Visual Servo Based Predictive Control Of Nonholonomic Mobile Robot

With the rapid development of robot kinematics, control theory, image processing technology, computer hardware technology and cameras, visual servo control system is becoming a hot topic in the field of robotics research. It has had a great achievement both in theory and application when combine machine vision, neural network, intelligent control and information fusion technology together. Mobile robot visual servo control is a technology that use visual sensors to obtain environmental information, controlling mobile robot to perform the corresponding action.This article discusses the related technologies of the wheeled mobile robot visual servo control system, the traditional control methods without considering the nonholonomic constraints brought in by the wheels that have an impact on the mobile robot control, so the main contribution and innovation of this article is to introduce the model predictive control method, and applying the primal-dual neural network to solve the convex optimization problem online.Model predictive control have three basic characteristics: predictive model, rolling optimization and feedback correction, it has unique advantage on solving the problem of the mobile robot motion control. Traditional research methods usually ignore or simplified the constraints of the robot kinematics and dynamics, but this kind of constraints have significant impacts on control performance, model predictive control can deal with the constraints conveniently and obviously. According to the differential kinematics equation that obtained from the Cartesian coordinates and polar coordinates, the objective function can be formulated as a quadratic programming problem by applying model predictive control, then using primal-dual neural network to solve the above mentioned quadratic programming problem over a finite receding horizon to achieve the control task of visual servo system. Feasible solution of linear matrix inequalities that guaranteed asymptotic stability of the visual servo system. Using the rolling optimization and feedback correction characteristics of model predictive control, it can effectively reduce or even eliminate the influence of the delay problem of closed-loop systems, and can be easily combined with the predicted future information to optimize for motion control, then achieved the purpose to enhance the control performance.In this paper, we achieved visual servo of nonholonomic mobile robots based model predictive control, including visual stabilization, trajectory tracking, and vision-based formation control. Finally, three experiment results are provided to illustrate the effectiveness and the reliability of the model predictive control scheme to achieve visual servo of nonholonomic mobile robots.