| Συγγραφείς: |
Cui, Zhanxin1 (AUTHOR) cuizhanxin19@nudt.edu.cn, Zou, Wen2 (AUTHOR) 8320763@qq.com, Feng, Zhiwei3 (AUTHOR) fengzhw@nudt.edu.cn, Zhang, Qingbin4 (AUTHOR) qingbinzhang@sina.com, Yang, Tao5 (AUTHOR) yt_yangtao@nudt.edu.cn |
| Περίληψη: |
Existing predictive control of parafoils is based on the local linearization of the dynamic system around its equilibrium state, which fails to capture global nonlinear characteristics. To address this limitation, we propose a novel data-driven framework for parafoil control based on the global linearization property of the Koopman operator. The dynamic model of the parafoil system is derived by applying the principles of momentum and angular momentum to comprehensively depict its intricate dynamics. The Koopman operator is used to approximate the nonlinear dynamics of the parafoil and enables a suitable linear representation for model predictive control (MPC) design in a high-dimensional state space. Based on the Koopman operator, a model predictive controller for the parafoil system has been successfully designed and implemented. Experimental data have verified the accuracy of the Koopman model in predicting the yaw angle and position of the parafoil system. The simulation results demonstrate that under the T-shaped trajectory, the root mean square errors (RMSEs) in the eastward, northward, vertical, and yaw directions were reduced by 34.27%, 14.41%, 8.70%, and 31.03%, respectively, compared to those of the traditional MPC. The tracking performance of this controller has been effectively validated by precisely tracking the J/T-shaped homing trajectory, where the Koopman MPC demonstrates comprehensive RMSE reductions ranging from 9.59% to 72.72% for the J-shaped path, except slightly higher eastward mean absolute error. Additionally, in the tracking control, the trained Koopman model improves efficiency by nearly 5.3 times compared with the local linearization method. The results demonstrate that the proposed Koopman-based MPC architecture controller is an effective approach for the trajectory tracking of the parafoil system, whose global linearization characteristic effectively suppresses tracking lag caused by frequent local linearization. This study verifies the model's prediction capability in airdrop tests and the controller's robustness under complex trajectories, providing new insights for parafoil system modeling and control. Practical Applications: This study offers a novel control framework for parafoil systems, which are widely used in precision airdrop missions, drone delivery, and aerial recovery operations. Traditional control methods often struggle to handle the complex nonlinear dynamics of parafoils, limiting their accuracy in real-world scenarios. This approach leverages advanced data-driven techniques to create a globally accurate predictive control model, enabling precise trajectory tracking even under challenging conditions. For practitioners, this innovation means improved reliability in critical tasks such as delivering supplies to remote areas, capsule recovery, or recovering unmanned aerial vehicles. The method's ability to predict and adjust the parafoil's yaw angle and position with high accuracy ensures safer landings, reduced risk of cargo damage, and enhanced operational efficiency. Experimental validation using J/T-shaped homing trajectories—common in real-world navigation—demonstrates its practical effectiveness. By translating complex dynamics into a user-friendly control strategy, the results directly address longstanding challenges in autonomous parafoil systems, offering scalable solutions for industries reliant on precision airdrops. [ABSTRACT FROM AUTHOR] |