Introducing Individuality into Students' High School Timetables

TL;DR

This paper extends the XHSTT high school timetabling format to support modular curricula by introducing two constraints: Student Choice and Balance Class Size. It formalizes an ILP extension of the Kristiansen et al. model, incorporating new variables, objective terms, and linking constraints, and demonstrates how to encode per-student course choices and balanced class sizes within the timetable. The authors evaluate 18 real-world German instances, translating Untis data into extended XHSTT, and report that the ILP approach yields feasible solutions for only 10 of the 18 cases within a six-hour window, highlighting the problem’s computational challenge. They propose leveraging the ILP as a component of metaheuristics (e.g., Large Neighborhood Search) and outline future work including SAT-based approaches and reinforcement-learning-inspired adaptive methods, with a public benchmark set to advance research in modular high school timetabling. The work provides a practical pathway toward personalized timetabling while acknowledging the gap between exact methods and real-world feasibility, and it lays groundwork for future solver enhancements and heuristic integrations.

Abstract

In a perfect world, each high school student could pursue their interests through a personalized timetable that supports their strengths, weaknesses, and curiosities. While recent research has shown that school systems are evolving to support those developments by strengthening modularity in their curricula, there is often a hurdle that prevents the complete success of such a system: the scheduling process is too complex. While there are many tools that assist with scheduling timetables in an effective way, they usually arrange students into groups and classes with similar interests instead of handling each student individually. In this paper, we propose an extension of the popular XHSTT framework that adds two new constraints to model the individual student choices as well as the requirements for group formation that arise from them. Those two constraints were identified through extensive interviews with school administrators and other school timetabling experts from six European countries. We propose a corresponding ILP formulation and show first optimization results for real-world instances from schools in Germany.