Robot movement sequence synthesis from images

A technique is developed to synthesize movement sequences from images without CAD models of the objects or the scene and without camera calibration. The Karhunen-Loeve (K-L) transform is utilized to acquire image features from sample images. This paper establishes the nonlinear relationship between robot joints and image features. A joint space trajectory is first generated and used to synthesize the robot movement sequence. The virtual movements of UP6 manipulator are obtained based on the experiments. The results show that the technique is convenient and accurate.