Studies on Control Design Method based on On-line Modeling and Optimization

我们研究了基于控制回路中的在线数据自适应地更新模型的方法：LPV系统建模的扩展和模型预测控制方案的扩展的组合。在这些讨论的基础上，我们制定了控制问题，并提出了解决方案和算法计算的控制。首先，以电液伺服系统为控制对象，研究了电液伺服系统中存在的非线性因素和不确定性，难以精确建模的问题。在电液系统各工作点附近建立线性局部模型，以负荷变化为调度参数建立LPV全局模型。对于LPV系统，我们推导了最坏情况下的最优控制策略，并在计算机仿真和真实的对象中证明了控制策略的有效性。其次，作为基于局部模型控制的一个典型难点，我们重点研究了状态、输入和输出中的信息延迟问题。特别是，我们研究了如何克服输入和输出的控制对象的延迟。提出了最坏情况下的最优控制方案，推导了构造控制方案的有限维算法，阐明了控制方案的预测-观测器结构，并给出了说明性的数值仿真结果。

We investigated methods of updating the model adaptively based on online data in the control loop: a combination of an extension of LPV systems modeling and an extension of model predictive control scheme. On the basis of these discussions, we formulated control problems, and proposed solutions and algorithms for calculating the controls. In particular, we paid attention to the role of the information delay and the obstacles raised by the information delays.First, as controlled objectives, we focused electro-hydraulic servosystems, which contain nonlinear elements and uncertainties and are difficult to be modeled exactly. We constructed a linear local model in the neighborhood of each selected operating point of the electro-hydraulic system, and built an LPV global model by using load variations as a scheduling parameter. For the LPV systems, we derived the worst-case optimal control strategy, and demonstrated the effectiveness of the control strategy in the computer simulations and in the real plants. Second, as a typical difficulty in controlling the plant based on the local models, we focused the problem caused by information delays in the state, input and output. In particular, we investigate how to conquer the input and output delays in controlling the plant. The worst-case optimal control scheme was proposed; the finite-dimensional algorithm for constructing the control scheme was derived, the predictor-observer structure of the control scheme was clarified, and the illustrative numerical simulations were reported.