Automated Part Placement for Precast Concrete Component Manufacturing: An Intelligent Robotic System Using Target Detection and Path Planning

Abstract

Placing embedded parts (EPs), e.g., junction boxes or plastic cable ducts, in a precast concrete (PC) component is a fundamental and repetitive trade in its manufacturing. Yet, such trade is far from being automated to enhance PC component manufacturing productivity. This study presents an intelligent robotic system for automated part placement for PC component manufacturing by using target detection and path planning. The proposed system consists of an Aubo-i5 robotic arm, a Robotiq 2F-85 clamping claw, and an Intel Realsense D435i depth camera. An improved YOLOv5 target detection algorithm is proposed to automatically detect EPs with high precision, and a two-way two-threaded informed RRT∗ path planning algorithm is developed to optimize the robot movement. Using junction box placement as an experiment, performance of the system was evaluated by examining EP detection, clamping, path planning, and placement. The visual detection model achieved a mAP value of 99.5%. The efficiency of the path planning algorithm was improved by 37.7% compared with Bidirectional RRT∗ with close pathfinding quality. The final success rate of EP placement reached 99.8%. The research contributes to the field of PC component production by providing an automated system for EPs placement.