Prof. Dr. Mirjam Dür (EURO Plenary)
Professor of Mathematical Optimization
Augsburg University, Germany
Title of Talk: Conic Optimization: An Application-Oriented Survey
A conic optimization problem is a problem involving a constraint that the optimization variable be in some closed convex cone. Linear optimization is a prominent example, where the nonnegativity constraint can be interpreted as requiring that the variable should be in the cone of nonnegative vectors. Other examples are second order cone problems (SOCP) where the variable is constrained to be in the second order cone, and semidefinite programming (SDP) where the matrix variable is required to be in the cone of positive semidefinite matrices. More general cones appear in special applications.
In this talk, we will highlight the enormous modeling power of conic optimization and review recent progres made in this field. While the past decades have seen research mainly in linear conic optimization, interest has now shifted to nonlinear and mixed-integer conic optimization. We will discuss algorithmic progress made in this direction as well as new fields of application. Special emphasis will be given to applications of conic optimization appearing in operations research.
Mirjam Dür was born in Vienna, Austria, where she received a M.Sc. degree in Mathematics from the University of Vienna in 1996. She received a PhD in Applied Mathematics from the University of Trier in 1999. After that, she worked as an Assistant Professor at Vienna University of Economics and Business Administration, as a Junior Professor at TU Darmstadt, as an Universitair Docent at the University of Groningen, The Netherlands, and as a Professor of Nonlinear Optimization in Trier. Since October 2017, she is a Professor of Mathematical Optimization in Augsburg.
Prof. Dür is a member of the editorial boards of the journals Mathematical Methods of Operations Research, Journal of Global Optimization, and Optimization Methods and Software, and of the book series Springer Briefs in Optimization. In 2010, she was awarded a VICI-grant by the Dutch Organisation for Scientific Research (NWO), and in 2012 a GIF Research grant by the German-Israeli Foundation for Scientific Research and Development. She is a former principal investigator and now member of the Scientific Advisory Board in the Research Training Group on Algorithmic Optimization at the University of Trier.
Prof. Dr. Georgia Perakis
William F. Pounds Professor of Operations Research and Operations Management
MIT Sloan School of Management, United States
Title of Talk: Analytics for Tackling COVID-19
In this talk I will discuss how Analytics have helped for tackling the COVID-19 pandemic. I will present work from various groups but will mostly focus on the work of my team related to COVID-19 this past year. I will discuss the MIT-Cassandra model that is a suite of models that are part of an ensemble method for COVID-19 case and death prediction. I will discuss the individual methods and what motivated them and then the ensemble method and show how they perform with data in the US. I will discuss how these models are comparing relative to other models also used by the CDC. I will further connect these predictions with detecting true infection (also referred to as prevalence). Finally, I will discuss how these methods and results can be used to distribute vaccines in different counties (or areas) within a state (or country) to a heterogeneous population, through optimization, ensuring fair distribution among the different counties. We will show how the proposed optimization model performs in the different counties in the state of Massachusetts.
(The MIT-Cassandra team includes in addition to myself my students (current and former): Amine Bennouna, David Nze-Ndong, Boyan Peshlov, Divya Singhvi, Omar Skali-Lami, Yiannis Spantidakis, Leann Thayaparan, Asterios Tsiourvas, Shane Weisberg)
Georgia Perakis is the William F. Pounds Professor of Management and a Professor of Operations Research, Statistics and Operations Management at the MIT Sloan School of Management. She has been on the faculty at MIT Sloan since July 1998. In her research, she investigates the theory and practice of analytics and its role in operations problems. She is particularly interested on how to solve complex and practical problems in pricing, revenue management, supply chains, healthcare, transportation and energy applications among others.
Georgia Perakis has received the CAREER Award from the National Science Foundation and the PECASE Award from the Office of the President on Science and Technology. In 2016, she was elected as an INFORMS Fellow, and in 2021 she was elected as an MSOM Distinguished Fellow. Both recognize individuals for lifetime achievements to their field. In addition, her work has received recognition with awards such as the TSL Best Paper Award, the Best Paper competition of the INFORMS Service Science Section several times as well as Best Application of Theory Award from NEDSI Conference. Her work on predicting demand for new products with Johnson & Johnson won first place at the Applied Research Challenge Competition in 2018. Her paper on subsidies received the 2019 Best Paper Award published in Management Science in the last three years.
Perakis is currently the editor in chief of the M&SOM journal. Prior to that role, she had also served as America’s editor in chief of the Journal of Pricing and Revenue Management, as a department editor for the journal Service Science in the area of Analytics and as an associate editor for the flagship journals Management Science, Operations Research, M&SOM, INFORMS Journal on Optimization, and as a senior editor for POM.
Georgia Perakis holds a BS in Mathematics from the University of Athens as well as an MS in Applied Mathematics and a PhD in Applied Mathematics from Brown University.
Talk of the Winner of the
2021 GOR Scientific Award
The winner will be announced during the Opening Session on Wednesday, September 1.
Prof. Dr. Claudia Archetti
Professor of Operations Research
ESSEC Business School, France
Title of Talk: Formulations and Exact Solution Approaches for the Inventory Routing Problem
In the last decades, Inventory Routing Problems (IRPs) have been attracting growing attention from the research community, due to the real-world applications, in integrated logistics and supply chain management, and the intellectual challenges that their study poses. The interest in studying IRPs is mainly motivated by the potential benefits coming from combining inventory management and routing decisions. Solving two separate optimization problems for inventory management and routing typically produces sub-optimal solutions to the integrated problem. Tackling directly the integrated problem causes an increase of the computational burden, but tends to provide considerably better solutions.
In the IRP the goal is to determine an optimal distribution plan to replenish a set of customers by routing a limited fleet of capacitated vehicles over a discrete planning horizon. Each customer consumes a per period quantity of product and has a maximum inventory capacity. The objective is to minimize the total distribution cost, that includes the routing and the inventory holding costs.
Different formulations have been proposed in the literature for modelling the problem, giving raise to various exact solution approaches, based on branch-and-price and branch-and-cut.
The goal of this talk is to analyse the formulations and study their pros and cons. We will mainly focus on compact formulations, focusing on properties and links between formulations with vehicle indices and aggregated formulations.
Since September 2019, Claudia Archetti is Associate Professor in Operations Research at ESSEC Business School in Paris. She was previously Associate Professor at the University of Brescia. The main areas of the scientific activity are: models and algorithms for vehicle routing problems; mixed integer mathematical programming models for the minimization of the sum of inventory and transportation costs in logistic networks; exact and heuristic algorithms for supply-chain management; reoptimization of combinatorial optimization problems.
Claudia Archetti has carried out the scientific activity in collaboration with Italian and foreign colleagues and published joint papers with some of the best researchers at the international level. She is author of more than 70 papers in international journals. She was area editor of Computers and Operations Research and is currently associate editor of Omega, Transportation Science, Networks, TOP and EURO Journal of Computational Optimization, and member of the editorial board of European Journal of Operational Research. Claudia Archetti is VIP3 of EURO, the Association of European Operational Research Societies, in charge of publications and communication.
Prof. Dr. Margaretha Gansterer
Professor for Production Management and Logistics
University of Klagenfurt, Austria
Title of Talk: Collaborative Vehicle Routing: Computational and Game Theoretical Aspects
The Sharing Economy is on the rise. Traditional business models have to be adapted and players have to learn how to survive in a world of shared idle capacities and digital platforms. The concept of shared transportation resources, also denoted as collaborative vehicle routing, is one of the hot topics in transportation and logistics. A collaboration can be described as a partnership between two or more companies to optimize operations by making joint decisions and sharing information, resources, or profits. While the willingness to enter coalitions does exist, the success of collaborations strongly depends on mutual trust and behavior of participants. Hence, proper mechanism designs, where carriers do not have incentives to deviate from jointly established rules, are needed.
In this talk, we elaborate horizontal collaborations, where logistics providers share resources with their competitors through the exchange of selected transportation requests. The aim is to increase the overall efficiency of the transport industry, by avoiding costly and pollutive empty trips. We focus on decentralized exchange mechanisms, which are based on the assumption that no fully informed decision maker exists. In such mechanisms, efficient solution methods for complex routing problems have to be tackled, while game theoretical aspects have to be taken into account. The talk gives insights on auction-based systems, where several strongly related decision problems have to be integrated. We analyze, for instance, whether carriers face a Prisoner’s Dilemma when selecting requests for trading. Recent findings as well as promising future research directions are presented.
Margaretha Gansterer is Full Professor of Business Administration with a focus on Logistics and Operations Management. She obtained her PhD and her habilitation at the University of Vienna. Currently, she is Head of the Department of Operations, Energy, and Environmental Management at the University of Klagenfurt. Previously, she held a professor position for Operations Management at the Otto-von-Guericke University Magdeburg. Margaretha Gansterer currently serves as the International Liaison (Europe, Middle East, Africa) at the INFORMS Transportation and Logistics Society (TSL) and is member of the EURO WISDOM Forum.
Her research focuses on optimization in shared logistics and manufacturing operations. She is particularly interested in auction-based collaborations, where both game theoretical aspects and the computational complexity of such mechanisms have to be taken into account. She also contributed to the field of vehicle routing, lot sizing, assembly line balancing, and hierarchical production planning. Margaretha Gansterer is principal investigator of two Austrian Science Fund (FWF) projects on collaborative logistics.
Prof. Dr. Daniel Kuhn
Professor of Operations Research
École Polytechnique Fédérale de Lausanne (EPFL), Switzerland
Title of Talk: A Unifying Framework for Robust and Distributionally Robust Optimization
Robust and distributionally robust optimization are modeling paradigms for decision-making under uncertainty where the uncertain parameters are only known to reside in an uncertainty set or are governed by any probability distribution from within an ambiguity set, respectively, and a decision is sought that minimizes a cost function under the most adverse outcome of the uncertainty. In this talk, we develop a general theory of robust and distributionally robust nonlinear optimization using the language of convex analysis. Our framework is based on a generalized ‘primal-worst-equals-dual-best’ principle that establishes strong duality between a semi-infinite primal worst and a non-convex dual best formulation, both of which admit finite convex reformulations. This principle offers an alternative formulation for robust optimization problems that may be computationally advantageous, and it obviates the need to mobilize the machinery of abstract infinite-dimensional duality theory to prove strong duality in distributionally robust optimization. We illustrate the modeling power of our approach through convex reformulations for distributionally robust optimization problems whose ambiguity sets are defined through general optimal transport distances, which generalize earlier results for Wasserstein ambiguity sets.
Daniel Kuhn is Professor of Operations Research at the College of Management of Technology at EPFL, where he holds the Chair of Risk Analytics and Optimization (RAO). His current research interests are focused on data-driven optimization, the development of efficient computational methods for the solution of stochastic and robust optimization problems and the design of approximation schemes that ensure their computational tractability. This work is primarily application-driven, the main application areas being engineered systems, machine learning, business analytics and finance.
Before joining EPFL, Daniel Kuhn was a faculty member in the Department of Computing at Imperial College London (2007-2013) and a postdoctoral research associate in the Department of Management Science and Engineering at Stanford University (2005-2006). He holds a PhD degree in Economics from University of St. Gallen and an MSc degree in Theoretical Physics from ETH Zurich. He serves as the area editor for continuous optimization for Operations Research and as an associate editor for several other journals including Management Science, Mathematical Programming, Mathematics of Operations Research and Operations Research Letters.
Prof. Dr. Frauke Liers
Professor of Applied Mathematics
Friedrich-Alexander University Erlangen-Nuremberg, Germany
Title of Talk: Mixed-Integer Robust Optimization: Some Algorithms and Some Applications
Protecting optimization problems against uncertainties is an exciting research area where new methods and algorithms are developing rapidly. One way of protecting against uncertainties that occur in real-world applications is to determine best possible robust decisions that are feasible regardless of how uncertainties manifest themselves within predefined uncertainty sets.
In this talk, we will review some of the recent developments in particular for mixed-integer robust optimization problems that often apply reformulation, decomposition as well as approximation approaches. Data-driven approaches are on the rise as well. (Mixed-integer) discrete decisions add another difficulty in algorithmic tractability, both in theory as well as in practice.
We will also look into some robust energy network applications together with some overview of the literature. In electricty networks, we show that robust protection can also be used for a robust safe approximation of joint chance constrained in DC Optimal Power Flow problems. For the robust operation of gas networks, we review reformulation and decompotion approaches for the occurring mixed-integer two-stage nonconvex robust problems, where the latter use an outer approximation for a bundle method that is able to deal with nonconvexities.
The main focus of Frauke Liers' research lies on design, analysis and implementation of global algorithms for mathematical optimization problems. In particular, she works in combinatorial and mixed-integer optimization with applications in Logistics, in Operations Research, in Engineering, and in the Natural Sciences. She is particularly interested in optimization under uncertainty, especially in robust optimization.
Frauke Liers studied Mathematics with a minor in Physics and obtained her doctoral degree in Computer Science at the University of Cologne in 2004 and her habilitation in 2010. She then held a grant within the Emmy Noether Programme of the German Science Foundation. In 2012, she moved to the Friedrich-Alexander University in Erlangen-Nuremberg as a Professor in Applied Mathematics. Frauke Liers serves in the editorial boards of Mathematical Methods of Operations Research, Discrete Optimization and Optimization and Engineering. She is involved in several third-party funded projects, in particular in the collaborative research centers TRR154 on modelling, simulation and optimization using the example of gas networks as well as in the CRC1411 on particle design. She currently coordinates the European Marie-Curie Innovative Training Network MINOA on mixed-integer nonlinear optimisation: algorithms and applications.
Prof. Dr. Renata Mansini
Professor of Operations Research
University of Brescia, Italy
Title of Talk: Solving Mixed Integer Linear Programming Problems by Kernel Search: Issues, Challenges and Future Directions
A wide range of optimization problems deriving from different application contexts can be formulated as mixed integer linear programming (MILP) problems. The solution of these complex problems is usually addressed with customized heuristic methods that can be seldom reused, even to solve similar problems. In the literature, several attempts have been made to overcome the drawback of problem-dependent heuristics. For example, metaheuristic algorithms introduce general schemes that explore the solution space regardless of the underlying problem structure, whereas general-purpose methods exploit commercial MILP solvers as effective off-the-shelf tools to optimize problems where no specific insight is used beyond the one provided by their mathematical formulations.
Kernel Search (KS) can be classified as a general-purpose heuristic framework based upon a straightforward decomposition paradigm. More precisely, KS constructs a sequence of restricted subproblems by identifying a subset of promising variables, called the kernel set, and partitioning the remaining variables into buckets. Each restricted subproblem takes into account the kernel set (possibly updated) plus a selected set of additional variables (the current bucket) and is solved by means of a commercial MILP solver. For this reason, KS is extremely easy to implement. The larger the kernel set, the more likely you are to get better solutions, but also higher computing times. According to the solved problem, KS can consider only a defined number of buckets or scroll the whole sequence of buckets more than once, use disjoint buckets or allow for their partial overlapping, consider equal or variable size buckets.
We will discuss the main features and the critical issues of the method by underlining its strong potential and indicating open challenges and future directions. Since the method has produced high-quality solutions for a number of specific (combinatorial) optimization problems, we will also investigate some of its applications providing useful insights for both researchers and practitioners.
Renata Mansini is a Full Professor in Operations Research at the Department of Information Engineering (DII) of the University of Brescia, Italy. She holds a PhD from the University of Bergamo and has been PhD Exchange Visitor Program P/1/153 at the Olin Business School at Washington University in St. Louis, Missouri. Her research focus is on combinatorial optimization models and algorithms. Her primary research interests lie in mixed-integer linear programming models and in the development of exact algorithms (branch-and-bound, branch-and-cut, branch-and-price), heuristic methods (local search methods, meta-heuristics, matheuristics), and approximation algorithms (computational complexity, worst-case analysis).
Renata Mansini has published in different application areas working on distribution logistics problems, vehicle and arc routing problems, procurement problems, financial risk and safety measures, linear programming-based portfolio optimization problems, knapsack problems, scheduling problems, and personnel rostering problems. She has founded and coordinates the group and the laboratory of Operations Research at DII, University of Brescia. She is an editor of the International Journal of Portfolio Analysis & Management and has been a member of the editorial board of Journal of Mathematics and ISRN Applied Mathematics. She is a member of the EURO Working Group on Combinatorial Optimization, the EURO Working Group on Vehicle Routing and Logistics, and the EURO Working Group on Commodities and Financial Modelling. She is chair of the events sub-committee of the EURO WISDOM Forum.
Prof. Dr. Sophie Parragh
Professor for Production and Logistics Management
Johannes Kepler University Linz, Austria
Title of Talk: Branch-and-Bound for Multi-Objective (Mixed) Integer Linear Programming: Key Ingredients, Challenges, and Motivating Applications
As promoted by the European Green Deal, policy makers and companies increasingly strive for minimizing environmental impact, in addition to other objectives such as keeping costs low or ensuring a high customer service level. Unfortunately, the minimum cost solution is rarely the best from an environmental perspective or from the perspective of customer service. Optimizing conflicting objectives concurrently results in a set of optimal trade-off or efficient solutions which have the property that neither objective can be improved without deteriorating at least one other objective. The image of these solutions in objective space is called the non-dominated frontier or Pareto front. The wide range of practical problems which can be modeled as mixed integer linear programs (MILPs), and involve more than one objective, motivates the development of generic exact methods as general purpose tools to solve them. In this talk, we first give a brief overview of recent advances in exact methods for solving bi- and multi-objective MILPs which compute at least one solution for each point on the Pareto front. They are commonly classified as either criterion space search methods, which work in the space defined by the objective functions, or as decision space search methods, which have been mainly generalizations of branch-and-bound algorithms. We then focus on the most recent successful branch-and-bound schemes, which do not exclusively work in the decision space. We discuss their key ingredients, such as bound set generation, branching rules, and primal heuristics. Finally, we highlight motivating applications in logistics, discuss open challenges and indicate promising directions for future research.
Sophie N. Parragh is Head of the Institute of Production and Logistics Management at the Johannes Kepler University Linz. She received her PhD from the University of Vienna in 2009 and completed her habilitation in 2016. She worked as a visiting researcher at the CIRRELT, Montreal and as a postdoc fellow at the IBM Center for Advanced Studies in Porto. She won a Hertha Firnberg Postdoc Fellowship (funded by the Austrian Science Fund (FWF)) and she was Visiting Professor at the Vienna University of Economics and Business (WU Wien). She serves as associate editor at INFORMS Journal on Computing and at OR Spectrum and she co-organizes the monthly webinar of the EURO Working Group on Vehicle Routing and Logistics (VeRoLog).
Sophie Parragh has been involved in several third-party funded applied and fundamental research projects in OR in health care, field staff routing and scheduling, production planning, and multi-objective combinatorial optimization (both as project leader/PI and project collaborator). Her main research focus is the development of exact and heuristic optimization algorithms. Currently, she is particularly interested in generic exact methods for multi-objective (mixed) integer linear programming, in emission-free vehicle routing, facility location, and short and medium term production planning under uncertainty.
Prof. Dr. Marc Uetz
Professor for Discrete Mathematics and Mathematical Programming
University of Twente, The Netherlands
Title of Talk: Network Routing and Beyond: Equilibria for Atomic Congestion Games
Congestion games are a rich and fundamental class of problems which lie at the core of the area algorithmic game theory, just like the TSP lies in the core of discrete optimization. One well-known example is the result by Roughgarden and Tardos, showing that the price of anarchy in network routing games with affine cost functions equals 4/3. The discrete version of the same problem, where each player chooses a single path instead of routing a flow, has a price of anarchy equal to 5/2. This 5/2 bound holds true for the more general class of atomic congestion games, where players choose arbitrary subsets of resources, while the cost of any resource increases with the number of players using it. There are interesting classes of atomic congestion games, however, which are not yet completely understood. The lecture addresses some open questions, along with some recent results in this context. We specifically consider games with restrictions of players’ strategy spaces, but also congestion games where players act sequentially. Some of the results are improvements on the known price of anarchy bounds, but sometimes also also counter-intuitive results where the quality of equilibria deteriorates.
Marc Uetz graduated in Mathematics from TU Berlin in 1997, and received his PhD in Mathematics from the same university in 2001. He was an Assistant and Associate Professor at the Department of Quantitative Economics at Maastricht University. Since 2007, Marc Uetz is the Chair of Discrete Mathematics and Mathematical Programming in the Department of Applied Mathematics at the University of Twente. He has been a Visiting Professor, among others, at the Max Planck Institute for Informatics, TU Vienna, Zhejiang University, IBM T.J. Watson, UC Berkeley, and MATH+ Berlin. His research is in discrete optimization and algorithmic game theory, specifically resource allocation, scheduling, mechanism design and the analysis of equilibria.
Marc Uetz serves as area editor for Operations Research Letters, and as associate editor for the journals Discrete Optimization, INFORMS Journal on Computing, and Journal of Scheduling. Since 2018, he is a member of the NWO Mathematics Roundtable. His research has been partially funded by grants from NWO, 4TU.AMI, and the Simons Foundation.
Prof. Dr. David Wozabal
Professor for Investment, Finance and Risk Management in Energy Markets
Technical University of Munich, Germany
Title of Talk: Short-Term Power Markets: Towards Optimal Trading Decisions
The talk explores optimal strategies for trading on short-term power markets. We take the perspective of a single firm that does not act strategically but treats market outcomes as exogenous and random. The problem is of high practical relevance for most players in the electricity sector and, correspondingly, there is an extensive literature on the subject. However, due to the high number of traded products and the increasing influence of variable intermittent production technologies which necessitates repeated rebalancing until shortly before delivery, the resulting decision problems are of considerable computationally complexity. In particular, finding optimal strategies for trading on continuous intraday markets remains a largely open problem as most authors consider only simplified versions of the market leading to policies that are not implementable in practice. We review the current state-of-the-art and discuss the specifics of different short-term markets and the resulting trade-offs associated with trading on them. We then go on to show that, for firms that operate on multiple markets, optimal policies for individual markets are interdependent and decisions therefore have to be coordinated. We demonstrate how stochastic optimization approaches can be combined with model predictive control to arrive at near optimal trading strategies on an hourly granularity and quantify the value of coordination between the day-ahead market and a continuous intraday market. To push the trading frequency to a sub-hourly level, we discuss a parametric weather-based trading heuristic based on intraday updates of renewable production forecasts. We evaluate the resulting decisions out-of-sample based on detailed order book level data.
David Wozabal studied Business Information Systems and Mathematics and obtained his doctoral degree in Statistics from the University of Vienna in 2008. After acquiring his postdoctoral teaching qualification (habilitation) in Business Administration at the University of Vienna (2012) he held a year-long visiting professorship at TUM (2012) and a short-term research post at Vienna University of Economics and Business (WU). In 2014 David Wozabal was given a faculty appointment in TUM’s newly established Center for Energy Markets.
The research of David Wozabal deals with the development of algorithms for stochastic optimization problems, risk measurement and risk management. The theoretical results of this research are applied to planning problems in energy management and classic problems in finance such as portfolio optimization. Another aspect of his research focuses on the structural problems of the European energy markets with particular regard to electricity markets. A major part of this work involves modeling price processes and examining the efficiency of electricity markets.
Talk of the Winner of the
2021 GOR Company Award
The winner will be announced during the Opening Session on Wednesday, September 1.