A reinforcement learning and nonlinear adaptive backstepping hierarchical control for autonomous remote construction using 5G network

Automated construction can effectively improve construction efficiency, reduce safety risks and labor costs. However, it requires high level precision and advanced artificial intelligent algorithms. This often requires extensive computational devices on-site, which may affect construction tasks. To address these issues, an automated remote construction method using 5G network is proposed to reduce on-site equipment deployment. A reinforcement learning algorithm is implemented for lifting path planning. Robot kinematics and a novel nonlinear backstepping hierarchical control framework are developed to ensure that cranes can precisely execute the lifting path. High-gain observers are designed to provide precise estimation of control feedback. This control framework considers nonlinear effects and the changes in moment of inertia caused by lifting heavy construction materials. The proposed remote construction method is validated through automated retaining wall construction experiments, showing that the robotized crane can be remotely controlled to accurately complete the tasks.