Intelligent Online Learning Strategy for an Autonomous Surface Vehicle in Lake Environments Using Evolutionary Computation

Abstract

An artificial intelligence strategy for the localization and monitoring of pollution-caused events caused by contamination in a lake, of a system consisting of an Autonomous Surface Vehicle (ASV) and a network of wireless beacons is proposed and evaluated. Particularly the event of algae bloom is considered. For this purpose, the path planning of the ASV is calculated at different phases with the help of an evolutionary algorithm. The main novelty of the proposed strategy is that it follows an intelligent online learning approach. Therefore, the ASV learns from the environment and makes decisions depending on the collected data. The proposed path planning is based on the adaption of the travelling salesman problem with constraints, using the beacons as geo-localization references and information support infrastructure. The approach consists of different phases that balance the exploration of the lake for searching new events and the exploitation of already discovered ones. A suitable configuration of the fitness function allows an efficient balance between exploration and intensification. Simulation results show that the level of coverage achieved are at least 85% for a situation where up to two dynamic algae blooms occurred at different locations in the lake.