Informative path planning based on swarm intelligence for a fleet of autonomous surface vehicles for water resource monitoring

• Autores: Micaela Carolina Jara Ten Kathen
• Directores de la Tesis: Isabel Jurado Flores (dir. tes.), Daniel Gutiérrez Reina (codir. tes.), Diego Luis Orihuela Espina (tut. tes.)
• Lectura: En la Universidad Loyola Andalucía ( España ) en 2023
• Idioma: inglés
• Número de páginas: 139
• Tribunal Calificador de la Tesis: Sergio Luis Toral Marín (presid.), Alfonso Carlos Martínez Estudillo (secret.), NICOLA BASILICO (voc.)
• Programa de doctorado: Programa de Doctorado en Ciencia de los Datos por la Universidad Loyola Andalucía
• Enlaces
  • Tesis en acceso abierto en: repositorio.uloyola.es
• Resumen

  • The preservation, monitoring, and management of extensive water resources have presented significant challenges in recent decades. To maintain the quality of these water resources, continuous monitoring of pollution levels is essential. This monitoring is achieved through the observation of water quality parameters. An efficient and effective approach to conducting this monitoring is through the utilization of an intelligent system employing autonomous surface vehicles equipped with sensors designed for measuring these parameters.

    The present work focuses on the development of a system designed to monitor water quality parameters in water resources using a fleet of autonomous surface vehicles. The proposed methodology focuses on an informative path planning based on particle swarm optimization and Gaussian Process techniques, combining the learning capability of the former with the predictions of the models from the latter. The main objectives are to generate reliable and accurate water quality parameter models and to identify potential pollution hotspots.

    The first approach informative path planning is called Aqua-PSO and serves as the foundational algorithm for subsequent proposed systems. The second informative path planning in question is the AquaFeL-PSO, characterized by a two-phase approach. In its initial phase, the Aqua-PSO is used with a primary focus on exploration. Subsequently, in the second phase, the fleet of vehicles is divided into sub-fleets, each of which is dedicated to the exploitation of a potential contamination area. In this last phase, the Aqua-PSO exploitation approach is complemented with the integration of the federated learning technique, providing the sub-fleets with autonomy to generate water quality parameters models.

    Finally, a multi-objective monitoring system designed for heterogeneous fleets, called AquaHet-PSO, is presented. Heterogeneous fleets encompass vehicles with different sensor capacities, with various types and quantities of sensors on board. The AquaHet-PSO is based on AquaFeL-PSO and has a mission structured in three distinct phases: exploration, assignment of vehicles to pollution zones based on their sensor configuration, and finally the exploitation phase.

    The evaluation of the proposed informative path plannings was conducted in a simulated Ypacarai lake environment, yielding satisfactory results in terms of mean square error and error metrics. These results indicate the ability to obtain reliable and accurate models of the water quality parameters, as well as the successful detection of contamination peaks in the lake. Furthermore, compared to other algorithms, the proposed informative path planners outperformed their counterparts, further emphasizing their effectiveness.