Optimizing Academic Timetables with GRASP and Simulated Annealing: A Case Study at a Federal Institute

This paper presents a hybrid solver for the problem of academic timetabling in a Federal Institute. The approach combines a GRASP-inspired constructive heuristic with Simulated Annealing, neighborhood operators, and automatic hyperparameter calibration. The problem was modeled by an objective function based on penalties for strong and weak constraints, contemplating teacher and class conflicts, unavailability, pedagogical limits, and grid quality criteria. Three instances derived from real data were evaluated, with 127, 255 and 382 weekly blocks of class, in 30 computational repetitions with distinct pseudorandom seeds. The results indicated the generation of viable solutions, without violations of strong constraints, with low variation in final costs and times compatible with institutional use. The commercial software used as an operational reference did not produce, in the tested configurations, complete grids without substantial manual intervention, reinforcing the need for customized solutions.