Speaker: Sameela Wijesundara – Monash University

Abstract:

Even when fully adopted and deployed, optimisation systems still face significant challenges. In particular, most optimisation systems suffer from one or more forms of drift over time, in which changes to the requirements of the associated optimisation problem – such as modifications to input or output data, constraints, or objective function – occur over the lifetime of the system. Currently, updating the constraint model to address problem drift requires an iterative modify-and-test process that is largely dependent on modelling experts.

In this talk, I first introduce the concept of problem drift and discuss the effects of problem drift using a real-world case study that we tackled for the UNHCR. I then present a new approach based on Modelling Domain-Specific Languages (MDSLs) to address problem drift by giving domain experts the flexibility to perform controlled modifications of parameters, constraints, and objective functions. This allows modelling experts and domain experts to collaboratively maintain constraint models. Further, I discuss how MDSLs developed for broader domains can enable code reuse and support rapid prototyping, allowing modelling and domain experts to co-develop models more quickly using pre-defined MDSLs. In doing so, I also establish an extended model development phase to mitigate the challenges that hinder the adoption of optimisation systems and to enhance their maintainability once adopted through the combined use of CERDs and MDSLs.

This event is Hybrid:
JOIN VIA ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635
20 Exhibition Wlk, Clayton-2-279-Meeting Room

SEMINAR: WED 22 JULY 2026 4.00PM-5.00PM (AEST, Melbourne Time)

Speaker: Sameela Wijesundara – Monash University

Abstract:

Optimisation problems are everywhere. These problems require finding a combination of choices (i.e., a solution) that satisfies a set of constraints and, optimises an objective function. Owing to their societal importance, computational approaches for solving optimisation problems have improved significantly. However, despite significant advances in constraint problem modelling and solving technology, the adoption of this technology in real-world decision-support systems remains lower than expected.

In this talk, I discuss some of the common barriers to the adoption of constraint technology, as well as additional challenges identified through my thesis. To address these challenges, I present a semi-formal visual specification technique designed to enhance communication between domain experts and modelling experts, support shared understanding, and facilitate collaborative constraint model development. This technique, termed Constraint Entity-Relationship Diagrams (CERDs), serves as an intermediate representation between the constraint problem and its model. Further, CERDs can serve as a living document that is used to elicit problem requirements and to provide a stable reference and a form of agreement between domain experts and modelling experts. CERDs are extensions of the well-known ERDs used for database design during the requirements specification phase of the software engineering process. The talk concludes by sharing insights gained from a series of user studies conducted to evaluate the CERDs across seven real-world case studies affiliated with Monash University. These studies enabled close collaboration with both modelling experts and domain experts, providing valuable evidence of the potential of CERDs to support more collaborative approaches to constraint model development.

This event is Hybrid:
JOIN VIA ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635
20 Exhibition Wlk, Clayton-2-279-Meeting Room

SEMINAR: WED 15 JULY 2026 4.00PM-5.00PM (AEST, Melbourne Time)

Speaker: Hamideh Anjomshoa  – Gurobi/OPTIMA

Abstract:

Every optimisation model is built to find answers. But in practice, models sometimes return a different result: infeasibility, a signal that the constraints you’ve defined cannot all be satisfied at once. For decision-makers, this is rarely just a technical problem. It often means a planning system has gone silent at exactly the moment a decision is needed.
This presentation offers a practical framework for understanding, diagnosing, and recovering from infeasibility in real-world optimisation models. We walk through how to identify the root cause using Gurobi’s Irreducible Infeasible Subsystem (IIS), a tool that isolates the smallest conflicting set of constraints, and how to restore a workable solution using feasibility relaxation. Along the way, we share the latest solver capabilities, common traps practitioners fall into, and a set of hands-on tricks for navigating infeasibility faster and with more confidence.

Whether you are building optimisation models, overseeing systems that depend on them, or conducting research at the intersection of mathematical programming and real-world decision-making, this talk offers practical tools and clear intuition you can take away immediately.

Bio:
Hamideh Anjomshoa is a Decision and Data Scientist with broad experience working across industry and academia. Currently, she is a Technical Account Manager at Gurobi Optimization, where she helps organisations across Australia and New Zealand solve large-scale industrial optimisation problems. She received her PhD in Mathematics from the University of South Australia in 2011, and has held roles as Associate Professor at the University of Melbourne and Senior Research Scientist at IBM Research Australia. Her work sits at the intersection of mathematical optimisation, machine learning, and data-driven decision-making, with applications in healthcare, mining, and transport. Hamideh is an Associate Investigator at OPTIMA and has contributed to the Australian and international optimisation community through peer-reviewed publications, patents, and industry collaborations.

—

This event is on ZOOM only:
JOIN VIA ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635

SEMINAR: WED 01 JULY 2026 4.00PM-5.00PM (AEST, Melbourne Time)

Speaker: Tea Tušar – Jozef Stefan Institute

Summary:
COBI is a new generator of scalable benchmark problems for COnstrained BI-iobjective optimization, designed to balance realistic properties with analytical tractability. It produces problems that capture key challenges of real-world scenarios, such as multimodality, ill-conditioning, and disconnected feasible regions, while still providing known Pareto sets and fronts. The construction of COBI problems is based on strictly convex-quadratic objectives and their multipeak extensions, combined with linear, convex-quadratic, or multipeak constraints, enabling analytical characterization and efficient computation of the Pareto set. In addition to introducing COBI problems and their properties, we will examine six strategies for approximating their Pareto sets, comparing them in terms of quality indicators and computational cost, and highlighting the trade-offs between accuracy and efficiency.

Bio:
Tea Tušar is a senior research associate at the Department of Intelligent Systems at the Jožef Stefan Institute and an assistant professor at the Jožef Stefan International Postgraduate School. She received her PhD for her work on visualizing solution sets in multi-objective optimization. Following her doctorate, she completed a postdoctoral fellowship at Inria Lille, France, where she contributed to benchmarking multi-objective optimization algorithms. Her work focuses on Evolutionary Computation, with particular emphasis on the visualization and benchmarking of evolutionary algorithms for single- and multi-objective optimization, including constrained scenarios, as well as their application to real-world problems.

She has participated in several collaborative projects involving the application of optimization techniques to problems such as electric motor design, energy scheduling, and tunnel alignment.

She has contributed to the development of the COCO platform (https://coco-platform.org/) for comparing optimization algorithms, expanding its capabilities to handle multi-objective and mixed-integer problems. She has held organizational and editorial roles at international conferences, including GECCO, PPSN, and EMO, and serves as an associate editor for Evolutionary Computation and ACM Transactions on Evolutionary Learning and Optimization. Since 2025, she has been serving as President of ACM Slovenia.

—
This event is Zoom only:
JOIN VIA ZOOM WITH MEETING ID: 873 1557 5255; PASSWORD: 778635

SEMINAR: WED 15 APRIL 2026 16:00-17:00 (AEST, Melbourne)/ 08:00-09:00 (CEST)

Speaker: Associate Professor Marcus Gallagher – University of Queensland

Abstract:
Fitness landscape analysis refers to a broad area of work concerned with understanding, modelling and utilizing the properties of the solution spaces of optimisation problems. This research includes both discrete and continuous problem instances and ranges from understanding the theoretical properties of landscapes, through to improving benchmarking of optimisation algorithms, as well as practical techniques that produce features for downstream use in automated algorithm selection and configuration methods. In this talk I will give a brief overview of the main ideas in this area before discussing some selected examples of techniques that have been developed specifically for understanding and modelling fitness landscapes, with an emphasis on continuous landscapes. Some of the main challenges in this area will be highlighted, leading to possible directions for future research.

Bio:
Marcus Gallagher received this Ph.D. degree in computer science from the University of Queensland, Brisbane, Australia in 2000. He is currently Associate Professor with the School of Electrical Engineering and Computer Science, University of Queensland and an Associate Investigator in OPTiMA. His research interests are in Artificial Intelligence, including evolutionary computation, black-box optimisation, algorithm benchmarking, machine learning and real-world applications. His contributions include the development of continuous Estimation of Distribution Algorithms, techniques problems for benchmarking optimisation algorithms and problem landscape analysis. Dr. Gallagher has co-authored over 120 papers and is an Associate Editor of the Swarm and Evolutionary Computation Journal and Chair of the Queensland Chapter of the IEEE Computational Intelligence Society.

—
This event is Zoom only:
JOIN VIA ZOOM WITH MEETING ID: 873 1557 5255; PASSWORD: 778635
SEMINAR: WED 06 MAY 2026 16:00-17:00 (AEST, Melbourne)

**Rescheduled from 22 April due to technical issues

Speaker: Maarten Flippo – TU Delft

Abstract:
Modern combinatorial solvers can be understood as searching for proofs of unsatisfiability or optimality. The proof system implemented by a solver, therefore, fundamentally shapes solver performance. While propositional resolution is simple and complete for propositional formulas, no existing practical resolution-based system offers unrestricted complete reasoning over integer linear inequalities.

We introduce hypercube linear resolution, a new proof system that is both sound and complete for integer linear reasoning. Hypercube linear resolution integrates propositional resolution with Fourier resolution through a new constraint type, the hypercube linear constraint, which captures linear relations within a discrete hypercube.

In this talk we will introduce the hypercube linear constraint, as well as the resolution operators we defined. Without going too deep into the theory, we will go through some examples where hypercube resolution improves over propositional resolution. Additionally, we look at what future work is most promising to make hypercube linear resolution truly practical.

Bio:
Maarten Flippo is a fourth year PhD student at TU Delft in the Netherlands. His research is focused on constraint programming solvers, and in particular the proof systems they use to solve real world problems. Concretely, he is motivated in improving user confidence in the correctness of CP solvers, as well as explore different proof systems to increase solver performance. As part of the PhD, Maarten is one of the core contributors of the Pumpkin constraint programming solver.

—

This event is Hybrid:
JOIN VIA ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635
ATTEND IN PERSON – Room 279, Alan Finkel Building, 20 Exhibition Wlk, Clayton, Monash University 

SEMINAR: WED 13 MAY 2026 16:00-17:00 (AEST, Melbourne Time)

Speaker: Harvey Zhang – University of Melbourne

Bio:
Harvey is undertaking an OPTIMA Industry PhD project titled “Optimization of Production by ‘Selected Turbine De-Rating’ at Macarthur Wind Farm” with AGL. His research interests revolve around optimization, graph theory, and game theory. The project aims to investigate wind turbine control strategies to maximize power output, addressing challenges posed by significant wake effects. He is undertaking this research at the University of Melbourne supervised by Dr. Mario Andrés Muñoz, Prof. Michael Kirley and Prof Kate Smith-Miles AO.
Prior to this, he completed his Bachelor’s and Master’s degree in Mathematics and Statistics at the University of Melbourne.

—

This event is Hybrid:
JOIN VIA ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635
ATTEND IN PERSON – Level 8, Room 8109, Melbourne Connect, 700 Swanston Street, Carlton 3053
Please arrive a little early to find a seat

SEMINAR: WED 20 MAY 2026 16:00-17:00 (AEST, Melbourne Time)

Speaker: Dr Ye Wang – OPTIMA/University of Melbourne

Abstract:
Optimizing pump operations is a challenging task for real-time management of water distribution systems (WDSs). With suitable pump scheduling, pumping costs can be significantly reduced. In this talk, we introduce an economic model predictive control (EMPC) framework for real-time operational management of WDSs. Optimal pump operations are selected based on predicted system behaviour over a receding time horizon with the aim of minimising the total pumping energy cost. Time-varying electricity tariffs are considered while all the required water demands are satisfied. In addition, we will discuss some theoretical guarantees on EMPC. We will show some results with WDS benchmark to demonstrate the effectiveness of the EMPC.

Bio:
Dr Ye Wang is an ARC DECRA Fellow and Lecturer in the School of Mathematics and Statistics at the University of Melbourne. He received the PhD degree (Cum Laude) in Automatic Control, Robotics and Vision from Universitat Politècnica de Catalunya-BarcelonaTech (UPC) in Spain. He was awarded the Best PhD Thesis in Control Engineering Award 2019 from the Spanish National Committee of Automatic Control and Springer, the UPC Special Doctoral Award 2020 in the field of Industrial Engineering, and a Discovery Early Career Research Award (DECRA) 2022 from the Australian Research Council. His current research interests include model predictive control, optimisation and learning-based control with application to water-energy systems and autonomous systems.

—

This event is Hybrid:
JOIN VIA ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635
ATTEND IN PERSON – Level 8, Room 8109, Melbourne Connect, 700 Swanston Street, Carlton 3053
Please arrive a little early to find a seat

SEMINAR: WED 27 MAY 2026 16:00-17:00 (AEST, Melbourne Time)

Speaker: Hasnain Ali  – Nanyang Technological University

Abstract:
Airport departure scheduling is a combinatorial optimisation problem complicated by uncertainty, human behaviour, and real-time constraints. In this talk I will present I-MATE (Intelligent Departure Metering Advisory Tool), a decision-support system that integrates deep reinforcement learning with an air traffic controller-facing interface to optimise aircraft pushback scheduling at Singapore Changi Airport. Beyond the algorithm, I will discuss what building and validating such a system actually involves: why controllers deviate from AI recommendations, what those deviations reveal about human-AI collaboration in safety-critical settings, and what remains unsolved. The talk bridges optimisation methodology with the practical, human realities of getting research into production.

Bio:
Hasnain Ali is an operations research and applied machine learning researcher, recently relocated to Melbourne from Singapore. He was a Postdoctoral Research Fellow at the Air Traffic Management Research Institute, Nanyang Technological University, where he led a national research programme in airport operations automation. His research focuses on combinatorial optimisation and machine learning for real-time scheduling in safety-critical transport systems.

—

This event is Hybrid:
JOIN VIA ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635
ATTEND IN PERSON – Join us at 3:30pm for a pre-seminar afternoon tea. Register your in-person attendance at: https://go.unimelb.edu.au/ygw2
Level 8, Room 8109, Melbourne Connect, 700 Swanston Street, Carlton 3053

SEMINAR: WED 03 JUNE 2026 4.00PM-5.00PM (AEST, Melbourne Time)

Speaker: Hendrik Bierlee – KU Leuven/OPTIMA

Abstract:
Global constraints are a central concept in Constraint Programming (CP), which allows modellers to compactly express complex relations, and allows solvers to efficiently handle them. Table constraints have especially been well-studied as they can express arbitrary finite relations, and are extensively used in CP benchmarks. In this paper we study how to best deal with table constraints when using Integer Linear Programming (ILP) solvers. We study two paradigms: linear encodings, and a lazy cut generation approach. For the encoding we propose a novel Boolean decomposition, as well as an MDD-based flow encoding. For the cut generation, in which lazy constraints are generated on-demand during branch-and-cut search, we investigate different ways of generating such integer and fractional cuts as well as how to strengthen them through shrinking and cut lifting. We experimentally compare the different approaches on CP competition instances with a wide variety of table constraints, showing clear benefits over the standard integer encoding.

This talk presents work from a project authoured by:
Hendrik Bierlee (KU Leuven/OPTIMA)
Wout Piessens (KU Leuven)
Tias Guns (KU Leuven)
Peter Stuckey (Monash University/OPTIMA)

—
JOIN VIA ZOOM – MEETING ID: 873 1557 5255; PASSWORD: 778635

SEMINAR: WED 10 June 2026 16:00-17:00 AEST (Melbourne Time)/ 08:00 CEST