A Robot 3C Assembly Skill Learning Method by Intuitive Human Assembly Demonstration

With the development of the Internet and computer technology, the demand for 3C products such as mobile phones has surged. At the same time, due to the continuous improvement of labor costs, it is urgent to automate the 3C assembly lines with industrial robots. The motivation of this paper is to propose an efficient off-line programming by demonstration method to automate 3C assembly lines. This process consists of two phases. In the first phase, the optical motion capture device is used to capture the position and orientation information of human hands during assembly process. In the second phase, those information of a couple of demonstrations are learned to derive a robot control policy. To do so, the local outlier factor based anomalous point detection algorithm as well as the trajectory segmentation algorithm inspired by density-based spatial clustering is utilized in advance to pre-process the demonstration data. Then the human assembly skill represented by a probabilistic policy is learned from those data to drive the robot to accomplish the same assembly task under new environment. The Gaussian Mixture Model is utilized to pre-structure the policy. The offline programming by demonstration method efficiently transfers human experience to the robot. Experimental results demonstrate the effectiveness of the method.