A carregar...
Publicação
Metaheuristic strategies for solving scheduling problems for gymnastics competitions
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 942.89 KB | Adobe PDF |
Orientador(es)
Resumo(s)
The positioning of teams regarding acrobatic gymnastics can become a challenge considering the complexity of the sport and the restrictions needed for a fair and adequate championship. Competition schedules are organized by blocks that symbolize the interval in the calendar where a predefined set of teams
with the same characteristics will act. The starting order defines the positioning of teams in each block.
This thesis portrays the research, development, and application of an algorithm that generates an admissible solution for assigning teams in the starting order, explore the hypothesis of generating solutions
with hard and soft constraints to provide an admissible solution.
Initially, the problem was formalized through theoretical research. Posteriorly, interviews were carried out with judges, athletes, and championship organizers to determine the associated hard and soft
constraints. These will define whether or not a starting order is admissible.
Due to the type of problem and the complexity of the associated constraints, different types of metaheuristics were tested, specifically local search and evolutionary algorithms. The following methods
were selected for implementation: Hill-Climbing (the baseline), Simulated Annealing, and Genetic Algorithm. These models underwent a testing phase in different types of competitions, and computational
complexity was also analyzed. Solutions were explored and compared between strategies. Results show
that local search methods can modulate different types of schedules in a reasonable amount of time, with
the simulated annealing technique providing the best results.
Descrição
Tese de mestrado, Ciência de Dados, 2023, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Otimização Hill Climbing Simulated Annealing Algoritmo Genético Metaheurísticas Teses de mestrado - 2024