Object Tracking for Fine-Tuning of Robot Positions

Abstract

In many complex applications an accurate model of the plant is not known. Consequently, complementary methods are needed to automatically achieve accurate dynamical positioning of a robot in relation to its surroundings. This thesis describes the development of a control strategy on vision-based object tracking for a robot manipulator. To ensure necessary robustness we assume that four distinct, circular shapes are visible on the face of the object to inspect. Based on this information, along with knowledge of the camera parameters, the position and the orientation of the object are estimated. The developed system relies on the use of an open-source vision library, ViSP. A Kalman filter is used to predict future states of the moving object, in order to reduce tracking errors introduced by the response time of the system.