European Institute for Statistics, Probability, Stochastic Operations Research
and their Applications

About | Research | Events | People | Reports | Alumni | ContactHome

November 3-4-5, 2014

Workshop on


 "Stochastic Service Systems"






8th Young European Queueing Theorists workshop.

The YEQT workshops are organized annually by the Eurandom research institute for stochastics located at the Eindhoven University of Technology in the Netherlands. The aim of these workshops is to bring together young researchers (recently appointed lecturers or assistant professors, post-docs and senior PhD students) and renowned scientists, to share and discuss research related to
operations research, applied probability and related areas. The workshop program will consist of a set of presentations from young researchers and several keynote presentations and tutorials by
senior researchers.

The theme of this year’s YEQT workshop is "Service operations", which encompasses operations of companies other than manufacturing and distribution. For the most part, service operations include the following three broad fields:

  • maintenance and service logistics

  • call center operations

  • healthcare operations

As examples, one may think of:
- the maintenance of aircraft by airlines and logistics for related resources such as service engineers and spare parts
- the operation of large call centers for various companies
- the logistics involved with the care of patients in hospitals across different departments

In all these cases, there is considerable uncertainty that needs to be addressed in optimizing these operations.
Applied probability is therefore a crucial component in modeling and optimizing the performance of service operations in these contexts. While the field of service operations is defined based on its
applications, it also motivates several more fundamental studies in (applied) probability. For example, renewal theory was initially conceived to study the decision of when to perform maintenance on assets. Renewal theory remains a primary tool in the study of many maintenance decisions. Furthermore, many service operations can be modeled as queueing systems (e.g., call centers).

Former YEQT workshops


Onno Boxma TU Eindhoven
Ahmad Al Hanbali University of Twente
Joachim Arts TU Eindhoven
Peter van de Ven CWI




Zeynep Akşin Karaesmen Koç University
Jeffrey Kharoufeh University of Pittsburgh
Raik Stolletz University of Mannheim



Refik Güllü Boğaziçi University
Geert-Jan van Houtum TU Eindhoven



Peter van Berkel Dalhousie University
Tejas Bodas IIT Bombay
Christian Bohner TU Münich
Arnoud den Boer University of Twente
Nicaise Choungmo Fofack IMT/OLN
Nicolas Gast INRIA
Willem van Jaarsveld Erasmus University
Marijn Jansen University of AMsterdam/Ghent University
Rouba Ibrahim University College London
Maialen Larrañaga CNRS-LAAS
Chiel van Oosterom TU Eindhoven
Justus Arne Schwarz University of Mannheim
Ingrid Vliegen University of Twente





09.30 - 09.45 Registration/Opening Remco van der Hofstad  
09.45 - 10.45 Keynote Raik Stolletz Decisions in time-dependent service queues
10.45 - 11.00 Break    
11.00 - 11.30 Invited Justus Schwarz A classification of approaches to the performance evaluation of time-dependent queues
11.30 - 12.00 Invited Marijn Jansen Overflow asymptotics for an infinite-server queue in different random environments
12.00 - 12.30 Invited Maialen Larrañaga Index policies for scheduling problems
12.30 - 14.00 Lunch    
14.00 - 15.00 Tutorial Refik Güllü Pricing in Service Systems with Strategic Customers
15.00 - 15.15 Break    
15.15 - 15.45 Invited Arnoud den Boer Convergence rates of Laplace-transform based estimators
15.45 - 16.15 Invited Rouba Ibrahim Managing Call Centers with Many Strategic Agents
16.15 - 16.30 Break    
16.30 - 17.30 Keynote Zeynep Akşin Karaesmen Modeling the Customer Role in Services: Examples from Call Centers
18.30 - Conference dinner    



09.30 - 10.30 Tutorial Refik Güllü Pricing in Service Systems with Strategic Customers
10.30 - 11.00 Break    
11.00 - 11.30 Invited Peter Vanberkel Designing Offload Zones to reduce Offload Delay
11.30 - 12.00 Invited Ingrid Vliegen Designing cyclic appointment schedules for outpatient clinics with scheduled and unscheduled patient arrivals
12.00 - 12.30 Invited Nicolas Gast Asymptotic properties of object-sharing systems
12.30 - 14.00 Lunch    
14.00 - 15.00 Keynote Jeffrey Kharoufeh Optimal Replacement Policies for Wind Energy Systems
15.00 - 15.15 Break    
15.15 - 15.45 Invited Chiel van Oosterom Maintenance Optimization for a Markovian Deteriorating System with Population Heterogeneity
15.45 - 16.15 Invited Willem van Jaarsveld Optimization of Industrial-Scale Assemble-To-Order Systems
16.15 - 16.30 Break    
16.30 - 17.30 Tutorial Geert-Jan van Houtum System-oriented inventory models for spare parts



09.30 - 10.30 Tutorial  Geert-Jan van Houtum System-oriented inventory models for spare parts
10.30 - 11.00 Break    
11.00 - 11.30 Invited Christian Bohner A Continuous Review, One-Warehouse Multi-Retailer Problem with Non-Homogeneous Poisson Demand
11.30 - 12.00 Invited Tejas Bodas Queues, Tolls and Welfare
12.00 - 12.30 Invited Nicaise Choungmo Fofack On-demand caches and content-oriented networks
12.30 - 12.35 Closing and Lunch    



Zenyep Akşin

Modeling the Customer Role in Services: Examples from Call Centers

Queues model random arrivals and random departures from service systems with limited capacity. The standard paradigm assumes both arrivals and departures as given features of the system and attempts to model these via stochastic processes and random variables. Yet as system features change, one can expect these model primitives to also change in response. In call centers, as well as many other service systems, the waiting entities are humans. The most visible manifestation of having a human being waiting in a queue is the phenomenon of abandonment. Customer patience is an important feature that needs to be accounted for in modeling services to properly account for departures from the system. Similarly, customer arrivals to a system are affected by queue features. As these change, a customer may decide to join the queue or not, and may choose to behave differently in the queue.
This talk will focus on such customer reactions to queuing features and discuss how these can be modeled as well as the use of data in understanding such reactions. The first part of the talk will focus on modeling customer patience in call centers and the design of delay announcement systems. In the second part, a different customer reaction, this time to sales attempts in a call centers is used as an illustrating example of modeling customer reactions.


Tejas Bodas

Queues, Tolls and Welfare

We consider a queueing system with multiple heterogeneous servers serving a multiclass population. The classes are distinguished by the time costs. All customers have i.i.d. service requirements.
Arriving customers do not see the instantaneous queue occupancy. Arrivals are randomly routed to one of the servers and the routing probabilities are determined centrally to optimize the expected waiting
cost. This is, in general, a difficult optimization problem and we obtain the structure of the routing matrix. Next we consider a system in which each queue charges an admission price. The arrivals are routed
randomly to minimize an individual objective function that includes the expected waiting cost and the admission price. Once again, we obtain the structure of the equilibrium routing matrix for this case. Finally, we determine the admission prices to make the equilibrium routing probability matrix equal to a given optimal routing probability matrix.
(joint work with Prof. A.J. Ganesh, University of Bristol, UK and Prof. D. Manjunath, IIT Bombay)


Christian Bohner

A Continuous Review, One-Warehouse Multi-Retailer Problem with Non-Homogeneous Poisson Demand

We extend the continuous review one-warehouse multi-retailer inventory problem to non-homogeneous Poisson demand. We apply unit-tracking to single-unit subproblems and find optimal time-dependent base-stock levels at the warehouse and the retailers. The procedure of deriving the base-stock levels is presented and uses a decomposition of the subproblems into two stages. Having the optimal base-stock levels, we find their time-dependent changes to be caused by either a quantity effect or an allocation effect that we both characterize. In addition, we show in a numerical study that the solution from a rolling horizon approach is far from optimal. For a linear demand rate, however, we prove that a time-shift of a half lead time applied to the rolling horizon solution yields the optimal retailer base-stock levels. From this result, we derive a simple shift heuristic.

Arnoud den Boer

Convergence rates of Laplace-transform based estimators



Nicaise Choungmo Fofack

On-demand caches and content-oriented networks

In this talk, we will start by a chronological review of caching from computer page memories to cache networks with strong emphasis on applications, challenges, and achievements.
Then, the second part of the presentation will show two simple cache architectures—called Time-To-Live (TTL)-based cache models—that greatly simplify the performance evaluation of heterogeneous and general cache networks with arbitrary topologies.
The last part of this talk will be devoted to possible extensions of these TTL-based models to address a set of interesting open issues related to content distribution in networks.


Nicolas Gast

Asymptotic properties of object-sharing systems

An object-sharing systems is a system where a number of objects are available at different locations. Users of such systems arrive at one of these locations; take an object; use it for a while; and return it to another station. Bike-sharing systems are a popular example of such systems. In my talk, I will show how to model such systems by stochastic queuing networks. I will derive asymptotic properties as the number of locations is large. This allows to compute the key performance metrics, such as the number of satisfactory trips, and to propose a few counter-intuitive heuristics, such as forbidding some trips, to improve the performance.


Refik Güllü (Tutorial)

Pricing in Service Systems with Strategic Customers

In this tutorial we consider systems where customers decide to join or not based on their valuation of the service, the waiting cost, and the price charged for the service. The service provider (depending on its objective) decides on the price to be charged, and in the presence of different customer classes, determines any prioritization of the customers. We present the fundamental models in this area, discuss the impact of various forms of information revealed to the customers, and summarize extensions of the fundamental models.


Geert-Jan van Houtum

System-oriented inventory models for spare parts

Stocks of spare parts, located at appropriate locations, can prevent long downtimes of technical systems that are used in the primary processes of their users. Since such downtimes are typically very expensive, generally system-oriented service measures are used in spare parts inventory control. Examples of such measures are system availability and the expected number of backorders over all spare parts. This is one of the key characteristics that distinguishes such inventory control from other fields of inventory control. In this tutorial, we survey models for spare parts inventory control under system-oriented service constraints. We link those models to two archetypical types of spare parts networks: networks of users who maintain their own systems, and networks of original equipment manufacturers who service the installed base of products that they have sold. We discuss both single location and multi-echelon models. We further focus on the use of lateral and emergency shipments, and we refer to applications in practice. We also discuss open research problems.

PRESENTATION (survey article)

Rouba Ibrahim

Managing Call Centers with Many Strategic Agents

Motivated by real-life work arrangements, such as those in place at Hydro-Quebec, Canada, we study optimal staffing and routing policies in large call centers where agents are strategic in selecting their own work schedules and handled call types. To mimic the operations of such call centers, we consider large, heavily-loaded, queueing systems with customer abandonment and many strategic agents.  We present a model where strategic agents make decisions by maximizing their individual utilities, and characterize the resulting symmetric Nash equilibrium in the system. We formulate and solve the operational problems faced by the system manager, both with and without strategic agents. As such, we quantify the operational impact of strategic agent behaviour in the system. We then propose compensation schemes that make strategic agents voluntarily behave in line with the optimal solutions of the system manager.  We also study optimal operational decisions in a centralized system, where a social planner maximizes the total social welfare.
(joint work with Kenan Arifoglu)


Willem van Jaarsveld

Optimization of Industrial-Scale Assemble-To-Order Systems

We provide insights and algorithms to improve inventory control in industrial-sized Assemble-To-Order (ATO) systems. We seek base-stock levels for components that minimize the sum of holding costs and product-specific backorder costs. Our initial focus is on first-come first-serve (FCFS) allocation of components to products. By developing a novel stochastic programming (SP) formulation for this setting, we compute solutions that are within one percent of the lower bound for realistically sized systems. We then answer the following questions for such systems: How do common heuristics used in practice compare to our performance, and how costly is the FCFS assumption?
For the first question, we investigate the effectiveness of ignoring simultaneous stock-outs, a heuristic that has been used by companies such as IBM and Dell to optimize inventory levels. We find that the performance of the heuristic, when compared to the optimal FCFS base-stock policy, increases as the average newsvendor (NV) fractiles increase.  For the second question, we adapt the SP formulation of Doğru, Reiman and Wang (2010), yielding an upper bound on the benefit of optimal allocation over FCFS. We find that FCFS performs surprisingly well for many practical cases, and that its performance improves, again, with increasing average NV fractiles.


Marijn Jansen

Overflow asymptotics for an infinite-server queue in different random environments

In this talk, we will consider an infinite-server queue in two different random environments. A random environment is represented by a stochastic background process, which modulates both the arrival process and the service process of the queue. The two different background processes that we will consider are a continuous-time Markov chain and a reflected Brownian motion. We will look at a queue that has to divide its attention between allowing customers to enter and serving the customers. The background process determines how much attention goes to either task. We will study the probability that the number of jobs in the system becomes unusually large, i.e. we will study overflow. Scaling the arrival process and using large deviations techniques, we compute the rate functions that describe the exponential rate of convergence of the overflow probabilities in the respective random environments. Surprisingly, the rate functions turn out to be the same for both random environments. Apparently, two very different modulating processes may lead to the same large deviations principle.
(Joint work with K. De Turck (Ghent University), M. R. H. Mandjes (University of Amsterdam) and S. Wittevrongel (Ghent University)

Jeff Kharoufeh

Optimal Replacement Policies for Wind Energy Systems

We present a framework within which real-time, condition-based data can be exploited to optimally time the replacement of a critical wind turbine component. First, we develop hybrid analytical-statistical tools to estimate the current health of the component and approximate the expected time at which it will fail by observing a surrogate signal of degradation. The signal is assumed to evolve as a switching diffusion process, and its parameters are estimated via a novel Markov chain Monte Carlo procedure. Next, we address the problem of optimally replacing a critical component that resides in a partially-observable environment. Two models are formulated using a partially-observed Markov decision process (POMDP) framework. The first model ignores the cost of turbine downtime, while the second includes this cost explicitly. For both models, it is shown that a threshold replacement policy is optimal with respect to the cumulative level of component degradation. Moreover, these thresholds depend on the decision maker’s assessment of the environmental conditions. Time permitting, a third model will be presented that considers cases in which the environment is partially observed and degradation measurements are uncertain. It will be shown that a threshold policy is also optimal for this challenging scenario. Several numerical examples illustrate the main results and the value of including environmental observations in the wind energy setting.

Maialen Larrañaga

Index policies for scheduling problems

We develop a unifying framework to obtain efficient index policies for restless multi-armed bandit problems with birth-and-death state evolution. This is a broad class of stochastic resource allocation problems whose objective is to determine efficient policies to share resources among competing projects. In a seminal work, Whittle developed a methodology to derive well-performing (Whittle's) index policies that are obtained by solving a relaxed version of the original problem. We first derive a closed-form expression for Whittle's index as a function of the steady-state probabilities. It can be efficiently calculated, however, it requires several technical conditions to be verified and, in addition, it does not provide qualitative insights into Whittle's index. We therefore formulate a fluid version of the relaxed optimization problem and we develop a fluid index policy. The latter does provide qualitative insights and is close to Whittle's index. The applicability of our approach is illustrated by two important problems:  optimal class selection and optimal load balancing. Allowing state-dependent capacities we can model important phenomena: e.g. power-aware server-farms and opportunistic scheduling in wireless systems.
Numerical simulations show that Whittle's index and our fluid index policy are both nearly optimal. 


Chiel van Oosterom

Maintenance Optimization for a Markovian Deteriorating System with Population Heterogeneity

We develop a partially observable Markov decision process (POMDP) model to incorporate population heterogeneity when scheduling replacements for a deteriorating system. The single-component system deteriorates over a finite set of deterioration levels according to a Markov chain. The population of spare components that is available for replacements is composed of multiple component types that cannot be distinguished by their exterior appearance but deteriorate according to different transition matrices. As a consequence, the type of an installed component is not known, but partial information can be inferred from the history of observed deterioration levels. We provide a set of conditions for which we characterize the structure of the optimal policy to minimize the total expected discounted operating and replacement costs over an infinite horizon. By a numerical experiment, we benchmark the optimal policy against a heuristic policy that neglects population heterogeneity.

Justus Arne Schwarz

A classification of approaches to the performance evaluation of time-dependent queues

Many queueing systems are subject to time-dependent changes of system parameters, such as time-varying arrival rates,
service rates, and number of servers. We develop a systematic classification of analytical approaches for the performance evaluation of single-stage, time-dependent queueing systems. The key ideas behind the main approaches are reviewed. Based on the classification scheme we establish and discuss links between different approaches.

Raik Stolletz

Decisions in time-dependent service queues

Many service systems, for example call centers or distribution centers, have to cope with uncertain and time-dependent arrivals or service capacities. An overview of evaluation and optimization models for time-dependent and stochastic service systems is given.

As an example, we provide a methodology to evaluate and optimize the arrival pattern for the time-dependent queueing system of truck handling operations at an air cargo terminal. Our optimization approach is based on the stationary backlog-carryover approach to analyze the system's performance. In the related optimization model, the time-dependent arrival rates serve as decision variables, i.e., changes in the original demand pattern are allowed and intentional. The objective of this non-linear optimization model is to minimize total expected waiting times while the corresponding change in the arrival pattern is limited. A numerical example compares the performance measures for original and optimized arrival patterns. It shows that significant reduction in waiting times can be reached even with minor shifts in arrival rates.

Peter Vanberkel

Designing Offload Zones to reduce Offload Delay

Offload delay occurs when the transfer of a patient from an ambulance service to an emergency department is prolonged. Offload delay negatively impacts patient care (e.g. poor pain control, delayed time to antibiotics, etc.) and ambulance coverage by delaying the return of an ambulance to service. In Nova Scotia the 90th percentile of offload delay has increased by a factor of 4 since 2002. The Halifax Infirmary and Dartmouth General Hospital have implemented Offload Zones as a solution to the offload delay problem. The Offload Zone is an area where patients can wait with a paramedic and a nurse allowing the ambulance to return to service immediately.
In this talk I will present two studies related to evaluating and (re)designing Offload Zones.  The first is a retrospective evaluation of the Offload Zone at the Halifax Infirmary following the Healthcare Failure Modes and Effect Analysis framework.  From this study we found that when the Offload Zone was implemented staff behavior changed.  In particular, patients arriving by ambulance went from “high priority to admit” to “low priority to admit”, since Offload Zone patients wait with a nurse and paramedic. Patients from the waiting room, who are waiting without a health care provider, are admitted instead.  This change in priority causes the Offload Zone to “fill-up”, leaving nowhere for arriving ambulances to transfer their patients and, hence, the continuation of offload delay.
The second study uses a continuous time Markov chain to study this effect.  Specifically we identify priority thresholds which 1) ensures the Offload Zone, at a minimum, performs as well as when there is not an Offload Zone and 2) identifies when additional priority reduces offload delay by a negligible amount. Both studies are supporting policy makers in determining the future feasibility of Offload Zones at the Capital District Health Authority.


Ingrid Vliegen

Designing cyclic appointment schedules for outpatient clinics with scheduled and unscheduled patient arrivals

We present a methodology to design appointment systems for outpatient clinics and diagnostic facilities that offer both walk-in and scheduled service. The developed blueprint for the appointment schedule prescribes the number of appointments to plan per day and the moment on the day to schedule the appointments. The method consists of two models; one for the day process that governs scheduled and unscheduled arrivals on the day and one for the access process of scheduled arrivals. Appointment schedules that balance the waiting time at the facility for unscheduled patients and the access time for scheduled patients are calculated iteratively using the outcomes of the two models. Two methods to calculate appointment schedules, complete enumeration and a heuristic procedure, are compared in various numerical experiments. Furthermore, an appointment schedule for the CT-scan facility at the Academic Medical Center Amsterdam, The Netherlands, is developed to demonstrate the practical merits of the methodology. The method is of general nature and can therefore also be applied to scheduling problems in other sectors than health care.
oint work with: Nikky Kortbeek,  Maartje Zonderland, Aleida Braaksma, Richard Boucherie, Nelly Litvak, Erwin Hans)










Eurandom, Mathematics and Computer Science Dept, TU Eindhoven,

Den Dolech 2, 5612 AZ  EINDHOVEN,  The Netherlands

Eurandom is located on the campus of Eindhoven University of Technology, in the brand new TU/e Metaforum building (4th floor) (about the building). The university is located at 10 minutes walking distance from Eindhoven main railway station (take the exit north side and walk towards the tall building on the right with the sign TU/e).
Accessibility TU/e campus and map.



To register for the workshop, please fill out the registration form. There is no fee for participation.


      Accommodation / Funding

Hotel will be booked for all invited speakers. Please give your arrival and departure date on the registration form.

Other participants have to make their own arrangements.

For hotels around the university, please see: Hotels (please note: prices listed are "best available"). 

More hotel options can be found on the webpages of the Tourist Information Eindhoven, Postbus 7, 5600 AA Eindhoven.



For those arriving by plane, there is a convenient direct train connection between Amsterdam Schiphol airport and Eindhoven. This trip will take about one and a half hour. For more detailed information, please consult the NS travel information pages or see Eurandom web page location.

Many low cost carriers also fly to Eindhoven Airport. There is a bus connection to the Eindhoven central railway station from the airport. (Bus route number 401) For details on departure times consult http://www.9292ov.nl

The University  can be reached easily by car from the highways leading to Eindhoven (for details, see our route descriptions or consult our map with highway connections.


      Conference facilities : Conference room, Metaforum Building  MF11&12

The meeting-room is equipped with a data projector, an overhead projector, a projection screen and a blackboard. Please note that speakers and participants making an oral presentation are kindly requested to bring their own laptop or their presentation on a memory stick.


      Conference Secretariat

Upon arrival, participants should register with the workshop officer, and collect their name badges. The workshop officer will be present for the duration of the conference, taking care of the administrative aspects and the day-to-day running of the conference: registration, issuing certificates and receipts, etc.



Should you need to cancel your participation, please contact Patty Koorn, the Workshop Officer.



Mrs. Patty Koorn, Workshop Officer, Eurandom/TU Eindhoven, koorn@eurandom.tue.nl



The organisers acknowledge the financial support/sponsorship of:










Last updated 22-01-15,
by PK

 P.O. Box 513, 5600 MB Eindhoven, The Netherlands
tel. +31 40 2478100  
  e-mail: info@eurandom.tue.nl