Demonstration of robotic deburring process motor skills from motion primitives of human skill model

Parvizi, Payam
This study presents a new method to learn motor skills of robotic deburring process from a human expert who is performing manual deburring operation. By utilizing this method, it is possible to automate the robotic deburring process of a workpiece with an unknown shape. In this work, the task-related movements of the human expert are recorded using 6DOF and 1DOF haptic devices. Then, the intrinsic movement primitives are parametrized by using Dynamic Movement Primitive (DMP) method. By collecting dataset of parameters and weights of this method, trajectories of complex behaviors can be generated. Moreover, this method can adapt itself with respect to different start and goal positions. In order to move from one position to other position in two-dimensional space, rhythmic movements of human expert are extracted by using two-dimensional DMPs. In addition, different fundamental studies have been done on general concentrated-position DMP and local DMP. In general concentrated DMP, the experiments are conducted on 6DOF haptic device, in which the manual deburring processes on a workpiece are imitated and the results are used as primitives to accomplish general automatic deburring on the workpiece. Furthermore, Local DMP performs like the human automatic controller which its aim is to control force using movement primitives of the skills of a human expert in one dimension. The experiments of local DMP are conducted on 1DOF haptic device interacted with 6DOF grinding robot and piezoelectric actuator. Moreover, automation of robotic deburring process on unknown workpiece requires studies on tool deflection, collision, optimal velocity and surface shape search. Different experiments have been performed in order to eliminate collision and reduce form error on a workpiece. The results of experiments are analyzed by scanning the surface of workpieces. Finally, our process is compared with an alternative process and the time duration of each process is analyzed.
Citation Formats
P. Parvizi, “Demonstration of robotic deburring process motor skills from motion primitives of human skill model,” M.S. - Master of Science, Middle East Technical University, 2018.