Beacon-based Monocular Vision Localization and Pose Estimation for AGV
Year: 2025
Type: article
Abstract: The accuracy of localization for Automated Guided Vehicles (AGVs) is a critical factor influencing their operational efficiency, safety, and the overall performance of automated systems. Visual localization for AGVs faces several challenges, including sensitivity to lighting variations, vulnerability to environmental changes, and computational demands that may impede real-time performance. Therefore, research into the visual localization of AGVs holds substantial practical significance, and employing monocular vision localization can effectively reduce localization costs. Differing from traditional localization methods, the proposed approach is a monocular vision localization technique that leverages visual beacons, providing significant technical advantages. At present, research on pure visual localization for AGVs remains relatively scarce. Employing monocular vision localization presents a viable approach to effectively reducing equipment costs. Studying this approach has theoretical significance and provides valuable guidance for enhancing its practical application. By strategically placing visual beacons on the AGV and performing camera calibration, a monocular camera is utilized to capture, recognize, locate, and estimate the posture of these beacons. Taking variables such as lighting conditions into consideration, a rigorous experimental design was formulated to validate the feasibility of the proposed methodology. Comprehensive experiments were conducted to evaluate the proposed localization approach, confirming that the AGV’s precision, pose estimation capability, and real-time performance meet the requirements of the specified application scenarios. The presented method exhibits significant potential for practical deployment. Compared to previous methods, this approach achieves the required localization accuracy while reducing deployment costs and enhancing real-time performance.
Authors: Jianjun Song, Chee-Loon Siow, Weiwen Zhao
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4411279728