In context
In 2024, automated guided vehicles (AGVs) were becoming central to smart manufacturing, yet their navigation in indoor workshops still suffered from drift, scale ambiguity, and reliance on GNSS that fails indoors. This paper addressed those gaps by fusing forward-view visual-inertial data with downward-view QR code landmarks for globally consistent, precise localization.
What was reported
Researchers from Shenyang Jianzhu University and partner institutions proposed a real-time, tightly-coupled multi-view visual-inertial navigation method for workshop AGVs. The system integrates a forward-facing camera and IMU with a downward-facing camera that detects floor-mounted QR codes, providing absolute pose references in a global coordinate frame.
Key technical contributions include a joint initialization scheme using a maximum a posteriori probability model to improve scale estimation, a QR-code-based pose correction model that periodically compensates keyframes to reduce cumulative error, and a pose-constrained optimization model to avoid local minima in bundle adjustment.
Experiments on a workshop AGV platform showed that the method outperformed state-of-the-art visual-inertial navigation approaches in both time efficiency and accuracy, achieving a translation root mean square error (RMSE) of less than 50 mm and a rotation RMSE of less than 2 degrees.
Why it mattered
This work demonstrated a practical path to high-precision AGV localization without external infrastructure like GNSS, combining the richness of vision with the robustness of inertial sensing and the absolute referencing of QR codes. It offered a scalable solution for indoor factory automation where repeatable, centimeter-level positioning is critical for material handling and assembly tasks.
“The results demonstrate the superiority of the proposed method in terms of time efficiency and positioning accuracy.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2024-07-20 · “多视角视觉－惯性融合的车间AGV精确导航方法”