Some projects

BESSEMANS, Pauline (2023 – …)

Title: Mathematical optimization methods for a dual bin packing problem applied to wood reuse

Abstract: The increasing demand for raw materials such as wood contributes to the depletion of natural resources and global warming. To curb this phenomenon, a more sustainable and circular use of wood could be developed through the intelligent management of wood waste. Recycling wood waste has several environmental and economic benefits. Among these, wood is an important CO₂ sink and not burning it allows this carbon to be stored for longer. This work focuses on a way of reducing wood waste, which aims to reuse wood rather than recycle it, i.e., not to reduce the value and properties of the material as would be the case if the wood were recycled into pellets, for example. The objective here is to produce CLT panels by assembling, gluing, and cutting wooden slats, which is a variant of the dual bin packing problem.

Supervisor: PAQUAY, Célia

DOSSE, Benjamin (2024 – … )

Title: New market designs for renewable electricity trading through mathematical optimization

Abstract: The increasing penetration of intermittent renewable energy sources has had a significant impact on the Europe market mechanism for electricity trading. The European market has been designed for controllable energy source, in which the production capacity is known in advance. The production capacity of intermittent energy source is known only on delivery time, after the trading takes places, inducing billions of euros of loss in the balancing market. This also brings new challenges on the European transmission network that cannot support the production load when renewable sources production at 100%. We are studying new local market designs adapted to intermittent energy sources and taking into consideration transmission constraints.

Supervisor(s): Jérôme De Boeck, (co-supervisor : Bernard Fortz)

LELOUP, Emeline (2020 – …)

Title: A three-dimensional loading vehicle routing problem with split pickups, promised service times and time windows under real-time disruptions

Abstract: The first-mile collection problem, which includes collecting and transporting products, has an increasing share in the intricacy of supply chains. Complex challenges must be considered such as time windows, time-dependent travel durations, maximum working duration, cargo stability and reachability. Moreover, parcels collection tends to be more uncertain than delivery as additional pickup requests may pop up, or the number of boxes to collect can change during the day. The pickup process, in which the vehicle is initially empty, is much more able to react to these types of disruptions on-the-fly. The scientific community has devoted a lot of attention to delivery, while the specifics of the pickup process have been substantially overlooked. We study this challenging optimization problem.

Supervisor(s): PAQUAY, Célia; PIRONET, Thierry

MAHARANI, Anisha (2020 – …)

Title: Integrated operational scheduling for hospital-at-home

Abstract: Hospital-at-home (HaH) programmes provide an alternative to acute hospitalization by delivering care in patients’ homes. This work proposes metaheuristic solution methods for a weekly HaH decision-making process in which patient admission, nurse routing, and rerostering decisions must be made simultaneously. To maintain the continuity of ongoing treatments and compliance with working hour regulations, this work incorporates information from consecutive weeks on nurse rosters and already admitted patients. In a first study, we consider admission decisions for the eligible patients of a given week, while in a second study, we also account for stochastic information available for the eligible patients of the coming week.

Supervisors: ARDA, Yasemin; FRANÇOIS, Véronique

Nelie Laura MAKENNE (2023 – …)

Title: Towards Green AI: Compressing Large Language Models for Sustainable Deployment

Abstract: Recent advancements in large language models have achieved state-of-the-art performance across a range of natural language processing tasks. However, their substantial computational, memory, and energy requirements limit their deployment in resource-constrained environments. This thesis proposes a compression framework based on singular value decomposition (SVD) to identify informative model parameters, facilitating effective low-rank approximations that reduce redundancy while preserving performance. A layer-wise compression strategy is introduced to regulate the compression process and adaptively adjust model components according to their sensitivity to dimensionality reduction. The environmental impact is assessed through a comprehensive carbon footprint analysis.

Supervisor(s):  Ashwin ITTOO

TONKA, Jenny (2021 – …)

Title: Air Transport Ground Operations Management: Digital Twin Building Blocks for Real-Time Monitoring and Process Optimization

Abstract: Transporting billions of passengers and millions of tons of cargo each year, the air transport industry has become a cornerstone of the global economy. However, the sector must address the complex logistical challenges posed by surging global demand and an increasingly challenging environment. This thesis specifically focuses on ground process operational efficiency, aiming to develop real-time monitoring and optimization strategies for both outbound baggage ground handling and air cargo ground operations. Leveraging the innovative technological framework of digital twins, as well as operations research and machine learning techniques, the thesis develops novel mathematical models, original heuristics, and tailored parametrization protocols for existing solution methods.

Supervisor: SCHYNS, Michaël