Presented by Professor Carlo Mannino (Professor at the Department of Mathematics, University of Oslo; Senior Research Scientist, OSLO)

Trains running through railway networks are monitored and guided in real time by train dispatchers. They control train movements by taking various relevant decisions, such as where trains should pass and meet each other, when trains should arrive or depart from stations, which routes they must take, etc. These decisions become crucial when one or more trains are delayed, in order to reduce knock-on effects and quickly regain punctuality.

In the past decades a significant research effort has been devoted to developing optimization models and methods for (automatic) real-time train rescheduling (dispatching). However, such academic effort did not translate into a widespread use of such techniques in the practice of train dispatching. To the best of my knowledge, only a few fully or partially autonomous dispatching systems are or have been rolled out over the years to support dispatchers. The mathematics and data technology required for supporting the practice, at least to a certain extent, have been approaching maturity for at least 10-15 years, and are possibly fully mature now. So, what has impeded a more widespread development and deployment of such systems? In the talk, I will try to formulate an answer to this question and also discuss some recent developments and “gamechangers” which show that the situation is rapidly evolving.

I will start by describing the basic mathematical models and methods for representing and solving train (re-)scheduling problems. I will then describe a new system to dispatch trains in the greater Oslo area, an important portion of the Norwegian railway network which includes the large Oslo central station and all the adjacent lines. Although not in operation yet, the system has the highest level of readiness (TLR 9) and is currently undergoing on-field testing by professional dispatchers at Oslo Control Center.

After this I will quickly review the few, real-life dispatching systems which (to the best of my knowledge) are or have been in operation for supporting dispatchers in practice.

I will end the talk by analyzing which were the main hurdles for the penetration of optimization methods into the practice of dispatching and why things are changing now. This discussion may be of inspiration for other sectors with similar obstacles (such as, for instance, air-traffic control).

3 SEPTEMBER 2021 WEBINAR

Presented by Professor Christina Pagel (Director of the Clinical Operational Research Unit, University College London)

In May 2020, I joined the group Independent SAGE. At the time, I thought I was signing up to one or two public meetings on YouTube, but instead our profile grew and we discovered there was a large public appetite for more information about Covid and its spread in the UK. From the summer onwards, I have been giving regular updates on the latest Covid situation in the UK during Independent SAGE weekly briefings and have been invited regularly on the media fielding questions about various aspects of the Covid pandemic. In this talk, I will reflect on how my experience of working in Operational Research applied to health care has shaped how I have understood and communicated the Covid pandemic over the past year. This ranges from thinking through problem definition and unintended consequences, combining knowledge across different disciplines to deciding on where the information lies in the data and how best to present it.

9 JULY 2021 WEBINAR RECORDING

Presented by Professor Joaquim Gromicho (Professor of Business Analytics, University of Amsterdam; Science and Education Officer, ORTEC)

I will tell you some stories that introduce mathematical optimization in python. Each story describes a problem and the solution will be demonstrated using standalone Jupyter notebooks which can be made available on request. The domains addressed include:

Optimization problems with analytical solution

Nonlinear continuous optimization problems

Mixed integer linear optimization problems

Coping with data uncertainty using robust models

Scalable mixed integer convex optimization expressed using second order cones

The illustrations will use python packages that are solver agnostic, meaning that you can easily replace the solver when needed. This can be the case when you start with a free solver but discover that you need a commercial one: replacing the solver will require no recoding of the model.

4 June 2021 WEBINAR RECORDING

4 June 2021 webinar presentation relevant information and links

Speaker prof. dr. Joaquim Gromicho is science and education officer at www.ortec.com and professor of Business Analytics at www.uva.nl, see https://abs.uva.nl/content/news/2021/01/improving-the-world-through-analytics.html

Presented by Dr. Cor van der Struijf, MBA (IBM Quantum Ambassador)

We experience the benefits of classical computing every day. However, there are challenges that today’s systems would never be able to solve. To stand a chance at solving some of these problems, we need a new kind of computing: Quantum Computing. During this session we will explore together what quantum computers are about, and how they might be applied to operational research and more specifically optimization problems. With this, you would be able to start exploring how quantum computing can help with your research.

7 May 2021 WEBINAR RECORDING

Presented by dr. Gyöngyi Kovács (Professor in Humanitarian Logistics, Hanken School of Economics)

Humanitarian logistics plays an important part in responding to disasters, crisis, but also epidemics and pandemics. This talk is about what we can learn from humanitarian logistics in tackling supply chain disruptions in pandemics. The talk also gives examples and shares results from current research projects on the COVID-19 pandemic, and also discusses the challenges with the ever-changing parameters of pandemic response.

9 April 2021 WEBINAR RECORDING

9 April 2021 webinar presentation

Presented by Tobias Andersson Granberg, Associate Professor at Linköping University

Emergency response systems, e.g. fire and rescue services and emergency medical services, have been a popular area of study for operational researchers. However, unlike for the airline industry for instance, there do not exist an abundance of success stories, where advanced operational research models and methods have been implemented in practice. In this talk, I’ll present a few emergency response problems, including the classical fire station location problem, ambulance dispatching and relocation, and volunteer management, where OR have been used in practice. I will also discuss some of the challenges with practical implementations, and the trade-off between working on something that is useful in practice, and something that can be published in an OR journal.

5 March 2021 WEBINAR RECORDING

Presented by Laurent Perron (Tech Lead Operations Research at Google)

CP-SAT won all gold medals in the Minizinc challenge in the tracks it participated in the last 3 years. It also proved 5 open problems in the MIPLIB 2017 suite, and improved bounds on a few more. In a sense, it realizes the old dream of having a good MIP solver and a good CP solver in the same engine. In this presentation, we will present how the SAT technology has enabled this merging of the two techniques in a competitive way, and concludes with the challenges and research opportunities in front of us.

5 February 2021 WEBINAR RECORDING

Presented by Prof. dr. Holger H. Hoos (Professor of Machine Learning at Leiden University)

In recent years, there has been a significant increase in the use of automated algorithm design methods, such as automated algorithm configuration and portfolio-based algorithm selection, across many areas within operations research, artificial intelligence and beyond. These methods are based on cutting-edge machine learning and optimization techniques; they have also led to substantial advances in those areas.

In this tutorial, I will give an overview of these automated algorithm design methods and introduce Programming by Optimization (PbO), a principled approach for developing high-performance software based on them. I will explain how PbO can fundamentally change the nature of developing solvers for challenging computational problems and give examples for its successful application to a range of prominent problems from OR and AI – notably, mixed integer programming, the travelling salesman problem, AI planning, automated reasoning and machine learning.

8 January 2021 WEBINAR RECORDING

Presented by Alain Nguyen (Combinatorial Optimization Expert at RENAULT)

RENAULT rolled out its first OR application in supply chain in 1992, with the very first version of CPLEX. It was a central planning tool for car production.

Since then, we tackled inbound and outbound transportation, car sequencing, truck and container loading. More recently, we moved to the shop floor to deal with scheduling, line balancing, operator assignment.

Car sequencing, Capacitated VRP, line balancing: the problems we face do not exactly match the academic models. There are always extra tricky constraints and objectives.

Historically, we developed OR applications that were used on a regular basis. More recently, we rolled out tools to carry one shot quick win studies.

Important challenges remain for full OR adoption: data availability in information systems (especially shop floor constraints), non- standard processes in the plants, OR visibility inside the company, lack of facilitators between the central OR team and the field-level operators.

4 December 2020 WEBINAR RECORDING

Presented by Professor Andrea Lodi (Canada Excellence Research Chair in ˝Data Science for Real-time Decision Making˝, Ecole Polytechiquie de Montreal)

In this talk, we cover some of the recent and exciting advances in the use of Machine Learning techniques for Combinatorial Optimization by highlighting and characterizing the major directions in which such use has been conducted.

6 November 2020 WEBINAR RECORDING