In this paper, an output feedback control approach based on online optimization for a class of constrained-input systems with uncertainties is presented. For this propose, the system is augmented with the aid of a Parallel Feedforward Compensator (PFC), which is designed based on an inaccurate linear model. The proposed method is performed by placing the transmission zeros of the augmented linear system at the left-hand side of the complex plane, creating a systematic and LMI-based procedure. It is worth noting that the method improves many commonly adopted restrictive assumptions on the linear model and is applicable for a class of nonlinear systems as well. In the augmented system, the augmented output is considered to form the external states, while the other linearly independent variables constitute the internal dynamics. The control signal is derived by finding online solution of a Quadratic Programming (QP) problem. This problem regulates the augmented output over a constrained input space, producing an optimal control law. Online solution of the QP is computed using a Recurrent Neural Network (RNN). Guaranteed stability and rapid convergence to the optimal solution are among advantages of such networks, which make the proposed control method more reliable. Furthermore, boundedness of the internal states during regulation of the external states is ensured in the presence of a class of uncertainties and nonlinearities. Effectiveness of the proposed method illustrated using a simulating example.

本文提出了一种基于在线优化的一类具有不确定性的约束输入系统的输出反馈控制方法。对于此建议，借助基于不精确线性模型设计的并联前馈补偿器（PFC）来增强系统。通过将增强线性系统的传输零点放在复杂平面的左侧，创建系统的，基于LMI的过程来执行所提出的方法。值得注意的是，该方法改进了线性模型上许多常用的限制性假设，并且也适用于一类非线性系统。在增强系统中，增强输出被视为形成外部状态，而其他线性独立变量构成内部动力学。通过找到在线二次编程（QP）问题的解来得出控制信号。这个问题在受约束的输入空间上调节增加的输出，从而产生最佳控制律。使用递归神经网络（RNN）计算QP的在线解决方案。这种网络的优点包括保证稳定性和快速收敛到最优解，这使得所提出的控制方法更加可靠。此外，在存在一类不确定性和非线性的情况下，确保了在外部状态调节期间内部状态的有界性。仿真实例说明了所提出方法的有效性。产生最优控制律。使用递归神经网络（RNN）计算QP的在线解决方案。这种网络的优点包括保证稳定性和快速收敛到最优解，这使得所提出的控制方法更加可靠。此外，在存在一类不确定性和非线性的情况下，确保了在外部状态调节期间内部状态的有界性。仿真实例说明了所提出方法的有效性。产生最优控制律。使用递归神经网络（RNN）计算QP的在线解决方案。这种网络的优点包括保证稳定性和快速收敛到最优解，这使得所提出的控制方法更加可靠。此外，在存在一类不确定性和非线性的情况下，确保了在外部状态调节期间内部状态的有界性。仿真实例说明了所提出方法的有效性。在存在一类不确定性和非线性的情况下，可以确保在调节外部状态时内部状态的有界性。仿真实例说明了所提出方法的有效性。在存在一类不确定性和非线性的情况下，可以确保在调节外部状态时内部状态的有界性。仿真实例说明了所提出方法的有效性。