Narayanan Vignesh, Jagannathan Sarangapani, Ramkumar Kannan
IEEE Trans Neural Netw Learn Syst. 2019 Jun;30(6):1651-1658. doi: 10.1109/TNNLS.2018.2870661. Epub 2018 Oct 12.
In this paper, adaptive neural networks (NNs) are employed in the event-triggered feedback control framework to enable a robot manipulator to track a predefined trajectory. In the proposed output feedback control scheme, the joint velocities of the robot manipulator are reconstructed using a nonlinear NN observer by using the joint position measurements. Two different configurations are proposed for the implementation of the controller depending on whether the observer is co-located with the sensor or the controller in the feedback control loop. Besides the observer NN, a second NN is utilized to compensate the effects of nonlinearities in the robot dynamics via the feedback control. For both the configurations, by utilizing observer NN and the second NN, torque input is computed by the controller. The Lyapunov stability method is employed to determine the event-triggering condition, weight update rules for the controller, and the observer for both the configurations. The tracking performance of the robot manipulator with the two configurations is analyzed, wherein it is demonstrated that all the signals in the closed-loop system composed of the robotic system, the observer, the event-sampling mechanism, and the controller are locally uniformly ultimately bounded in the presence of bounded disturbance torque. To demonstrate the efficacy of the proposed design, simulation results are presented.
在本文中，自适应神经网络（NNs）被应用于事件触发反馈控制框架，以使机器人操纵器能够跟踪预定义轨迹。在所提出的输出反馈控制方案中，通过使用关节位置测量值，利用非线性神经网络观测器重建机器人操纵器的关节速度。根据观测器在反馈控制回路中是与传感器还是控制器位于同一位置，提出了两种不同的配置来实现控制器。除了观测器神经网络外，还利用第二个神经网络通过反馈控制来补偿机器人动力学中的非线性影响。对于这两种配置，通过利用观测器神经网络和第二个神经网络，控制器计算出扭矩输入。采用李雅普诺夫稳定性方法来确定事件触发条件、两种配置下控制器的权重更新规则以及观测器。分析了具有这两种配置的机器人操纵器的跟踪性能，结果表明，在存在有界干扰扭矩的情况下，由机器人系统、观测器、事件采样机制和控制器组成的闭环系统中的所有信号都是局部一致最终有界的。为了证明所提出设计的有效性，给出了仿真结果。
