Robust Model Predictive Control Strategies For Visual Servoing Of Mobile Robots

As an important means of observing and understanding the world,80% of the information of the outside world is obtained through vision.With the development of information technology and image processing technology,mobile robots can be endowed with visual cognitive function,which enables them to understand the world through images.In recent years,with the increasing application scenarios of mobile robots,using visual information to control them,namely visual servoing,has significant theoretical significance and potential application prospects.Considering the nonholonomic constraint and the visual visibility constraint in the field of visual servoing,the paper uses model predictive control(MPC)to design visual servoing controller.The problems studied and the main achievements achieved are as follows:(1)Considering the visual visibility constraint and dynamic constraint of mobile robots,a two-step visual servo stabilization MPC control strategy for mobile robots is proposed.The control strategy is divided into two parts: steering predictive controller and linear predictive controller.The steering predictive controller controls the robot steer to target position so that the feature points appear in the field of vision.The linear predictive controller controls the robot move along the straight line to reach the target point.(2)Considering the uncertainty of mobile robot model,a quasi-min-max predictive control strategy for visual servoing of mobile robot is proposed.First,linear parameter varying(LPV)model is used to describe the uncertainty of visual servoing model of mobile robot.Then,quasi-min-max is used to establish the optimization problem of receding horizon control of visual servoing of mobile robot.At last,linear matrix inequality(LMI)technology is used to transform the optimization problem into a semi-definite optimization problem and obtain the robust stability conditions of closed-loop system.(3)Considering the controller designed in(2)is too conservative and has a large amount of online computation burden,a quasi-min-max predictive control strategy with high efficiency for visual servoing of mobile robots is proposed.Referring to the design idea of multi-lyapunov functions min-max predictive controller,for various modes of mobile robot visual servoing LPV systems,Lyapunov functions are designed to obtain the upper bound of optimization functions to reduce the conservativeness of the controller designed in(2).Considering the additional computational burden,the offline strategy of the controller is designed.