Possibilistic Robot Localization Using Visual Landmarks

ABSTRACT This contribution introduces a novel camera‐based localization method for omnidirectional mobile robots using fisheye lenses to capture images of landmarks to estimate robot position and orientation. In this work, the landmarks are red poles that are indistinguishable from one another. Ambiguity is purposefully allowed for ease of installation. In order to face challenges like landmark detection ambiguity and noisy camera feeds, the proposed localization method employs possibility theory for robust state estimation, embracing the inherent uncertainties. This approach diverges from traditional precise localizations by considering a robust range of possible robot positions, presenting the opportunity of enhancing safety in applications such as collision avoidance. Early integration into a possibilistic filter algorithm shows promise for improving dynamic state estimation, marking an advancement in applying possibility theory to real‐world scenarios.