Adaptive Localization Configuration for Autonomous Scouting Robot in a Harsh Environment

Abstract

Greenpatrol project aims to develop a robotic prototype for pest detection and treatment in greenhouse crops. The robot platform uses a sensor fusion approach with Global Navigation Satellite System (GNSS), odometers, inertial and range sensors in order to obtain a position and heading solution with the required precision to navigate inside the greenhouse and to localize accurately the pests. Previously some tests were carried out inside a greenhouse to verify its localization subsystem performance, but due to the difficulties to get a reliable ground truth, additional tests in open sky conditions has been done. As there are some circumstances than can degrade the GNSS signals, especially in a harsh environment like a greenhouse, an adaptive configuration of the Augmented Monte Carlo Localization (AMCL) based in GNSS quality indicators is proposed. The open sky data collected has been used to check the behavior of the proposed approach simulating gaps in the GNSS signals and the results show that this localization subsystem can deal with these outages maintaining the position solution close to the system specifications.