Multi-Robot Informative Path Planning

Francisco da Silva Parreira de Castro Fonseca

Key information

Authors:

Francisco da Silva Parreira de Castro Fonseca (Francisco da Silva Parreira de Castro Fonseca)

Supervisors:

Alberto Manuel Martinho Vale (Alberto Manuel Martinho Vale); Rodrigo Martins de Matos Ventura (Rodrigo Martins de Matos Ventura)

Published in

November 27, 2024

Abstract

The use of information in path planning has made evident the need to make monitoring and exploring missions more efficient using Multi-Robot (MR) teams. This is especially relevant in Chemical, Biological, Radiological, and Nuclear (CBRN) threats. Such teams can reduce the human risk associated with hazardous scenarios and lead to more efficient missions. This work proposes a MR Informative Path Planning (IPP) modular framework to efficiently map a scalar quantity applied to scenarios of radiological monitoring. It is built upon IPP theory, Gaussian Processes (GP) and Bayesian Estimation (BE), Exploration versus Exploitation strategies, and path generation using Rapidly Exploring Information Gathering (RIG)-trees. Two methods are explored in detail: a BE approach, which uses the current set of observations made by the agents to estimate the location and intensity of the sources in the workspace, and a GP approach, which uses mutual information to choose the future locations to sample. The proposed framework is developed modularly and is a two-stage solution combining the BE as a first exploration phase and the GP approach to exploit high-interest points. Both stages use RIG-trees to generate candidate paths, using a control variable to separate the two stages. A Python simulator is developed to run experiments with the planner for different stage separations against a Boustrophedon cell decomposition baseline. Each experiment uses a set of simulated scenarios for different mission configurations. The final result offers a MR IPP framework, capable of being used for other variables of interest within the emerging concern of CBRN threats.