Neural Network-Based Adaptive Swing Suppression Control for Tower Cranes With Obstacle Avoidance

Obstacle avoidance presents a great challenge for the control system of tower cranes due to their uncertain dynamics, underactuated nature, and strong coupling. Aiming at these issues, a novel neural network-based adaptive controller is proposed for tower cranes with obstacle avoidance. The proposed controller addresses several critical and practical application-oriented control issues simultaneously, including obstacle avoidance, uncertain disturbances, state constraints, and nonlinear dynamics. Technically, the obstacle avoidance task is transformed into a state-constrained control problem by utilizing the coupling characteristic of tower cranes. Then, several state constraint auxiliary terms are designed to limit the payload and trolley position within a collision-free range, achieving obstacle avoidance for the payload. In addition, the neural networks are employed to approximate uncertain/unknown dynamics of the crane systems. Importantly, the proposed method achieves precise positioning, fast swing suppression, and obstacle avoidance simultaneously, with a short transportation time and strong robustness against unknown disturbances. The convergence and stability of the proposed control system are theoretically proven by using the Lyapunov stability theory. Finally, the effectiveness of the proposed controller is verified with hardware experiments. The experimental results indicate that compared with the existing method, the proposed method could reduce the transportation time by 70% and reduce the maximum swing angle by 15% in the presence of the same external disturbances and parameter uncertainties.