Abstract In this paper, a Gaussian process-based nonlinear model predictive control (GP-based NMPC) algorithm is presented to deal with the visual servoing problem for constrained mobile robots. Firstly, a GP-enhanced model is established by incorporating a GP model and a visual servoing kinematic model where the GP-model is used to capture the robot dynamics with on-line updating. Then, a nonlinear model predictive control (NMPC) strategy is proposed to transform the visual servoing task into a nonlinear optimization problem with robot-physical and camera-visibility constraints. Subsequently, a variant iterative linear quadratic regulator algorithm is presented to solve the constrained NMPC problem in real time. Finally, simulations and experiments are conducted to show the effectiveness of the presented method.

中文翻译：

---

基于高斯过程的非线性未知动力学约束移动机器人视觉伺服预测控制

摘要 本文提出了一种基于高斯过程的非线性模型预测控制(GP-based NMPC)算法来处理约束移动机器人的视觉伺服问题。首先，通过结合GP模型和视觉伺服运动学模型建立GP增强模型，其中GP模型用于通过在线更新捕捉机器人动力学。然后，提出了一种非线性模型预测控制 (NMPC) 策略，将视觉伺服任务转换为具有机器人物理和相机可见性约束的非线性优化问题。随后，提出了一种变体迭代线性二次调节器算法来实时解决受约束的 NMPC 问题。最后，通过仿真和实验证明了所提出方法的有效性。