Análise comparativa de algoritmos de vigilância persistente multi agente para a limpeza de manchas de óleo na costa alagoana

Abstract

This paper presents a comparative analysis of persistent surveillance algorithms applied to the simulation of oil slick cleanup off the coast of Alagoas, with the aim of evaluating the performance of different multi-agent control strategies in dynamic environments. The methodology adopted integrates MATLAB® and the PyGNOME library to build five scenarios, in which environmental factors such as wind and tide are considered in the modeling of oil dispersion through Lagrangian simulation. Three movement policies were implemented: random, reactive and adapted, the latter two using the A* algorithm for robot decision-making, with support for density maps generated by kernel estimators (KDE). The statistical analysis of the data, based on the Shapiro-Wilk, Friedman and Wilcoxon tests, allowed an evaluation of the efficiency of each approach. The results obtained analyze how each policy behaved according to the scenario evaluated and their performance was compared. The research contributes to the development of clean-up strategies in the context of response to coastal environmental disasters.