Project Abstract

In most medium- and large-sized schools worldwide, classroom scheduling presents a recurring critical challenge. Since schools must deliver a wide range of subjects each semester (typically 20-30 subjects across multiple student cohorts), preparing schedules is a demanding task. Teachers' availability and environmental factors impose strict constraints, making the scheduling process highly complex. Although existing optimization methods, such as heuristic search and traditional optimization techniques, have been applied, they often prove unsuitable in school environments where constraints are non-standard and change rapidly.

In this project, we aim to implement an AI-assisted classroom scheduling system capable of handling both hard and soft constraints to generate an optimal schedule using reinforcement learning techniques. The constraints, along with each teacher's workload, will be encoded into the AI model to evaluate and classify feasible schedules. Furthermore, the project will develop an AI model to help formulate a structured set of constraints, enabling the system to adapt flexibly to diverse scheduling requirements.