Bring your company case
Arc routing: applications in salt spreading, garbage collection and unmanned aerial vehicles (UAV, drones) task planninh. MILP and heuristics
Multidimensional bin packing to allocate pods in nodes within kubernetes
Integrated healthcare timetabling
Scheduling and Routing
Tighter formulations for TSP with Konstantin Pavlikov
Student Project Assignment with two-sided preferences. Related to the stable marriage problem. Extension of an existing tool that consider preferences only from the side of the students.
Capacity Expansion in Energy Production in collaboration with Energinet. Large scale optimization for long term decision making, which plants is best to construct and where, which energy source is likely to give the best performance of the overall system? The optimization problem includes both discrete and continuous variables as well as uncertainty issues.
Sport analytics: analysis of soccer data in collaboration with Divisionsforening, DBU and SDU Idræt Institute. Data available: Tracking (25 data per second) + event data: data preparation, alignment, search, pattern mining.
After a successful collaboration that resulted in the Master Thesis “Automated Tools for Detecting Mistakes and Frauds in Annual Tax Assessments”, the Danish Tax Office (Skatte Styrelsen) proposes a continutation of the project on a related topic. In the new project, we should investigate which elements contribute to extended handling time for the cases handlers and extended waiting time for the citizen when the citizen asks for a decision to be reconsidered by the Danish Tax Office. At the beginning of the project in August, we will have a dataset consisting of multiple thousand examples of time for case handling and waiting time, connected to the annual tax report from the citizen. The project should focus on building a causal explainable ML model to evaluate which factors contribute to longer time consumption. Further, we expect to be able to access more data in September which would also include the complaint from the citizen in free text in the dataset as well as attached documents from the citizen. We hypothesize that long complaints with many attachments are more complicated cases which take longer time, but it might be that certain themes in the text are the complicating aspects or that long well documented complains are faster to handle since all material is already present.
The HR department of the Danish Tax Office will need to receive a cv and a criminal record from the student interested in the project. The student will then be connected to the Danish Tax Office as an intern and receive all data, a computer, and a server to make the calculations on. The server runs both a R and a Python interface. The data is confidential, and raw data can therefore never leave the server. Anonymized data can be used in the thesis for public examination. We will expect the student to be physically present in our office in Ribe a few times during the term (when receiving the IT equipment among other), but there will often be a lift available from a coworker from Odense. Although the student can be provided with a desk at our office in Ribe, it is entirely up to the student when they will be present, with a few exceptions.
Please do not hesitate to contact Marco Chiarandini for further information.
Transport Optimization. Bus line planning and/or estimation of origin destination demand with data from the city of Odense.
Bus Map Drawing
Research on Graph Coloring. Instance space analysis and generation. Mario A. Muñoz and Kate Smith-Miles. “Generating New Space-Filling Test Instances for Continuous Black-Box Optimization”. eng. In: Evolutionary computation 28.3 (2020), pp. 379–404. ISSN: 1063-6560. Kate Smith-Miles and Simon Bowly. “Generating new test instances by evolving in instance space”. In: Computers Operations Research 63 (2015), pp. 102–113. ISSN: 0305-0548. DOI: https://doi.org/10.1016/j.cor.2015.04.022. Nikolaus Hansen, Anne Auger, Dimo Brockhoff, Dejan Tušar, and Tea
Education Management Tools
Student sectioning. Starting material: Mads’ speciale, articles.
Course Timetabling: exact algorithms (max sat, cp, milp) or black box heuristic solvers
Multiple objective solvers for timetabling
Fairness in Timetabling. See: talk by John Hookoer; tutorial or report
Optimize Binary Neural Networks by heuristics.
Comparison of local search solvers: local solver, paradiseo, oscar
General Local Search Solver Development. Constraint Based Local Search.
Automatic Algorithm Configuration (with Jacopo Mauro)
AI for Good. Artificial Intelligence for Computational Sustainability
Postnord. Daily demand prediction or Route optimization or 3D vehicle packing. Contact and discuss.
Predictive maintanance at Sanovo or other companies.
Image processing: Dexterity test assessment in children. Automatically assess the goodness of line drawed by children.
AI for Teaching and Learning
Develop an optimization game for educational purposes. The problem could be portfolio optimization or timetabling or others. See beer game and burrito game at Gurobi for examples.
LLM for report classification and annotation
Automated Feedback and Grading
