Design and Intelligent Control of a Special-Shaped Crane for Confined Spaces

Hoisting operations in confined spaces demand exceptional precision and safety. This paper proposes a hybrid intelligent control strategy for a special-shaped crane, characterized by its multi-articulated structure designed for high dexterity, to meet these challenges. Based on a multi-rigid-body dynamic model, our strategy integrates Deep Reinforcement Learning (DRL) for high-level path planning and Robust Sliding Mode Control (RSMC) for low-level motion control. The DRL agent generates optimal, collision-free paths in complex environments filled with dense obstacles, while the RSMC ensures precise trajectory tracking and actively suppresses payload vibrations. Simulations demonstrate that this hybrid method reduces end-effector vibration by over 75% and achieves superior positioning performance. This research provides a robust mechatronic solution for critical applications such as nuclear facilities and aerospace assembly.