Predefined-Time Optimized Tracking Control for QUAVs With Extended State Observer and Global Prescribed Performance

This paper investigates the predefined-time optimized control problem for quadrotor uncrewed aerial vehicles (QUAVs) subject to the global prescribed performance constraints and partially unknown states. Under the backstepping framework, a predefined-time global prescribed performance optimized controller is developed through the actor-critic architecture, where the actor executes control actions and the critic evaluates their effectiveness. Furthermore, a neural network-based extended state observer is integrated into the optimized control approach to estimate unknown states and disturbances, enhancing the robustness of the control system, while the performance index is minimized. Notably, unlike most existing results, the requirement for the initial values of prescribed performance functions is removed, thereby reducing conservatism in the optimization of prescribed performance control. Additionally, a second-order command filter is introduced to overcome the “explosion of complexity”. It is demonstrated by the theoretical analysis that all closed-loop signals are bounded, while the position and attitude tracking errors subject to the global prescribed performance constraints converge into a neighborhood near zero within a predefined time. Finally, the feasibility and effectiveness of the proposed control strategy are validated through a simulation example. Note to Practitioners—Trajectory tracking serves as the foundation for QUAVs to accomplish complex flight missions. Among the key considerations in trajectory tracking are the requirements for tracking speed and precision. Moreover, unknown disturbances and partially unmeasurable states frequently occur. To overcome the aforementioned challenges, a predefined-time global prescribed performance optimized controller is developed, and the control cost is minimized. Compared to previous approaches, this paper eliminates the initial condition requirements of the prescribed performance approach. Furthermore, the extended observer is introduced to compensate for the effect of unknown disturbances and partial states. In addition, a second-order command filter is employed, avoiding the “explosion of complexity”. Through the theoretical analysis, all closed-loop signals are proved to remain bounded, and the tracking errors satisfy global prescribed performance, converging near zero within a preset time. A simulation confirms the validity of the proposed control strategy.