Metaheuristics for solving optimization industrial problems

• Autores: Enol García González
• Directores de la Tesis: José Ramón Villar Flecha (dir. tes.) , Javier Sedano Franco (codir. tes.)
• Lectura: En la Universidad de Oviedo ( España ) en 2024
• Idioma: inglés
• Títulos paralelos:
  • Metaheurísticas para la resolución de problemas de optimización en la industria
• Tribunal Calificador de la Tesis: Elías Fernández Combarro Álvarez (presid.) , Noelia Rico Pachón (secret.) , Anca Andreica (voc.) , Petrica Claudiu Pop Sitar (voc.) , Lidia Sánchez González (voc.)
• Enlaces
  • Tesis en acceso abierto en: RUO
• Resumen

  • The Multirobot Path Planning (MPP) problem, a highly significant optimization problem, is extensively cited in the literature. Its applications span across various fields, including the industrial sector, space exploration, intelligent laboratories, and education.

    The MPP is based on the approach of a situation in which multiple robots move simultaneously inside a room. These robots must reach their destinations in the shortest possible time and ensure that no collisions occur, avoiding both static obstacles in the room and the other robots moving in the environment.

    While the MPP literature predominantly focuses on heuristic techniques, particularly variants of A*, recent research has started exploring the use of metaheuristics. The potential of metaheuristics for this problem is evident, but a comprehensive evaluation of these techniques is still required to understand their full range of advantages and disadvantages.

    The main content of this thesis consists of applying different metaheuristic techniques in solving the MPP problem, focusing mainly on its application in an industrial environment. The work carried out in the thesis analyzes and evaluates the strengths and weaknesses of different proposals in different environments and with different numbers of robots to determine in which cases it is better to use each technique in order to minimize the time that the robots take to complete the route and the time that the planner takes to calculate the optimal route.

    Finally, this thesis presents a new metaheuristic inspired by the foraging behavior of the Slime Mould. This new metaheuristic technique is compared with other metaheuristics well-known in the literature to solve a path planning problem.

    The main conclusions of this thesis are that, after extensive research in the field of MPP, new methods have been proposed that significantly improve the quality of the solutions proposed in the literature. In particular, genetic and co-evolutionary algorithms were used in the generation of collision-free paths in multi-robot problems. Moreover, the proposed solutions are efficient not only at the level of results obtained, but also at the level of computational times. In these new contributions, the real-world needs for their application are better identified, describing in a simple way the adaptation of each problem to be solved with the proposal made in this thesis. The analysis of the literature on metaheuristic methods and the proposal of new procedures for the evaluation and comparison of metaheuristics, as well as a new metaheuristic for optimization based on Slime Mold and its hybridization with other well-known optimization techniques, also stand out as relevant conclusions.