Autonomous Block Assembly for Boom Cranes with Passive Joint Dynamics: Integrated Vision MPC Control

A research team from TU Wien (Gerald Ebmer, Minh Nhat Vu, Tobias Glück, Wolfgang Kemmetmüller) has published a paper detailing a novel autonomous framework for construction cranes. The system tackles the critical challenge of precise block assembly, where the pendulum-like sway from a crane's passive joints makes accurate placement difficult. Their integrated approach combines computer vision for real-time block tracking, intelligent path planning to avoid collisions, and a sophisticated Nonlinear Model Predictive Control (NMPC) system to actively suppress motion. This allows a boom crane to autonomously perform the complete sequence of locating, picking up, and precisely placing prefabricated building blocks.

The technical core uses a collision-aware planner to generate smooth B-spline reference paths in real-time on standard CPU hardware. The NMPC controller then tracks this path while continuously using vision feedback to counteract sway. Experimental validation on a laboratory testbed demonstrated successful autonomous stacking and obstacle-avoidance. The key result is a dramatic improvement in efficiency: the AI-driven sway damping reduces settling times by more than an order of magnitude (over 10x) compared to relying on the crane's natural, uncontrolled passive dynamics. This proves the real-time feasibility of a fully integrated, vision-guided control system for automating complex construction tasks.