Data-Driven Computation of Minimal Robust Control Invariant Set
Data-Driven Computation of Minimal Robust Control Invariant Set
复制标题
最小鲁棒控制不变集的数据驱动计算
DOI：
10.1109/cdc.2018.8619312
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
N. Ozay
中科院分区：
文献类型：
--
作者：
Yuxiao Chen;H. Peng;J. Grizzle;N. Ozay
我们提出了一个数据驱动的框架来计算一个最小鲁棒控制不变量集（mRCI）的近似对于一个不确定的动态系统，其中系统的模型也是未知的，应该从数据中学习。首先，通过从实验数据中提取的一组线性约束来表征可接受模型集。可接受模型集中的每个模型都包含有关名义模型的信息，以及模型不确定性的表征，包括加性和乘性不确定性。然后提出了一种基于鲁棒优化的迭代算法，在计算最小鲁棒控制不变量集的同时，从允许集中选择最优模型。数值结果表明，与采用最小二乘方法依次选择模型并计算模型的基准方法相比，该方法大大减小了不变量集的大小。
We propose a data-driven framework to compute an approximation of a minimal robust control invariant set (mRCI) for an uncertain dynamical system where the model of the system is also unknown and should be learned from data. First, the set of admissible models is characterized via a set of linear constraints extracted from the experimental data. Each model in the set of admissible models contains information about the nominal model, as well as the characterization of the model uncertainty, including additive and multiplicative uncertainties. Then an iterative algorithm based on robust optimization is proposed to simultaneously compute a minimal robust control invariant set while selecting an optimal model from the admissible set. The numerical results show that the proposed method greatly reduces the size of the invariant set compared to a benchmark method that sequentially selects a model with least squares and then computes the invariant set.