Advanced path planning for safe human-robot collaboration in construction: a pose-aware approach

The construction industry has faced stagnant productivity and a high fatality rate among workers. To address these issues, robotization has been implemented, particularly through physical human–robot collaboration (pHRC), where robots perform physically demanding tasks in collaboration with workers in shared workspaces. However, inefficient motion information exchange between robots and workers can lead to collisions and safety risks. To address this concern, this study proposes a pose-aware human–robot collaboration method. This approach develops a 3D pose estimation deep network to analyze workers’ postures. Based on these motion data, the method refines the Rapidly Exploring Random Tree path-planning algorithm, enabling robots to generate collision-free paths. A material-delivering task was employed to evaluate the method. The results reported that the proposed method allows robots to make safe adjustments based on workers’ motion with an 86.5% success rate. The findings support the development of safe and reliable pHRC solutions, enhancing robotics implementation in construction.