The Young European Queueing Theorists (YEQT) workshop is a yearly event that brings together early-career and world-leading researchers in operations research, fostering a dynamic forum for knowledge exchange and collaboration. This edition will focus on Scheduling and Control in Stochastic Systems, a topic of growing significance due to increasing complexity, scale, and unpredictability in modern service and networked systems. Efficient scheduling and control mechanisms are essential for optimizing performance and ensuring stability. This workshop will explore challenges and innovative solutions from various perspectives, including operations management, stochastic modelling, machine learning, and control.

Organisers:

• Fiona Sloothaak: Eindhoven University of Technology
• Rouba Ibrahim: School of Management of University College London
• Izzy Grosof: Northwestern University

 

Keynote speakers

• Céline Comte, CNRS & LAAS (Keynote speaker APT)
• Mor Harchol-Balter, Carnegie Mellon University
• Benny van Houdt, University of Antwerp
• Nan Liu, Carroll School of Management, Boston College
• Rhonda Righter, UC Berkeley

Invited Early-Career Researchers: 

• Elene Anton, University of Pau and Pays de l’Adour
• Eyal Castiel, University of Twente
• Diego Goldsztajn, Universidad ORT Uruguay
• Yige Hong, Carnegie Mellon University
• Yue Hu, Stanford University
• Lucy Huo, Cornell University
• Daniela Hurtado-Lange, Kellogg School of Management, Northwestern University
• Sanne van Kempen, Eindhoven University of Technology
• Ziv Scully, Cornell University
• Yueyang Zhong, London Business School

Registration

Registration has been closed! If you want to know if there is still a possibility to participate, please send an email to eurandom.office@tue.nl

YEQT is sponsored by

 

 

 

 

 

 

 

Programme

Monday November 3, 2025
10:00-11:00	Keynote: Nan Liu
11:00-11:15	Break
11:15-12:00	Yueyang Zhong
12:00-12:45	Daniela Hurtado-Lange
12:45-14:00	Lunch
14:00-15:00	Keynote: Mor Harchol-Balter
15:00-15:30	Break
15:30-16:15	Ziv Scully (remote)
16:15-17:00	Yige Hong (remote)
17:00-18:30	Welcome drinks
Tuesday November 4, 2025
10:00-11:00	Keynote: Benny van Houdt
11:00-11:15	Break
11:15-12:00	Sanne van Kempen
12:00-12:45	Eyal Castiel
12:45-14:00	Lunch
14:00-15:00	Keynote: Rhonda Righter
15:00-15:30	Break
15:30-16:15	Elene Anton
16:15-17:00	Yue Hu (remote)
Wednesday November 5, 2025
09:30-10:15	Lucy Huo
10:15-11:00	Diego Goldsztajn
11:00-11:15	Break
11:15-12:15	Keynote: Céline Comte (Keynote lecture APT)
12:15:	Closing + Lunch

Abstracts

 

Elene Anton

Comparing the performance of the redundancy-d models under the FCFS and ROS scheduling policies
In this talk, we analyse the response time of redundancy-d systems under the first-come first-served (FCFS) and random order of service (ROS) scheduling policies.  Redundancy has received considerable attention as a dispatching paradigm that promises the potential for significant response time improvements. Previous work on redundancy-d models – with exponentially distributed service times and independent and identically distributed copies – has focused on response time results under FCFS or stability conditions under a wider range of scheduling policies, including ROS, and shows that the stability region coincides under both scheduling policies.  We provide the first proof that the FCFS scheduling policy improves the mean response time of the redundancy-d model compared to the ROS policy. To do so, we present a novel coupling technique that exploits the knowledge of the steady-state distribution under the FCFS policy.

 

Céline Comte

Product-Form Queueing Systems, Exponential Families, and Policy-Gradient Reinforcement Learning
This talk builds on the relationship between product-form queueing systems and exponential families. Queueing systems with a product-form stationary distribution have played a central role in the development of queueing theory. Popular examples are Jackson, Whittle, and BCMP networks from the 1970s, as well as order-independent queues from the 1990s. Recent applications of these product-form queueing systems include redundancy scheduling and online stochastic matching. Exponential families are parametric sets of probability distributions that are used extensively in machine learning, as they are numerically tractable and appear naturally as maximum-entropy probability distributions. After observing an equivalence between product-form distributions and exponential families, we focus on a particular consequence of this equivalence, in the design of policy-gradient reinforcement learning algorithms to optimize performance in product-form queueing systems. We show multiple applications of this result in admission-control and load balancing. Lastly, we discuss future research perspectives along this research line.

This talk is based on joint works with Abdelkrim Alahyane (EMINES-UM6P & LAAS), Matthieu Jonckheere (CNRS & LAAS), Éric Moulines (École Polytechnique), Pascal Moyal (Institut Élie Cartan de Lorraine), Jaron Sanders (Eindhoven University of Technologies), and Albert Senen-Cerda (CNRS, IRIT & LAAS).

 

Mor Harchol-Balter

Optimal Core Allocation for Parallel Speedup Jobs
Most queueing-theory models assume that each job runs on a single server. But this one-server-per-job model is not a good representation of today’s compute jobs. In this talk we define the parallel speedup job model, which is representative of many machine learning training jobs, database queries, and scientific computing jobs. Each job can run on any number of cores, but the job’s speed depends on the number of cores on which the job is run, where there is typically a depreciating benefit from each additional core. The question is to understand how to best share a limited number of cores among a stream of jobs, each governed by a potentially different speedup function. We discuss some recent optimality results in this nascent area.

This talk is designed to be accessible to all and also features many open problems.

Joint work with Ben Berg, UNC Chapel-Hill

 

Diego Goldsztajn

Asymptotically Optimal Policies for Weakly Coupled Markov Decision Processes
We will consider the problem of maximizing the expected average reward obtained over an infinite time horizon by $n$ weakly coupled Markov decision processes; this setup is a generalization of the multi-armed restless bandit problem that allows for multiple actions and constraints. We will establish a connection with a deterministic and continuous-variable control problem where the objective is to maximize the average reward derived from an occupancy measure that represents the empirical distribution of the states of the processes in the limit as $n \to \infty$. We will prove that a solution of this fluid control problem can be used to construct simple policies for the weakly coupled processes that achieve the maximum expected average reward as $n \to \infty$, and we will give sufficient conditions for the existence of such solutions. Under certain assumptions on the constraints, we will prove that these conditions are automatically satisfied if the unconstrained single-process problem admits a suitable unichain and aperiodic policy. In particular, these assumptions hold for multi-armed restless bandits and a broad class of problems with multiple actions and inequality constraints. Moreover, we will show that the policies can be constructed in an explicit way in these cases.

 

Yige Hong

A new Lyapunov approach for fully heterogeneous weakly-coupled MDPs
Heterogeneity poses a fundamental challenge for many real-world large-scale decision-making problems but remains largely understudied. In this paper, we study the \emph{fully heterogeneous} setting of a prominent class of such problems, known as weakly-coupled Markov decision processes (WCMDPs). Each WCMDP consists of $N$ arms (or subproblems), which have distinct model parameters in the fully heterogeneous setting, leading to the curse of dimensionality when $N$ is large. We show that, under mild assumptions, a natural adaptation of the ID policy, although originally proposed for a homogeneous special case of WCMDPs, in fact achieves an $O(1/\sqrt{N})$ optimality gap in long-run average reward per arm for fully heterogeneous WCMDPs as $N$ becomes large. This is the first asymptotic optimality result for fully heterogeneous average-reward WCMDPs. Our techniques highlight the construction of a novel projection-based Lyapunov function, which witnesses the convergence of rewards and costs to an optimal region in the presence of heterogeneity.

 

Benny van Houdt

Distributed Join-Idle-Queue Load Balancing with Redundant Tokens
The class of Join-Idle-Queue load balancing algorithms was designed for Cloud service data centers. These algorithms have low overhead and achieve high performance for low to medium loads. In high load scenarios several variations have been proposed to improve the performance. These improvements either put work on the critical path or require two-way communication.
In this talk we discuss the impact of adding  redundant tokens to the Join-Idle-Queue load balancing algorithm to improve the performance under high loads.  We discuss some associated mean field models and identify a number of open problems.

Yue Hu

The Impact of Information-Granularity and Prioritization on Patients’ Care Modality Choice
The past few years have witnessed a significant expansion in telemedicine adoption by healthcare providers. On one hand, telemedicine has the potential to increase patients’ access to medical appointments. On the other hand, due to the limitations of remote diagnostic and treatment methods, telemedicine may be insufficient for patients’ treatment needs and may necessitate subsequent in-person follow-up visits. To better understand this tradeoff, we model the healthcare system as a queueing network providing two types of service: telemedicine and in-person consultations. We assume that an in-person visit guarantees successful treatment, whereas a telemedicine visit may fail to meet the patient’s treatment needs with a probability that is contingent on individual patient characteristics. We formulate patients’ strategic choices between these care modalities as a queueing game, and characterize the game-theoretic equilibrium and the socially optimal patients’ choices. We further examine how improving patients’ understanding of their telemedicine suitability through predictive analytics at the online triage stage affects system performance. We find that increasing information granularity maximizes the stability region of the system but may not always be optimal in reducing the average waiting time. This limitation, however, can be overcome by simultaneously deploying a priority rule that induces the social optimum under specific conditions. Finally, leveraging real-world data from a large academic hospital in the United States, we perform a comprehensive case study that encompasses both the development of a prediction model for in-person follow-up needs and the implementation of effective information provision and prioritization strategies.

 

Lucy Huo

Exploring the Operational Potential of Queue-Ratio Policies in Multiclass Queueing Networks
Motivated by the recent stability analysis of multiclass queueing networks under queue-ratio policies by Zhao, Gurvich, and Hasenbein (2024), we build on their results to explore the operational implications of this policy class. Under their assumptions, together with a P-matrix reflection matrix, we establish that, in a multi-scale heavy traffic regime, the stationary distribution of the scaled workload converges to a product-form limit with exponentially distributed marginals. Together with state space collapse, the scaled queue lengths are also exponentially distributed in the limit. These results yield a tractable closed-form approximation for steady-state performance evaluation without simulation. Leveraging the closed-form approximations, we investigate their potential as a robust engineering tool for operational optimization, for example, identifying near-optimal policies for objectives such as delay minimization and resource balancing. We present numerical experiments and report our ongoing exploration in this direction, concluding with a discussion of open problems.

 

Daniela Hurtado Lange

Heavy-Traffic Analysis of Markov-Modulated Queues
In general, analyzing the delay distribution in queueing systems is intractable. However, key performance measures require knowledge of the distribution. For example, the service level requires computing the tail probabilities. To reduce this gap, a common practice in queueing theory is to analyze the system in an asymptotic regime, such as heavy traffic (where one loads the system close to its maximum capacity). Heavy traffic analysis has gained special attention in the last decade, when ‘direct methods’ have been developed. Here, one analyzes the steady state of the scaled queue lengths instead of a reflected Brownian motion. The Transform Method is one of them, and it focuses on studying the drift of an exponential transform of the scaled queue lengths (the test function) to show convergence in Moment Generating Function. A key step in this method is carefully designing a test function that yields a meaningful result. This method has been successfully used in systems where the inter-arrival and service times are i.i.d., and generalized to an input-queued switch with Markov-modulated arrivals. We propose using the Poisson equation to design the test function in this work. Specifically, we show that using the Poisson equation immediately yields the correct test function to study systems where the inter-arrival and service times depend on the queue lengths. Specifically, we study queueing systems with control in the arrival or the service process using the same essential steps. In this talk, I will present the methodology and examples of its use.

 

Sanne van Kempen

Learning payoffs while routing in skill-based queues
Motivated by applications in service systems, we consider skill-based queueing systems where each customer must be handled by a server with the right skill set. We focus on optimizing the routing of customers to servers in order to maximize the total payoff of customer–server matches. In addition, customer–server dependent payoff parameters are assumed to be unknown a priori. We present a machine learning algorithm that adaptively learns the payoff parameters while maximizing the total payoff. The algorithm considers the basic feasible solutions of a static linear program as candidate solutions in a Multi-Armed Bandit setting.

The first part of the talk is dedicated to the mathematical analysis of the algorithm. We show that the algorithm is asymptotically optimal, up to logarithmic terms, by deriving a regret lower bound. The regret analysis overcomes the complex interplay between queueing and learning by analyzing the convergence of the queue length process to its stationary behavior. In the second part of the talk, we will discuss an extensive simulation study of the algorithm using real-life call center data. We demonstrate the potential of the algorithm with a focus on robustness, adaptability to changing environments, and other challenging scenarios.

 

Nan Liu

Dynamic Scheduling in Healthcare: Revisiting Classic Paradigms and Introducing Advance Notice
There are two classic paradigms of dynamic scheduling in healthcare: advance scheduling and allocation scheduling. Advance scheduling requires providers to guarantee patients specific service times, offering the convenience of knowing exactly when they will be seen. In contrast, allocation scheduling places patients on a wait list and gives providers the flexibility to call them for service at the last minute. These two paradigms represent opposite ends of the scheduling spectrum, favoring patients and providers, respectively. Advance scheduling is common in outpatient settings (e.g., primary care, dental care), whereas allocation scheduling is often used for elective surgeries in publicly funded health systems (e.g., in Canada or the UK).

In this talk, I will first review the modeling frameworks and structural properties of the two classic paradigms. Then, I will introduce a new approach, advance notice, inspired by challenges we observed in managing hospital inpatients for diagnostic services. Under advance notice, patients are placed in a common queue waiting to be called for service, and they will be provided both a fixed preparation time and a guaranteed service time window in advance (neither a last-minute notice nor an exact future appointment time). Advance notice represents a new scheduling paradigm to strike a fine balance between the two classic ones: it enjoys the benefit of allocation scheduling (giving the provider flexibility in using her capacity) and that of advance scheduling (reducing uncertainties in patient waiting for service). We formulate and analyze a Markov Decision Process model to study advance notice decisions. Time permitting, I’ll also present a case study using data from a large academic medical center in the U.S. to illustrate the performance of this new scheduling paradigm.

 

Rhonda Righter

Scheduling in Service and Matching Systems with Compatibility Constraints
I will discuss systems, such as matching platforms, with Poisson arrivals of items of two types (e.g., riders and drivers) that are to be matched with compatible items of the other type, where compatibilities between classes of each type are described by a bivariate matching graph. Such models have many applications, including to classical parallel-server queues. I will describe conditions under which simple scheduling (matching policies) minimize waiting times, including strict class-based (index) priority policies and the match the longest compatible queue policy. I will also consider the performance of two service (matching) disciplines that are neither state-dependent nor class-dependent: FCFM (first-come-first-matched, among compatible items) and ROM (random order of matching, among compatible items), and show that under appropriate symmetry conditions FCFM has better performance than ROM.

Joint work with Kristy Gardner, Esa Hyytiä, and Runhan Xie

 

Ziv Scully

Strongly Tail-Optimal Scheduling in the Light-Tailed M/G/1
We study the problem of scheduling jobs in a queueing system, specifically an M/G/1 with light-tailed job sizes, to asymptotically optimize the response time tail. For some time, the best known policy was First-Come First-Served (FCFS), which has an asymptotically exponential tail. FCFS achieves the optimal exponential decay rate, but its leading constant is suboptimal. Designing a policy that minimizes this leading constant is a long-standing open problem.
We solve this open problem with a new scheduling policy called 𝛾-Boost. Roughly speaking, 𝛾-Boost operates similarly to FCFS, but it pretends that small jobs arrive earlier than their true arrival times. This reduces the response time of small jobs without unduly delaying large jobs. We prove 𝛾-Boost’s asymptotic tail optimality, and we show via simulation that 𝛾-Boost has excellent practical performance.
The 𝛾-Boost policy as described above requires knowledge of job sizes. In preliminary work, we generalize 𝛾-Boost to work with unknown job sizes, proving an analogous asymptotic optimality result in the unknown-size setting. Our generalization reveals that 𝛾-Boost is a type of Gittins index policy, but with an unusual feature: it uses a negative discount rate.

Joint work with George Yu (Cornell) and Amit Harlev (Cornell). Paper links: [1] [2].

 

Yueyang Zhong

When Strategic Customers Meet Strategic Servers: Individual and Social Optimization in Many-Server Queueing Systems
We initiate the study of interactions between strategic customers and strategic servers in many-server queueing systems. In the seminal works of Naor (1969) and Knudsen (1972), arriving customers decide whether to join the system based on a trade-off between their reward from service and the cost of waiting, while servers operate at fixed exogenous rates. The joining threshold (maximum queue length) induced by self-interested customers at equilibrium is higher than the socially optimal threshold (i.e., self-interested customers “over-join”), a result known as Naor’s inequality. This occurs because they fail to internalize the congestion externality they impose on other customers. Although Naor’s inequality is robust across a wide spectrum of queueing models, its validity when servers are also strategic (in choosing their work speeds) is unknown. Within a large-system asymptotic framework, we show that strategic servers provide slower-than-socially-optimal service at equilibrium, thereby deterring customers from joining the system. When this effect outweighs customers’ tendency to over-join, Naor’s inequality is reversed. Next, we propose an incentive scheme that charges an entry fee from customers and offers a performance-based compensation for servers, which realigns individual incentives with the social optimum. Finally, we analyze how the overall welfare gains at the social optimum (relative to the equilibrium) are distributed across customers and servers, in order to uncover potential equity trade-offs in policy design.