Smart schemaläggningssystem för optimal personalhantering

Employee scheduling in the healthcare and social care sector is a complex challenge that must balance legal regulations, employee preferences, and unexpected absences such as sick leave. Manual scheduling is still prevalent in many organizations, leading to inefficient processes, staff overload, and dissatisfaction. This thesis addresses the need for a smart scheduling system that automates shift allocation while allowing for manual adjustments, with the aim of improving efficiency, fairness, and legal compliance. The significance of this problem lies in its impact on both staff well-being and organizational efficiency. Automating the process has the potential to reduce administrative workload and mitigate issues like uneven shift distribution. Although several commercial tools exist, they often lack flexibility or are inaccessible to smaller organizations. Thus, this project seeks to create a viable alternative using open-source technologies. The solution involves the development of a full-stack web application that integrates Timefold, a Java-based AI constraint solver, to handle both hard and soft constraints. They’re distinguished by their flexibility, where hard constraints are things like legally mandated work-hour limits and rest periods that must always be followed. On the other hand, soft constraints—such as employee preferences and availability, may be violated if necessary to produce a feasible schedule. The system was built using Supabase for real-time database interaction, Quarkus for backend logic, and React for the user interface. The interface supports drag-and-drop scheduling, error validation, and real-time feedback. System validation included constraint testing, solver performance analysis, and limited user interface evaluations. The system successfully generated legally compliant schedules and respected employee preferences when feasible. It also prevented errors like over-assignment via real-time feedback. However, limitations such as the inability to support shifts spanning two days, absence of multithreading, and infrastructure-related delays were identified. The outcome is a functional prototype that demonstrates the potential of combining AI planning tools with user-centered design to improve scheduling in care environments. Future work includes full-scale usability testing, extended feature support, and optimization for larger datasets.