Real-Time LQG Robotic Visual Tracking

In this paper, a modified LQG technique is proposed for the solution of the robotic visual tracking problem (eye-in-hand configuration). The problem of mbotic visual tracking is formulated as a problem of combining control with computer vision. A crosstorrelation method provides the object's motion measurements which are used to update the system's measurement vector. These measurements arc fed to a discrete steady state Kalman filter that calculates the estimated values of the system's states and of the exogenous disturbances. Then, a discrete LQG controller computes the desired motion of the robotic system. Experimental results are presented to show the effectiveness of the approach.