Autonomous Navigation of Construction Robots Based on Visual SLAM Technology

Xinjun Wang

Abstract: AbstractIn the field of automated construction, the autonomous navigation system of construction robots plays an important role in improving work efficiency and safety. Visual SLAM technology has demonstrated its superiority in the dynamic scene of construction due to its strong adaptability and relatively low cost. Therefore, based on the characteristics of the construction environment, the traditional visual SLAM technology has been improved to enhance the positioning and map construction performance of robots. By combining RGB-D cameras with the ORB-SLAM2 framework, the accuracy of robot environmental perception and positioning has been improved, and the YOLOv4 tiny network has been used to eliminate dynamic interference and optimize the stability of the navigation system. The simulation results show that the autonomous navigation system for construction robots based on improved technology reduces path planning and travel time by 19.08% and 13.81%, respectively, and significantly improves the efficiency and accuracy of 3D map construction. This research achievement provides a new technological approach for precise navigation of construction robots, which is of great significance for achieving autonomous robot operations in construction scenarios.