05-09-2024 Jeroen Mulder - Operations Research Challenges inside Air France-KLM
In my presentation, I will provide a high-level overview of our airline planning process and our efforts to optimize each step within this process. This process ranges from long-term fleet planning to network planning, flight scheduling, day-of-operations planning, and managing disruptions during the day.
Our planning efforts are challenging due to varying timelines, numerous stakeholders, and the stochastic nature of our processes. I will highlight the main Operations Research challenges we currently face that remain unsolved due to computational or organizational constraints. More importantly, I will discuss our current strategies for managing these challenges on the day of operations.
07-10-2024 Jeroen van Kasteren - Allocation of persons to parallel queues, the case of mental health homes
Persons who cannot live independently can make use of long-term residential facilities. The matching process between client and facility is complex, as both restrictions and client preferences play a crucial role. In this talk, the aim is to shed some light on how to allocate persons to different queues. For small instances, an optimal allocation can be achieved using a Markov Decision Process formulation, but this is more involved for larger instances. We illustrate this for the complex setting of persons with severe mental illness.
24-10-2024 Mathijs Pellemans - Estimating Vessel Arrival Times in Global Supply Chains: A Historical Maritime Data Approach
This study aims to improve the reliability of vessel estimated arrival times (ETA) predictions, which are essential for optimizing port operations and improving supply chain management. We use historical satellite AIS data and advanced machine-learning techniques to predict vessel arrival times. To do so, we use a Markov chain to learn the patterns of vessels across ports and compute its performance in predicting vessel destinations across 120 ports distributed over four continents. In addition, we develop a method to detect waypoints and compress the AIS trajectories such that the solution becomes scalable in real-world settings.
31-10-2024 Magnus Botnan - An introduction to topological data analysis.
This presentation provides an accessible introduction to topological data analysis (TDA), with a focus on persistent homology and non-linear dimension reduction techniques inspired by topology.
07-11-2024 Wang Yu - Joint inventory & pricing model solved with ML-enhanced branch & price algorithm
With the trend of mass customization and smart manufacturing, the automotive industry is under a rapid transition to improve its responsiveness and handle highly diversified customer orders while reducing the cost of automotive inbound logistics. To address this challenge, this paper proposes a new variant of the multi-period inventory routing problem, which focuses on coordinating discrete time-varying demands of auto parts on the assembly line and predetermined packages of the parts at the suppliers over a finite short-term time horizon (e.g., on an hourly basis). The goal is to minimize the total transportation and inventory cost while making aperiodic collecting quantities and routing decisions simultaneously for an inbound warehouse next to the assembly plant. A time-indexed integer programming model is formulated for the problem to analyze the feasibility conditions. Then, a reformulation is designed to make the problem more tractable, based on which a novel machine learning enhanced branch-and-price algorithm (BPL) is proposed, where learning-based prediction cuts are embedded to accelerate the pricing procedure. Experimental results based on real-scale random instances show that the proposed algorithm can always generate near-optimal solutions with an average gap to the best lower bound equal to 4.42% and save over 90% of the average calculation time compared to solving the integer programming model directly by CPLEX. The proposed learning technique is also computationally efficient, capable of shortening the total calculation time by 13% on average. Our work supports timely decision-making and provides new insights into the multi-period inventory routing problem involved in inbound logistics.