AOO 2026 - Monday 27th of April 2026
- Jan 28
- 4 min read
Updated: Feb 23
DORS is delighted to welcome you to the annual conference, Applications of Optimization 2026.
It is great opportunity to hear our 5 speakers regarding the application of Operation Research in the industry and the latest development in academy in this field.
Andrea Cassioli, Lead Decision Scientist at Maersk
Julian Hall, professor at Universty of Edinburgh and founder of HIGHS (Open source solver)
Scott Egan, Leading Advanced Analytics at Abacus Medicine
Get ready for a day filled with inspirations, networking time and fun. We can’t wait to see you there!
Address:
Industriens Hus
H. C. Andersens Blvd. 18, 1553 København (map)
Thanks to our sponsors for support our community:
Programme :
09:00 – 10:00 Registration and refreshments
10:00 – 10:15 Welcome and announcements
10:15 – 11:00 1st speaker
11:15 – 12:00 2nd speaker
12:00 - 13:00 Lunch
13:15 – 14:00 3rd speaker
14:15 – 15:00 4th speaker
15:00 - 15:30 Coffee break
15:30 – 16:15 5th speaker
16:15 – 16:30 Closing
As our tradition, after the conference, all participants can continue to talk around a glass at the restaurant Bryggeriet. It is always a very good time. We hope to see you there too.
Abstracts
Title: Optimizing Maersk’s Network: How Operations Research Powers Container Logistics
Speaker: Andrea Cassioli, Lead Decision Scientist at Maersk
Abstract:
Maersk is one of the world’s largest container shipping companies, moving millions of containers across the world, using hundreds of vessels. Planning such operations is complex due to the scale of the network, execution uncertainties, and the need to comply with numerous business rules and regulations. With Maersk’s strategy focused on becoming a global integrator of container logistics and its ambitious goal of achieving net-zero emissions by 2040, these challenges are only intensifying.
To address this, Maersk has developed a suite of Operations Research (OR)-based decision support tools that assist both network designers and front-line operators. In this talk, I will highlight key challenges faced by network designers and demonstrate how these tools enable efficient operations across Maersk’s global network.
Title: HiGHS: From gradware to software and impact
Speaker: Julian Hall, professor at Universty of Edinburgh and founder of HIGHS (Open source solver)
Abstract
Since it was founded in 2018, by combining two linear programming (LP) solvers written by Edinburgh PhD students, HiGHS has become the world’s best open-source linear optimization software, with a user base from solo academics to multinational companies. This talk will give an overview of its creation, with an insight into the techniques underpinning its linear programming, mixed-integer programming and quadratic programming solvers. Situations where it is preferable to use HiGHS rather than commercial software will be discussed. Finally, you will learn about its community of developers and users, its funding, and our vision for the future.
Title: Deep generative scenario search: Bridging machine learning and stochastic optimization
Abstract
Many decision-making problems involve uncertainty in parameters such as costs, demand, and processing times. A common way to address this uncertainty is to formulate the problem using a set of possible future scenarios and to select the decision that minimizes the expected cost across these scenarios. However, adequately representing the range of possible future outcomes often requires a large number of scenarios, which can result in highly complex models that are computationally difficult to solve.
Deep Generative Scenario Search (DGSS) addresses this challenge by combining a deep generative model with an iterative, objective-guided search over sets of scenarios. The method can be viewed as a black-box optimization loop operating on the scenario generator. In each iteration, a candidate scenario set is generated, the corresponding stochastic optimization problem is solved, and the resulting decisions are evaluated against the full empirical scenario set. The evaluation outcome is then used to guide the generation of improved scenario sets in subsequent iterations.
Computational experiments are conducted on several NP-hard problems, including transportation and facility location problems. Across these instances, DGSS consistently produces high-quality solutions, avoids infeasible second-stage decisions, and exhibits low variability in solution quality. Because the method is iterative, it generates a trajectory of candidate solutions, which can subsequently be evaluated with respect to additional soft or practical criteria.
Beyond identifying near-optimal first-stage decisions, the algorithm also produces a compact set of realistic future scenarios that support and justify these decisions. By reducing the number of scenarios while preserving decision quality, DGSS enhances interpretability and transparency, thereby improving the explainability of the selected solutions.
Title: Learning To Stow: Improving stowage planning with RL & Mathematical Optimisation
Abstract
Roll-on/Roll-off (RoRo) stowage planning is challenging both mathematically and operationally: it combines hard combinatorial decisions with safety, stability, and tight port-time constraints. This talk presents the problem through both an industrial and an academic perspective.
First, we describe how RoRo stowage planning is performed and why deploying optimisation in practice is difficult. We highlight operational realities such as incomplete or late-arriving information, different vehicle attributes, rule-based constraints, and the difficulty of connecting all the data.
Second, we connect these realities to recent advances in neural combinatorial optimisation and reinforcement learning. We explain what makes RoRo stowage combinatorially hard and show how learning-based methods can complement classical OR by learning effective heuristics and producing feasible stowage plans.
Title: coming soon
Speaker: Scott Egan, Leading Advanced Analytics at Abacus Medicine
Abstract
coming soon
Comments