (WSS) systems for the purpose of optimizing energy costs.
Energy costs are responsible for more than 90% of the operational costs of such systems and as such are
the most important factor governing WSS operation. The major constraints for such an optimization are
water volume constraints such as water supply and demand, tank minimum and maximum water levels etc.
This is in contrast to Water Distribution Systems (WDS) which may also be governed by water pressure
constraints. Existing modeling tools, such as the EPAnet software, are focused on WDS and as such are
very complex, focusing on the physical properties of the system (pipes, valves etc.) rather than on water
flow and electrical power consumption and costs.
We propose a simple generic modeling language for WSS. Although simple and generic, in our experience
it suffices for describing WSS in enough details for energy cost optimization. The language has the major
benefit that it easily translates into an LP model for optimizing energy costs.
We describe the language and a graphic tool for building WSS models. We demonstrate how real life
constructs translate easily into the modeling language and into an LP model. We discuss how the model
may be efficiently solved. We show how the language may be easily expanded to incorporate other
constraints such as pressure and other operational constraints.
We express our hope that such a modeling language may be used to share network models by interested
researchers and enrich the OR community.
Beni Lew, Olga Tarnapolski, Vladimir Yudachev, Amos Bick
Membrane Treatment of Brackish Groundwater for Unrestricted Use for Irrigation and
Sustainable Agricultural Production: Decision Analysis via The Hasse Diagram Technique
(HDT)
Field experiments are in progress for brackish groundwater upgrading for unrestricted use for irrigation and
sustainable agricultural production at the Arava valley (Israel). The treatment system is based on
implementing of potable water and two main treatment stages: Nanofiltration and Reverse-Osmosis
membrane processes. The feed are subsequently applied for irrigation of pepper crops. The comparison
criteria include pepper yield and pepper quality. Data analysis is based on Hasse Diagram Technique
(HDT), that is the application of the partial order theory referring to the objects, and characterized by vector
based quantities with an easy visualization of the obtained results. This study demonstrates the possibility
and appropriateness of providing a systematical decision making framework with several characteristics:
31
(i) different technological performances can be evaluated using multiple attributes - both quantitative and
qualitative - rather than profitability alone, (ii) the use of ratings makes it possible to evaluate the
applicability of different options for the end user, (iii) the HDT is a useful tool with an easy visualization
of the obtained results, and; (iv) the proposed approach forms the basis for a continuous process of planning
and managing technology selection, so that the priorities of the technologies can easily be modified and
updated.
M5c: Strategic Behavior in Queues Chair and Organizer: Moshe Haviv
Moshe Haviv, Binyamin Oz
Self-Regulation of a Queue via Random Priorities
We consider an unobservable M/M/1 queue where customers are homogeneous with respect to their reward
(due to service completion) and with respect to their cost per unit of time of waiting. Left to themselves, it
is well known that in equilibrium they will join the queue at a rate that is higher than it is socially optimal.
Hence, regulation schemes, under which the resulting equilibrium joining rate coincides with the socially
optimal one, should be considered. In this talk we suggest a classification of regulation schemes, based on
a few desired properties, and use it to classify schemes from existing literature. To the best of our
knowledge, there is no existing scheme that possesses all properties, and in this talk we suggest such one.
This novel scheme is based on assigning random priority to each customer, prior to the decision whether or
not to join. We also introduce variations of this regulation scheme as well as additional schemes based on
randomization.
Moshe Haviv, Liron Ravner
Accumulating Priority Queue with Strategic Customers
In an accumulating priority M/G/1 queue each customer is assigned a positive priority coefficient. This
assignment can be either class dependent or a choice made by the customers themselves. The actual
(accumulated) priority of a waiting customer is a linear function of his time since arrival whose slope
coincides with his priority parameter. The service regime is such that upon service completion the next
customer to be admitted is the one who has accumulated the most priority. We study a non-cooperative
game where customers can purchase their own priorities. For the case of homogeneous customers with
respect to their waiting cost parameter, we explicitly compute the unique pure strategy Nash equilibrium.
We further show that this model can display both avoid the crowd and follow the crowd behaviour, for
different levels of bidding. For a game with heterogeneous customer types that differ in their waiting costs,
we characterize the Nash equilibrium as the solution to a set of polynomial equations and suggest using an
iterated best response dynamics in order to compute it. We further consider a non-atomic version of this
game where each class of customers can coordinate their bids. We construct a unique pure strategy Nash
equilibrium for the resulting game.
Nahum Shimkin
What to (Truthfully) Tell Customers to Make Them Join a Queue
We consider a service system, such as a queueing system, to which customers arrive sequentially. Upon
arrival, each customer receives from the system manager some information about his or her expected quality
of service (for example, the expected waiting time in the queue, based on the current queue size which is
unobservable by the arriving customer), and may then decide whether to balk or join the system. The
manager is committed to truth telling, but can provide partial information (e.g., a range of possible waiting
times, which must include the true one). We ask what information should be provided to arriving customers
to maximize the throughput, namely the fraction of customers that choose to join. This question is
formulated as an optimization problem, in terms of the service demand curve and the probability
distribution of the service quality. Concrete solutions are derived, whose form depends on the convexity
or concavity properties of the demand curve.
32
Name
List of Participants: Speakers and Chairs*
Email
Affiliation
Amid David
Anily Shoshana
Arieli Itai
Atar Rami
Avinadav Tal
Bachmat Eitan
Barron Yonit
Bendavid Illana
Ben-Tal Aharon
Bick Amos
Buchbinder Niv
Carmeli Nitzan
Chanukov Gabi
Chernonog Tatyana
Cohen Yaarit.M.
Cohensius Gal
Datner Sharon
David Israel
Dreyfuss Michael
Eisenhandler Ohad
Elalouf Amir
Eliazar Iddo
Gavious Arieh
Gerchak Yigal
Gerstl Enrique
Gilboa Freedman Gail
Goldberg Noam
Halman Nir
Hanany Eran
Hassin Rafi
Haviv Moshe
Herer Yale
Hindi-Ling Hila
Ilani Hagai
Kaspi Haya
Kaspi Mor
Kerner Yoav
Levin Asaf
Lipetz Vladimir
Long Zhenghua
Mansour Yishay
IBM Research
[email protected]
Tel Aviv Univ.
[email protected]
Technion
[email protected]
Technion
[email protected]
Bar-Ilan Univ.
[email protected]
Ben Gurion Univ.
[email protected]
Ariel Univ.
[email protected]
Ort Braude
[email protected]
Technion
[email protected]
Bick & Assoc.
[email protected]
Tel Aviv Univ.
[email protected],
Technion
Bar-Ilan Univ.
[email protected]
Bar Ilan Univ.
[email protected]
Technion
[email protected]
Ben Gurion Univ.
[email protected]
Tel Aviv Univ.
[email protected]
Ben Gurion Univ.
[email protected]
Jerusalem College of
Technology
[email protected]
Tel Aviv Univ.
[email protected]
Bar-Ilan Univ.
[email protected]
Tel Aviv Univ.
[email protected]
Ben Gurion Univ.
[email protected]
Tel Aviv Univ
[email protected]
Hebrew Univ.
[email protected]
Technion
[email protected]
Bar-Ilan Univ.
[email protected]
Hebrew Univ.
[email protected]
Tel Aviv Univ
[email protected]
Tel Aviv Univ.
[email protected]
Hebrew Univ.
[email protected]
Technion
[email protected]
Ben Gurion Univ.
[email protected]
Shamoon College.
[email protected]
Technion
[email protected]
Tel Aviv Univ.
[email protected]
Ben Gurion Univ.
[email protected]
Technion
IBM Research
[email protected]
Technion
[email protected]
Tel Aviv Univ.
33
Session
S5c
S2b
M1c
S4
M4d
M1b
M4d
S5d
M3, S1*
M5b, M5b*
S2d
M1b
S2a
S5b
S2d
S2b
M5a
S5d
S2d
S2c
S2a, S5d*
M2
M1c
S5b
M4b
M1c, S4*
S2b
S5a
S4, S2a*
M1a, M1a*
M5c, M4a*
S5d
M5a
M4b
S4*
M5a
M4d, M4d*
M1a
S5c
M1b
S3
Mansour Rafiq
Megiddo Nimrod
Masin Michael
Miller Eliaz
Mor Baruch
Mosheiov Gur
Naor Seffi
Noham Reut
Oz Binyamin
Pauwels Edouard
Perel Nir
Perlman Yael
Poznanski Renata
Rachmilevitch Shiran
Raviv Tal
Ravner Liron
Raz David
Sabach Shoham
Sadeh Arik
Sagron Ruth
Shabtay Dvir
Sher Mali
[email protected]
Haifa Univ.
[email protected]
IBM Almaden
[email protected]
IBM Research
Shimkin Nahum
Shindin Evgeny
Shtern Shimrit
Sinuany-Stern Zilla
Smirnov Dina
Tzur Michal
Wachtel Guy
Weiss Gideon
Yechiali Uri
Yom-Tov Galit
Zeevi Assaf
Zussman Gil
‫אורבך אלי‬
‫אפרתי חיננית‬
‫ארקוסין שירלי‬
‫הירש תהילה‬
‫סבתו ינון‬
‫סעידיאן אורטל‬
Tel Aviv Medical Ctr
Ariel Univ.
[email protected]
Hebrew Univ.
[email protected]
Technion
[email protected]
Tel Aviv Univ.
[email protected]
Hebrew Univ.
[email protected]
Technion
Shenkar
[email protected]
Bar-Ilan Univ.
[email protected]
Tel Aviv Univ.
[email protected]
Haifa Univ.
[email protected]
Tel Aviv Univ.
[email protected]
Hebrew Univ.
[email protected]
Holon Inst. of Tech.
[email protected]
Technion
[email protected]
Holon Inst. of Tech.
[email protected]
Ben Gurion Univ.
[email protected]
Ben Gurion Univ.
[email protected]
Israel Traffic Police
[email protected]
Technion
[email protected]
Haifa Univ. & IBM
[email protected]
Technion
[email protected]
Ben Gurion Univ.
[email protected]
Technion
[email protected]
Tel Aviv Univ.
[email protected]
Bar-Ilan Univ.
[email protected]
Haifa Univ.
[email protected]
Tel Aviv Univ.
[email protected]
Technion
[email protected]
Columbia Univ.
[email protected]
Columbia Univ.
‫מכון לב‬
‫מכון טל‬
‫מכללת אפקה‬
‫מכון טל‬
‫מכון לב‬
‫מכללת אפקה‬
34
M4c
S1
S5c, s5c*
S5d
M4b
M1a
S2d*
S2c
M5c
S5a
S2a
S5b, S5b*
M4c
M1c
M1a, M5a*
M4a
M5b
S5a, s5a*
M4a
M4d
M4b, M4b*
M1d, M1d*
M5c, S3*
M4c
S5a
S5c
S2c
S2c*
S5d
M4c*
S2a, M2*
M1b, M1b*
M2
M4c
M1d
M1d
M1d
M1d
M1d
M1d
Operations Research Society of Israel
(ORSIS)
Annual Meeting
May 10-11, 2015
Organized by
Technion—Israel Institute of Technology
The William Davidson Faculty of Industrial Engineering and Management
Faculty of Electrical Engineering
Sponsored by
The Organizing Committee:
Nahum Shimkin (Chair)
Michal Penn
Gail Gilboa-Friedman
35

ORSIS 2015 Program Abstracts - Faculty of Industrial Engineering

Transcription

Similar documents

5 201 SI SR O

נוהל המעבדה לאישור יבוא ושחרור קסדות ליבוא מסחרי

Computerized task - Rooms of the house

הכנס ה-43 של האגודה לקידום הלימודים הקלאסיים בישראל

אפידמיולוגיה תזונתית - בית הספר לבריאות הציבור

אי/ - מנל"מ

© T echnion - Israel Institute of T echnology, T echnion Libraries

בקובץ המצורף

מבוא לבריאות הציבור - בית הספר לבריאות הציבור

(ס " ש 2 היקף שעות לימוד –( אפידמיולוגים פענוח נתונים וחלק מהקבוצות ב

תכנית הכנס (אנגלית ועברית) (pdf, 285 KB) - Konrad-Adenauer

××¨××× ××ª×××××ª

הנאמנים חבר - The Jewish Agency for Israel

מניעת נפילה בטיפול תזונתי

הנאמנים חבר - The Jewish Agency for Israel

2010 Conference Chair

הטכניון – מכון טכנולוגי לישראל, הפקולטה להנדסת מכונות

26 ) of VirtexII FPGA family Self Partial Dynamic Reconfiguration

Exact and Approximate Algorithms for New Variants of Some Classic

Location Awareness in Wireless Ad-Hoc Networks (or How to RLISE a TIGR?)