Visual servo strategies using linear quadratic Gaussian (LQG) for Yaw-Pitch camera platform

摘要

Visual servo is a process of moving a robot to fulfill certain specific conditions using visual feedback. For a 2DOF platform such as yaw-pitch camera, the main task is to maintain the object in the camera field of view, precisely in the center of the image plane. A high performance of visual servoing is used to smooth the object tracking process. In order for the controller to calculate the necessary robot movements, a control law base on current image measurements and model is needed. This paper present visual servoing method with variation of Jacobian interaction matrix in Image-Based Visual Servo (IBVS) system, combine with Linear Quadratic Regulator (LQR) that allows the system to reach convergence and Kalman filter in order to compensate for the perturbing motions and LQR-Kalman filter pair can be denominated as a Linear Quadratic Gaussian (LQG) controller. The proposed system consists of a SONY FCB-EV7520 camera placed on a Yaw-Pitch Platform which consists of two dynamixel AX-12 servo motors. Both camera and servo platform are connected to a personal computer for the object tracking and servo control. Simulations of several tasks have verified the performance of the proposed method.