期刊
IEEE TRANSACTIONS ON CYBERNETICS
卷 50, 期 11, 页码 4619-4632出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TCYB.2020.2969377
关键词
Neurons; Artificial neural networks; Observers; Recurrent neural networks; Delays; Delay effects; Dynamic event-triggered protocol (DETP); multidelayed neural networks (NNs); proportional– integral observer (PIO); recurrent neural networks (RNNs); sensor saturations
类别
资金
- National Natural Science Foundation of China [61873148, 61873169, 61933007, 61973102, U1509205]
- Zhejiang Provincial Natural Science Foundation of China [LR16F030003]
- Royal Society of the U.K.
- Alexander von Humboldt Foundation of Germany
In this article, the design problem of the proportional-integral observer (PIO) is investigated for a class of discrete-time multidelayed recurrent neural networks (RNNs). In the addressed RNN model, the delays occurring in the information interconnections are allowed to be different, and the phenomenon of sensor saturation is taken into consideration in the measurement model. A novel dynamic event-triggered protocol is employed in the data transmission from sensors to the observer with hope to improve the efficiency of resource utilization, where the threshold parameters are adaptive to the dynamical environment. By virtue of the Lyapunov-like approach, a general framework is established for examining the boundedness of the estimation errors in mean-square sense, and the ultimate bound of the error dynamics is also acquired. Subsequently, the explicit expression of the desired PIO is parameterized by using the matrix inequality techniques. Finally, a simulation example is utilized to verify the effectiveness and superiority of the proposed PIO design scheme.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据