- This event has passed.
Special Edition of the Queueing Colloquium
On the occassion of the 60th birthday of Ivo Adan and Jacques Resing, we are organizing a special edition of the well-known Queueing Colloquium.
Ivo and Jacques are prominent members of the Queueing community in The Netherlands and worldwide, the authors of the renowned lecture notes "Queueing Systems".
Five distinguished researchers and close collaborators of Ivo and Jacques have agreed to give talks at this celebratory event, which will make this colloquium scientifically, as well as socially, a very interesting event.
|Alp Akçay||TU Eindhoven|
|Marko Boon||TU Eindhoven|
|Sem Borst||TU Eindhoven|
|Geert-Jan van Houtum||TU Eindhoven|
|Bernardo D'Auria||Chieti-Pescara University “G. d’Annunzio”|
|Antonis Economou||University of Athens|
|Stella Kapodistria||TU Eindhoven|
|Vidyadhar Kulkarni||University of North Carolina|
|Gideon Weiss||University of Haifa|
The (tentative) programme can be found through this link
Workload analysis of adaptive-service systems: a matrix geometric approach
In many situations service times are affected by the experienced queueing delay of the particular customer in service, examples may be found in health care, call centers and telecommunication networks. We focus on a simple M/M/c queue in which joining customers are labeled according to the level of the system’s workload (Low/High) at the moment of their arrival. While in service, customers are then served at a speed depending on the label assigned to them.
The model is challenging as the workload depends on the service speeds assigned to the customers while, on the other hand, these depend on the workload by the assigned labels. Assuming exponential distributions, we show that the system may successfully be analyzed by a matrix geometric approach.
The impact of information on strategic customer behavior in queueing systems
The economic analysis of queueing systems with strategic customers is a fast-growing field that complements the earlier studies that concerned the performance evaluation, the design and the dynamic control of service systems.
In such studies, a certain reward-cost structure is imposed on a queueing system that quantifies the customers' desire for service and their dislike for waiting. The customers are allowed to make decisions as to whether to join or balk, to stay or renege, to buy priority or not etc. Then, the collective behavior of the customers is analyzed as a game among the potential customers and a fundamental problem is to determine the corresponding symmetric customer strategy equilibrium profiles. Then, the system is further studied under the equilibrium profiles.
One central question in the literature is what level of information should be given to the customers regarding the state of the system. Using an appropriate information structure, the administrator of a system can modify the collective customer behavior and improve the performance of the system, in terms of throughput, social welfare or profit.
In the present talk, we will review various information structures that have been proposed in the literature. At the one end, there are the observable models, where the customers make their decisions after observing perfectly the state of the service system. At the other end, there are the unobservable models where the customers decide, relying only to their knowledge of the system parameters. There are many other intermediate structures, such as partially observable models, models with a mixture of observing and unobserving customers and models with delayed observations. There are also other pieces of information that can be provided to the customers apart from the number of waiting customers, such as the elapsed service time, the server's status, the remaining time for some event etc. We will present representatives models from each family of information structures and comment on their applicability in the management of service systems.
Integrated learning and decision making
In this presentation, we provide a general theory for a class of maintenance problems (binary actions with state dependent costs) characterized by parameter uncertainty. For this class of problems, we develop mathematical models that integrate learning from real-time data with decision making for scenarios in which components' deterioration processes or lifetime distributions are characterized by an a-priori unknown parameter. This approach, the structured learning and decision making approach, is able to leverage the opportunities that arise due to the ongoing developments in sensor technology and the IoT.
We consider stylized, yet representative models (e.g. Brownian motion with drift, Poisson process, (Hidden) MCs) to analyze the essential maintenance trade-off: costly premature preventive maintenance versus costly tardy corrective maintenance in different scenarios that are prevalent in modern, data-driven manufacturing industries. Moreover, all scenarios (except for the model discussed in the last chapter) share one commonality: population heterogeneity through parameter uncertainty. Although there has been done some work in that area (see previous section), we aim to provide a uni ed mathematical framework to analyze and tackle these problems based on our main methodology: Bayesian learning, MDPs, and their integration.
(joint work with Collin Drent)
Optimal Vaccination Rollout Policies
A limited supply of vaccines is available to a community consisting of several distinct groups. The groups differ in size, their readiness to be vaccinated, and the potential benefits to their members from the vaccine. We develop the optimal vaccine rollout schedule: the policy that dynamically opens and closes the access to the vaccines to each group so as to maximize the total benefit to the society. We find that the optimal vaccine roll-out schedules can be surprisingly complicated and counter-intuitive.
(joint work with Puyao Ge and Jayshankar Swaminathan)
Design for parallel skill based service systems
Service systems with several types of customers and servers subject to a bipartite compatibility graph are in general quite intractable. I will discuss tractable examples of such systems, and their relation to the simpler tractable model of FCFS bipartite matching, and present a conjecture, and a heuristic based on it, to answer problems of design for large scale general systems. A major part of this work was in collaboration with Ivo Adan.
The registration is now closed.