Research Of Mobile Robot Visual Servoing Control Strategy Based On Active Vision

Development of the discipline by means of automatic control theory, image processing,filtering techniques and computer vision, vision sensors have been widely applied to thefield of robotics. The applications of Vision sensors are mainly visual navigation, visualtracking, visual servo and so on in the robot control. The research in these areas greatlyimproves the flexibility of the robot, intelligence and precision. Among them, the visualservoing is defined as to make the stabilization control for the machine through the use ofvisual feedback,drives a robot to reach the desired position/orientation.It is a hotspots anddifficult problems in visual control research field. Although scholars in related areashave been proposed many methods to solve this problem, Today, however, many of the keyissues has not yet been satisfactorily resolved.The constraints for the field of view (FOV) in visual servoing study, the authorproposes a mobile robot based on the dual-freedom pan-tilt camera adaptive initiativevisual servoing theoretical framework.Then makes further research for related issues. Thework of this paper is mainly focused on the following aspects:(1) A mobile robot based on the dual-freedom pan-tilt camera active visual servoingcontrol theoretical framework is presented. Specifically, through adjusting in real time thepitching Angle and yaw Angle of pan-and-tilt camera, the camera’s optical axis alwayspoints to the reference geometry center, in order to solve the FOV constraint problems.Themethod can achieve optimum path,which is different from the nonactive adaptive visualservo system. Different from the conventional monocular vision,the changes of the pitchangle of pan-tilt make the vision of mobile robot system more open,and make full use oftargets having obvious characteristics in the surrounding environment, Such as ceiling,ventilation system, signs, and even line border, roof beam and wall Angle and so on,fulfilling Visual servoing task. In addition,it can be easily combined with sensors such aslaser rangefinder, to make up for the monocular visual defects such as lack of informationof depth.(2) The kinematics model of the system is set up. This article introduces pin-holeimaging model of the camera, the camera intrinsic parameters and extrinsic parametersmodel; establishes a conversion relationship equation of the pan-tilt system and the robotcoordinate system; establishes a geometric model of the ground mobile robot active visualtracking system; And deduces in detail the kinematic model of the target geometric centerof the image plane, Namely derives an equation of the movement speed of arget geometric centerin the image plane and linear velocity of the robot in the world coordinate system, theangular velocity of the robot in the world coordinate system, the rotational angular velocityof pan-tilt in the robot coordinate system and translation external parameters of the camera,camera intrinsic parameters, image error. This text proves the reversibility of image Jacobimatrix produced by a rotating angular velocity of pan-tilt in the robot coordinate system.This is no doubt that a solid foundation has been laid for the control design and systemperformance analysis.(3) Design the active adaptive visual controller with the camera motion compensation.The method achieves the camera’s optical axis always pointing to the reference geometrycenter by adjusting in real time the pitching Angle and yaw Angle of pan-tilt camera,inorder to solve the FOV constraint problems.(4) This text proposes a mobile robot based on the the dual quaternion active visualpositioning method for the more degrees of freedom relative movement between cameracoordinate system and world coordinate system in this article, relative motion complexcharacteristics. The method is effective to avoid the calculation of singularities inconversion processing of the camera and the coordinates of the world coordinate system,ing the optimal solution of the rotation matrix and translation vector at the same time, thesmaller amount of calculation,higher calculation precision. Finally, simulation resultsshow the effectiveness of the algorithm.(5) Some related research for wheeled mobile robot of Non-integrity constraints issuesare done. On the premise of implementation of monocular vision mobile robot real-timemotion state estimation, it achieves based on the location of the mobile robot visualservoing control. the simulation verifies the finiteness of stabilization control algorithm.