@incollection{AssafLoefflerHofstedt, author = {Assaf, George and L{\"o}ffler, Sven and Hofstedt, Petra}, title = {Optimized scheduling of medical appointment sequences using constraint programming}, series = {Integration of constraint programming, artificial intelligence, and operations research : 22nd International Conference, CPAIOR 2025 Melbourne, VIC, Australia, November 10-13, 2025 : proceedings, Part I}, booktitle = {Integration of constraint programming, artificial intelligence, and operations research : 22nd International Conference, CPAIOR 2025 Melbourne, VIC, Australia, November 10-13, 2025 : proceedings, Part I}, editor = {Tack, Guido}, publisher = {Springer Nature Switzerland}, address = {Cham}, isbn = {978-3-031-95973-8}, issn = {0302-9743}, doi = {10.1007/978-3-031-95973-8_1}, pages = {1 -- 16}, abstract = {We propose a novel constraint-based model to efficiently tackle the medical appointment sequence scheduling problem (MASSP), inspired by a real-world problem at Charit{\´e} - Universit{\"a}tsmedizin Berlin. In many practical medical scenarios, scheduling a sequence of appointments, rather than a single appointment, has become increasingly essential for patients undergoing multi-stage treatments. The goal of the MASSP is to identify a set of medical resources with sufficient, consecutive, and available time slots in their calendars to create a sequence of appointments for effectively managing a treatment plan. The problem comprises various constraints, including the availability of both the intended patient and required medical resources, as well as the time and resource dependencies among the individual appointments that constitute the sequence. To address this, we formulate the problem as a constraint optimization problem () that not only captures the basic constraints of the MASSP but also optimizes resource assignment to ensure a fair workload distribution within the medical facility. The results of our experiments demonstrate that the model performs effectively under diverse conditions, which confirms the utility and robustness of the proposed model in optimizing resource allocation and ensuring equitable workload distribution.}, language = {en} } @incollection{LoefflerHofstedt, author = {L{\"o}ffler, Sven and Hofstedt, Petra}, title = {A constraint-based savings algorithm for the Traveling Salesman Problem}, series = {Machine Learning, Optimization and Data Science : 10th International Conference, LOD 2024, Castiglione Della Pescaia, Italy, September 22-25, 2024, Revised Selected Papers}, volume = {2}, booktitle = {Machine Learning, Optimization and Data Science : 10th International Conference, LOD 2024, Castiglione Della Pescaia, Italy, September 22-25, 2024, Revised Selected Papers}, editor = {Nicosia, Guiseppe and Ojha, Varun and Giesselbach, Sven and Pardalos, M. Panos and Umeton, Renato}, publisher = {Springer Nature Switzerland}, address = {Cham}, isbn = {9783031824838}, issn = {0302-9743}, doi = {10.1007/978-3-031-82484-5_15}, pages = {203 -- 217}, abstract = {There exists a plethora of NP-hard problems for which obtaining an exact solution is computationally intensive while an approximate solution can be rapidly computed. The Traveling Salesman Problem (TSP) stands out as a prominent example within this problem class, being equally significant in both research and industry sectors. Exact methods for solving the TSP are typically infeasible within acceptable timeframes beyond a certain instance size. Due to the NP-hardness of the problem, computational time requirements would quickly escalate to weeks, months, years, or even millennia. In contrast, heuristic approaches often yield sufficiently good solutions rapidly, typically within seconds or minutes. However, the drawback lies in their tendency to produce only locally optimal solutions, which may deviate significantly from a global optimum. Nevertheless, heuristic methods remain preferred in industry due to their ability to find satisfactory solutions within reasonable timeframes. It is worth noting that even marginal improvements in solutions can translate into significant cost savings in practice. Hence, there is a demand for methods capable of swiftly providing high-quality solutions while also continuously searching for further, better solutions over time. This work aims to combine two approaches: heuristic search (utilizing two simple greedy algorithms and the savings algorithm developed by Clarke and Wright) and exact search (constraint programming), to solve TSPs. Using various TSP instances, we demonstrate that our approach yields a better solution than purely local methods in at least 80\% of the cases. Furthermore, within the given time limit of 5 min, the new approach identifies a solution that is more than twice as good compared to traditional COP methods.}, language = {en} }