Multi-Objective Path Planning for Environmental Monitoring using an Autonomous Surface Vehicle Federico Peralta fperalta@us.es Higher Technical School of Engineering, University of Seville Seville, Spain Michael Pearce scrambledpie@googlemail.com Zenith AI Belfast, United Kingdom Matthias Poloczek matpol@amazon.com Amazon San Francisco, USA Daniel Gutierrez Reina dgutierrezreina@us.es Higher Technical School of Engineering, University of Seville Seville, Spain Sergio Toral storal@us.es Higher Technical School of Engineering, University of Seville Seville, Spain Juergen Branke juergen.branke@wbs.ac.uk Warwick Business School Coventry, United Kingdom ABSTRACT Intelligent environmental monitoring is an exciting and impactful use case of autonomous vehicles. In this work, we explore measur- ing water quality parameters via a battery-powered Autonomous Surface Vehicle to obtain an accurate spatial model of the quality of a water resource while minimizing the distance traveled (alter- natively, time, or battery consumption). We develop an objective function for this class of problems and propose an evolutionary multi-objective optimization algorithm that selects the measure- ment locations while using a quick novel construction heuristic for determining the path. Our method leverages new crossover and mutation operators that seem to be of independent interest. We demonstrate the efcacy of the proposed approach in obtaining results for two diferent waterbodies: Lake Ypacarai in Paraguay, and Mar Menor in Spain. KEYWORDS water monitoring, path planning, spatial crossover, indirect encod- ing, multi-objective optimization ACM Reference Format: Federico Peralta, Michael Pearce, Matthias Poloczek, Daniel Gutierrez Reina, Sergio Toral, and Juergen Branke. 2022. Multi-Objective Path Planning for Environmental Monitoring using an Autonomous Surface Vehicle. In Genetic and Evolutionary Computation Conference Companion (GECCO ’22 Companion), July 9–13, 2022, Boston, MA, USA. ACM, New York, NY, USA, 5 pages. https://doi.org/10.1145/3520304.3528978 1 BACKGROUND The value of a water resource is based on the health and dynamics of the ecosystem it supports. Monitoring water quality parameters (WQPs) is important because the information they can provide helps in decision-making and strategies to maintain the health of Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). GECCO ’22 Companion, July 9–13, 2022, Boston, MA, USA © 2022 Copyright held by the owner/author(s). ACM ISBN 978-1-4503-9268-6/22/07. https://doi.org/10.1145/3520304.3528978 the water. However, there are examples where inadequate behavior and practices have led to polluted waters, e.g., Lake Ypacarai in Paraguay and Mar Menor in Spain. Normally, WQPs are monitored to understand the changing physical and chemical characteristics of water resources. In small lagoons, the normal procedure is to continuously monitor with sensors placed at specifc locations. However, this procedure does not scale to water resources with larger surface areas. For that reason, Autonomous Surface Vehicles (ASVs) have been utilized to accomplish the mission of monitoring [2, 5]. These ASVs, equipped with sensors to measure pH, temperature, Dissolved Oxygen, etc., travel on the surface of the water body and perform measurements. This paper proposes an intelligent water quality measurement planning system to perform a frst assessment of water quality in large-scale water environments using an ASV. The system considers one of the WQPs and searches for a preset number of measurement locations that allow to build a reliable surrogate model of the WQP over the entire surface. The measurement locations are chosen ac- cording to the following two goals: i) minimize the uncertainty of the resulting surrogate model and ii) minimize the distance traveled by the ASV. To achieve this, we frst formulate a quantitative objec- tive function that captures the desired behavior, and then propose an evolutionary algorithm for the given problem. Our contributions are as follows: • We formulate the task as a multi-objective problem for wa- ter resource monitoring, minimizing model uncertainty and path length. • We propose a multi-objective genetic algorithm to tackle this problem. The algorithm leverages novel genetic operators for efcient 2-dimensional space coverage. • The path planning part of the problem is a generalization of Path TSP, which we solve via an adapted Cheapest Insertion method. 2 PROBLEM DEFINITION The proposed system aims to efciently obtain a reliable surrogate model of the real water quality parameters of a particular water resource. An ASV equipped with water quality sensors will be used to measure water quality at a fxed number of locations on 747