Adaptive neural network-based saturated control of robotic exoskeletons
Adaptive neural network-based saturated control of robotic exoskeletons
复制标题
DOI：
10.1007/s11071-018-4348-1
复制
发表时间：
2018-06
影响因子：
5.6
通讯作者：
Hamed Jabbari Asl;T. Narikiyo;M. Kawanishi
中科院分区：
文献类型：
--
作者：
Hamed Jabbari Asl;T. Narikiyo;M. Kawanishi
在本文中，提出了用于连接机器人外骨骼的具有输入饱和度的新型自适应神经网络（NN）控制器。控制器由状态反馈控制器和输出反馈控制器组成。通过利用辅助动力学，控制器为这些机器人系统的输入饱和控制提供了一个新的框架，该框架可以具有状态反馈控制的全局稳定性。为了补偿系统的未知动态，采用了基于神经网络的自适应方案。此外，利用自适应鲁棒项来处理未知的外部干扰。稳定性研究表明，闭环系统与状态反馈控制器具有全局一致最终有界（UUB），其中基于神经网络的控制器的全局特性是利用平滑切换函数和鲁棒控制项来实现的。此外，该系统是带有输出反馈控制器的半全局UUB。通过模拟和实验测试验证了控制器的有效性。
In this paper, novel adaptive neural network (NN) controllers with input saturation are presented forn-link robotic exoskeletons. The controllers consist of a state feedback controller and an output feedback controller. Through utilizing auxiliary dynamics, the controllers provide a new framework for input saturated control of these robotic systems which can feature the global stability for state feedback control. To compensate for the unknown dynamics of the system, adaptive schemes based on NNs are exploited. Furthermore, adaptive robust terms are utilized to deal with unknown external disturbances. Stability studies show that the closed-loop system is globally uniformly ultimately bounded (UUB) with the state feedback controller, where the global property of the NN-based controller is achieved exploiting a smooth switching function and a robust control term. Also, the system is semi-globally UUB with the output feedback controller. Effectiveness of the controllers is validated by simulations and experimental tests.