EXISTING SOLUTIONS FOR AUTOMATED UNIVERSITY TIMETABLING: A COMPARATIVE REVIEW OF ALGORITHMS AND METHODS

This article provides a comparative review of the existing solutions reported in the literature for automated university timetabling, faculty–course assignment, and the automatic extraction of scheduling constraints from institutional documents. Building on the three threads identified in a preceding review, the solutions surveyed here fall into exact methods — mixed-integer linear programming (MILP), constraint programming (CP), satisfiability/maximum-satisfiability (SAT/MaxSAT), and satisfiability modulo theories (SMT) — and heuristic or metaheuristic methods — genetic algorithms, local search, and their hybrids. The principal difference between these two families is a trade-off between an optimality guarantee and scalability: exact methods return a provably optimal solution or a proof of infeasibility but their running time grows sharply with instance size, while heuristic methods give up that guarantee in exchange for tractable running times on realistically sized instances. A head-to-head comparison on a real dataset from Chalmers University of Technology quantifies this trade-off directly: Gurobi (MILP) solved all five annual instances in under 100 seconds each, while Z3 (SMT) failed to solve any instance within the time limit. For document-driven constraint extraction, the review compares prompting, fine-tuning, schema optimization, and retrieval-augmented generation (RAG) approaches. The article also situates existing locally authored solutions from Uzbekistan within this same comparative framework. Consistent with the intended scope of this stage of the research, no new solution is proposed here; that is left to a subsequent article.

Hali tarjima qilinmagan