Direct teaching algorithm based on task assistance for machine tending

Abstract

Much research has been done on direct teaching in which an operator directly teaches a robot its task by hand guiding. In this study, we propose a novel algorithm to add virtual stiffness to a robot based on its task without changing the mode of teaching. The proposed algorithm defines the task assistance domains with spherical, cylindrical, and hexahedral shapes depending on the task environment. Inside this domain, stiffness is added to assist the operator in teaching the robot. At the boundary of the task assistance domain, stiffness is gradually changed to avoid the instability due to an abrupt change in stiffness. The proposed method was implemented on a 6 DOF industrial robot performing direct teaching for machine tending to demonstrate its feasibility. The experiments show that the operator can perform direct teaching more efficiently and this algorithm is particularly good for repetitive teaching on a small batch production.