This article designed a soft bending actuator (SBA) and presented a neural-network-based tracking control strategy for such an actuator. To achieve high-precision control, a visual feedback system was established by using a high-speed camera to dynamically compute the corresponding central angle which described the curvature of the SBA. Considering its nonlinear features, the echo state Gaussian process, a fusion of echo state network and Gaussian process, was used to approximate the SBA's dynamics with a nonlinear autoregressive exogenous model. On this basis, a single-layer neural network was trained to calculate the control signal in light of the idea of predictive control, due to its simplicity and good approximation capabilities. The stability of the closed-loop system was guaranteed, and experimental results indicated that the proposed strategy showed relatively high tracking accuracy under various reference trajectories.

中文翻译：

---

具有回波状态高斯过程的基于视觉伺服的软弯曲驱动器预测控制

本文设计了一种软弯曲致动器（SBA），并提出了一种基于神经网络的跟踪控制策略。为了实现高精度控制，通过使用高速摄像机动态地计算描述SBA曲率的相应中心角，建立了视觉反馈系统。考虑到其非线性特性，使用回波状态高斯过程，回波状态网络和高斯过程的融合，通过非线性自回归外生模型来近似SBA的动力学。在此基础上，由于其简单性和良好的逼近能力，根据预测控制的思想对单层神经网络进行了训练，以计算控制信号。保证了闭环系统的稳定性，