Teaching cranes to move smoothly without building detailed models

Automating crane control requires smooth trajectories that suppress load oscillations while meeting safety and efficiency targets. This work bypasses traditional model-based control and large-scale learning by using behavioral data directly—applying Willems' fundamental lemma to identify system dynamics from input-output measurements alone. The approach uses convex optimization to generate smooth, energy-efficient trajectories with minimal expert tuning. Laboratory tests on a rotary crane show 35% reduction in load sway, 43% lower tracking error, and 50% faster operation compared to established model-based methods. The method handles underactuated crane dynamics and requires limited data, making it practical for deployment.