Design Methodology for Super Energy-Saving Control Systems-Discrete-valued/Saturated-input Control Approach based on Online Optirnization

本文研究了一类输入状态幅值饱和或由离散值信号驱动的约束系统。研究的目的是为此类系统开发系统的设计方法，并在低分辨率执行器下获得令人满意的控制性能。得到了以下结果：首先，针对输入饱和的线性系统，提出了基于离线优化的稳定性分析方法和参考整形方法，提高了系统的跟踪性能。通过数值算例验证了该方法的有效性。其次，对于基于在线优化确定受系统约束的控制的模型预测控制方法，我们开发了两种类型的方法。一是利用对象的先验知识来减少在线优化的计算负担。二是在模型预测控制方法中对对象模型进行自适应改进。通过仿真验证了其有效性。第三，对于离散值输入控制，我们得到了一些结果。其中一个要点是利用动态量化器，根据原始连续值信号的过去序列，将其转换为给定的离散值集合中的信号。给出了线性连续系统动态量化器的一种简单设计方法，并提出了一种基于被控对象详细信息的离散系统动态量化器的优化设计方法。许多数值和实验结果支持这种动态量化器的实用性。在此基础上，提出了一类非线性系统的鲁棒控制方法。

This study considers a class of constrained systems, which have amplitude saturation in input/state or are actuated by discrete-valued signals. The purpose of the research is to develop systematic design methods for such systems and to achieve satisfactory control performance with low resolution actuators. The following results have been obtained.First, for linear systems with input-saturation, we have developed a stability analysis method and a reference shaping method which improves the tracking performance based on off-line optimization. Its effectiveness is demonstrated through various numerical examples.Second, as for model predictive control methods which determine the control subject to the systems constraints based on online optimization, we have developed two types of the methods. One is to reduce the computation burden concerning online optimization by exploiting a priori knowledge of the plant. The other is to improve the plant model in an adaptive way during the model predictive control method. Their effectiveness is validated through simulations.Third, as for discrete-valued input control, we have obtained several results. One of the main points is to utilize the dynamic quantizers which transform the original continuous-valued signals to the signals in the given discrete-valued set based on the past sequence of the both signals. We give a simple design method of such dynamic quantizers for linear continuous-time systems, and develop an optimal design method for discrete-time case which is based on the detailed information of the plant to be controlled. Many numerical and experimental results support the utility of such dynamic quantizers. Furthermore, a robust control method for a class of nonlinear systems has been presented.