Multi-neighborhood simulated annealing for the sports timetabling competition ITC2021

Abstract

We describe the solver that we developed for the Sports Timetabling Competition ITC2021, a three-stage simulated annealing approach, that makes use of a portfolio of six different neighborhoods. Five of these neighborhoods are taken from the literature on round-robin tournament scheduling, whereas the last one, denoted as PartialSwapTeamsPhased, is a novel contribution and it is specifically designed for the phased version of the problem. We perform a comprehensive and statistically principled tuning procedure to find the best combination of parameters for the competition instances. We dedicate specific focus to evaluate the contribution given by the new neighborhood PartialSwapTeamsPhased, which yielded better results on most phased instances. Overall, the final outcome is that the three-stage simulated annealing solver is able to find a feasible solution on 44 out of 45 instances and ranked second in both the first competition milestone and the final round. We also propose an Integer Linear Programming model implemented in CPLEX, which, unfortunately, did not produce significant results on the instances of the competition.

The international timetabling competition on sports timetabling (ITC2021)

“…Team UoS (winners) has used an IP-based matheuristic, in which they iteratively x many variables in their IP model, in order to reduce the computation time (Lamas-Fernandez et al, 2021). Team Udine (second) has used a simulated annealing strategy, using 5 neighborhoods from the literature, as well as one neighborhood they developed specically for the competition instances (Rosati et al, 2021). Team Saturn (third) has used a more classic though novel IP decomposition approach (Sumin & Rodin, 2021).…”

Effective metaheuristics for the latency location routing problem

“…Note that the presented SA scheme is also known in the literature as multineighborhood SA (Bellio et al., 2021; Rosati et al., 2022).…”