A-LIQ: A Warehouse AGV Navigation AIgorithm Based On Laser- InertiaI QR Code Fusion

• Published in: 淡江理工學刊 Vol. 27; no. 9; pp. 3233 - 3242
• Main Authors: Xin Wang, Bo Zhang, Jianwei Zhao, Hezhen You
• Format: Journal Article
• Language: English
• Published: 淡江大學 01.01.2024
• Subjects: Automatic Guided Vehicle (AGV) Navigation; Laser Odometry; QR Code; IMU Preintegration; Factor Graph Optimization
• ISSN: 2708-9967
• DOI: 10.6180/jase.202409_27(9).0012

The method based on laser inertial navigation technology has been widely used in the navigation field of automatic guided vehicles (AGV) in warehouse workshops. The existing algorithms are prone to scale drift, large cumulative error, and LIDAR degradation leading to serious reduction in the number of sensing points, we aim to design an AGV navigation framework based on the fusion of laser, inertial, and quick response (QR) code technologies (named A-LIQ). First, an inertial measurement unit (IMU) pre-integration model with QR code is proposed, and the obtained QR code constraint information is added between two key frames to form a new composite unit, reducing scale drift, and improving positioning accuracy. Secondly, a local map optimization model is proposed, keyframes and QR codes are selectively introduced, local stratified bundle adjustment (BA) optimization is performed based on sliding windows, and keyframe poses and map point locations are updated. Finally, a LiDAR/IMU/QR code tight coupling optimization method is proposed, and the pre-integration factor, closed-loop factor, QR factor, and laser odometer factor are incorporated into the factor graph system to achieve multi-level data fusion. In this paper, the method is verified on the developed AGV navigation platform, and its performance is evaluated by using measured data and compared with LeGO-LOAM, BALM, LIO-SAM. The results show that the method does not significantly increase the calculation amount. Effectively improve the track closure effect at the closed loop, with lower positioning error, positioning accuracy error is less than 0.02 meters, attitude error is less than 2◦.