针对一类工作点时变的光滑非线性多变量被控对象，采用离线辨识的RBF-ARX模型描述系统的全局非线性动态特性，并在此基础上提出一种具有自适应微分作用的非线性模型预测控制方法。该方法将整个预测时域内模型输出和期望输出的偏差变化率考虑到优化目标中，利用与偏差变化率有关的函数动态修改该优化项的权重，从而能在保证控制器稳态性能的情况下有效地改善系统的动态性能。最后利用该方法对四旋翼飞行器实验装置进行了实际的姿态控制，控制效果验证该方法的有效性。%For a class of smooth nonlinear multivariable systems whose working-points vary with time, a Gaussian radial basis function (RBF) neural networks-based local linearization autoregressive with exogenous (ARX) model was built to describe the system’s global behavior, and an improved nonlinear model predictive control (NMPC) method with adaptive differential effect based on RBF-ARX model identified offline was presented. Difference from conventional NMPC, the differential of errors between model outputs and designed outputs in whole prediction horizon were considered and their weights were adapted by functions of themselves in each optimization process, thus the controller can improve dynamic performance when the steady-state performance is ensured. A case study on a quadrotor for its real attitude control indicates that the proposed method is effective.
展开▼
