Offline modeling-online correction integrated fault-tolerant control for unmanned aerial vehicle landing

To address the problem of degraded attitude control accuracy caused by control surface damage faults during unmanned aerial vehicle landing, this paper proposes a fault-tolerant control method for unmanned aerial vehicle landing integrating offline network modeling and online correction. Firstly, the longitudinal dynamics model of the unmanned aerial vehicle is established to provide a foundation for subsequent analysis and design. Secondly, an identification strategy of “offline network-online correction” is designed. In the offline phase, the powerful nonlinear fitting capability of deep neural networks is utilized to extract aerodynamic patterns from historical flight data for training a high-precision aerodynamic baseline model. In the online phase, a recursive least squares incremental identification method is adopted to achieve efficient incremental updates of aerodynamic parameters using only real-time observed data, balancing modeling accuracy and real-time performance. Finally, a super-twisting sliding mode fault-tolerant control method is proposed, where the real-time aerodynamic parameters obtained through online identification are dynamically embedded into the controller structure, and adaptive adjustment of the attitude control law is realized by correcting the model parameters. Mathematical simulation comparison results show that, under the faults of control surface damage, the proposed adaptive fault-tolerant control method significantly reduces the overshoot and steady-state error of key attitude parameters compared with classical control methods. Moreover, it exhibits superior robustness under external disturbances, effectively improving the landing control performance.