A design of detecting defect devices using ML in simulation robots

摘要

Industrial robots have motion patterns in performing tasks. These repetitive patterns contain data regarding position and electrical current of the motor moving the joints of the robot, but they are not always identical. Vibration of the working environment and degradation of the motor affect the joints of the robot so that they may not be in the same position as previous pattern. Because of this, motion control algorithms allocate different command values for joints to achieve the same target positions. When the difference between the normal and abnormal joint positions increases, the robot needs to be stopped for reconfiguration. This paper proposes a layout to detect the abnormal patterns of joint position data using machine learning techniques, and demonstrates the system in simulation.