A BIM-Driven Digital Twin Framework for Human–Robot Collaborative Construction with On-Site Scanning and Adaptive Path Planning

As the Architecture 4.0 paradigm advances, integrating robotic systems into construction workflows has become vital to address labor shortages and enhance execution precision. However, conventional BIM-based automation struggles to cope with dynamic, cluttered on-site environments. This study presents a closed-loop digital twin framework that fuses 3D BIM modeling, real-time sensor-based site scanning, and human-robot interaction to enable adaptive collaboration in architectural construction. The system continuously updates its digital twin using LiDAR and RGB-D DATA to capture spatial deviations, unexpected obstacles, and environmental changes. Based on these inputs, the robot’s motion trajectories are recalculated through an online path replanning module. We validate the system using a wall panel dry-hanging case, involving a 6 degrees of freedom (6 DoF) ABB robotic arm and a Unity-VR interface for immersive human supervision. Across 24 experimental runs in cluttered environments, the adaptive system achieved a $92.4 \%$ average placement accuracy, reducing positioning error by $47 \%$ compared to static BIM-based workflows. Obstacle avoidance success rate reached $\mathbf{9 5. 8 \%}$, and average task completion time decreased by $18.6 \%$ due to reduced manual intervention. These results demonstrate the framework’s potential to transition construction robotics from pre-scripted automation to intelligent, real-time collaboration.