TY - JOUR
AU1 - Zhu, Huazhong
AU2 - Jin, Zhehao
AU3 - Liu, Andong
AU4 - Ni, Hongjie
AB - Mobile robots equipped with visual sensors are widely used in challenging unstructured environment due to their flexibility. However, dynamic properties of actuators are generally neglected when modeling mobile robots, which may reduce the performance of the servo controller. In this paper, we present a cautious model predictive control method for visual servoing of mobile robots with unknown actuator dynamic properties. Firstly, an enhanced model constructed by a nominal and an additive Gaussian process (GP) model is learned on-line, where the GP model is stochastic and captures dynamic properties of actuators by using the training data. Furthermore, a stochastic model predictive control (SMPC) formulation is presented for cautious control where the chance constraints of predictive states are considered to ensure visibility of the feature point. For solving the SMPC problem, an augmented deterministic model (ADM) that represents the uncertainty propagation of the stochastic state is presented to transform the SMPC formulation to a deterministic model predictive control (DMPC) formulation. Then, the DMPC problem is solved by employing a modified iterative linear quadratic regulator (iLQR) with a Lorentzian ρ\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$\rho $$\end{document}-function introduced in the terminal cost function. Finally, the validity of the proposed method is validated by several examples.
TI - Gaussian process-based cautious model predictive control for visual servoing of mobile robots
JF - Nonlinear Dynamics
DO - 10.1007/s11071-023-08987-6
DA - 2023-12-01
UR - https://www.deepdyve.com/lp/springer-journals/gaussian-process-based-cautious-model-predictive-control-for-visual-8TcmCTL6NS
SP - 21779
EP - 21796
VL - 111
IS - 23
DP - DeepDyve
ER -