Visual Control of a Mobile Robot Using Homography and Learning from Demonstration Methodology

Abstract

The paper analyzed the robustness of the control system based on information obtained from the camera to unpredictable disturbances in a static environment such as the estimation of the distance to the selected object and/or a sudden change in lighting. Experimental results on the anthropomorphic robot named "NeuroArm Manipulator System" (with the use of the CMUCam 3 camera) indicate the necessity of implementing an empirical subsystem based on machine learning in order to successfully manipulate work-pieces within manufacturing systems for the production of sheet metal parts.