The Design And Control Of Visual Servo Platform Based On Double-loop Structure

In recent years,vision has become a key component of mobile robot,which plays an important role in environmental monitoring,human-computer interaction and visual guidance.Especially when the robot enters the unknown environment,because the visual sensor has the advantages of massive information,non-contact and high precision,the robot can adapt to the environment and get the target feature quickly.However,most robots simply carry a fixed camera and change the direction of camera's measurement by moving the robot,it is difficult to track the moving target because of the large inertia of the robot and the small camera view field.The camera is mounted on a small platform with several degrees of freedom of movement,then the visual servo platform is fixed on the mobile robot,which makes the camera more flexible and can realize tracking fast maneuvering target.In this paper,the visual servo platform of the mechanical,hardware,control scheme and controller design are discussed and studied.Aiming at the defect of low sampling rate of camera,this paper proposed a double-loop control scheme based on the conversion between angle and pixel,the internal and external loop controller design method is given respectively,the reliability and validity of the visual servo control scheme and controller design are verified on the platform,the results of the experiments show that the visual servo performance is greatly improved by using the design method of this paper,which satisfies the conditions of the general mobile robot.The innovation of this paper is mainly embodied in the design of angle loop and image loop controller.In the design of angle loop controller,in order to reduce the tracking error of low-frequency signal and reduce the overshoot of system response,the proportional integral(PI)controller is added in the state feedback design,and a novel parameter design method which can restrain overshoot is adopted.The design of parameters is divided into two steps,the state feedback gain is designed after the inner mode expansion of the controlled object,then the adjustable parameters are adjusted for the overshoot of the step response of the angle loop.Because the state of the angle loop is not measurable,the state observer is designed,then the output feedback control based on LQR is formed in the whole.In the image loop's controller design,the feedback controller is designed firstly,then the equivalent compound control is implemented based on the pixel miss distance and the position of the platform.Because of the delay in estimation of the target's position and the phase lag of feedforward controller,the visual tracking effect of tracking target is still unsatisfactory.Finally,the compound control based on Kalman prediction is adopted.