Author/Editor		Ude, Aleš; Dillmann, Ruediger
Title		Trajectory reconstruction from stereo image sequences for teaching robot paths
Type		članek
Source		In: Proceedings of the 24th international symposium on industrial robots; 1993 Nov 4-6; Tokyo. Tokyo: Japan industrial robot association,
Publication year		1993
Volume		str. 407-14
Language		eng
Abstract		This paper presents a new method for programming of robot trajectories. We employ the "'Teaching by Showing" programming paradigm which enables the user to efficiently program robot tasks by means of vision. The user specifies the desired trajectory by moving the object to be manipulated with his own hand. The motion is measured by a stereo vision system. We take advantage of a CAD model of the object in order to facilitate the process of measurement. Smoothing vector splines are utilized to reconstruct the trajectory either in Cartesian or in joint coordinates. Uncertainties resulting from the noise of the sensors are considered explicitly in the process of reconstruction. Important points that must be reached by the robot's end-effector exactly can be specified and limitations on robot's velocities and accelerations can be set. The conversion of the reconstructed trajectory into the robot language commands is discussed. Some experimental and simulation results are presented to substantiate our approach.
Descriptors		ROBOTICS SOFTWARE COMMUNICATION