AquaFeL-PSO: A Monitoring System for Water Resources using Autonomous Surface Vehicles based on Multimodal PSO and Federated Learning

Abstract

The preservation, monitoring, and control of water resources has been a major challenge in recent decades. Water resources must be constantly monitored to know the contamination levels of water. To meet this objective, this paper proposes a water monitoring system using autonomous surface vehicles, equipped with water quality sensors, based on a multimodal particle swarm optimization, and the federated learning technique, with Gaussian process as a surrogate model, the AquaFeL-PSO algorithm. The proposed monitoring system has two phases, the exploration phase and the exploitation phase. In the exploration phase, the vehicles examine the surface of the water resource, and with the data acquired by the water quality sensors, a first water quality model is estimated in the central server. In the exploitation phase, the area is divided into action zones using the model estimated in the exploration phase for a better exploitation of the contamination zones. To obtain the final water quality model of the water resource, the models obtained in both phases are combined. The results demonstrate the efficiency of the proposed path planner in obtaining water quality models of the pollution zones, with a 14$\%$ improvement over the other path planners compared, and the entire water resource, obtaining a 400$\%$ better model, as well as in detecting pollution peaks, the improvement in this case study is 4,000$\%$. It was also proven that the results obtained by applying the federated learning technique are very similar to the results of a centralized system.