In response to the problem of insufficient navigation stability of transmission and transformation line transportation robots in complex environments such as lighting changes, obstacle occlusion, and terrain undulations, this study proposes a navigation method that integrates computer vision and multi-source information processing. Firstly, a multi-source perception module was developed, which integrates a global shutter industrial camera, a 16 line mechanical LiDAR, and an ADIS16470 inertial measurement unit. Based on the above hardware, further design an improved federated extended Kalman filter fusion architecture to achieve global pose estimation and dynamic obstacle recognition. In the path planning and real-time obstacle avoidance section, this study combines the optimized A algorithm with the YOLOv5 dynamic detection model to form a collaborative strategy. Experiments have shown that under complex operating conditions, compared with traditional single GPS/INS navigation schemes, its positioning error significantly decreases from ± 1.2 meters; Compared to the loose coupling fusion method of vision and LiDAR with an error of ± 0.6 meters, the multi-source fusion scheme proposed in this study improves the positioning accuracy to ± 0.2 meters. At the same time, the system path tracking error remains within 0.15 meters, the obstacle avoidance response efficiency is improved by 40%, and the overall reliability of navigation is significantly enhanced.
Yuan et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: