Robot system configured to perform learning control

A robot system having a robot mechanism unit including a sensor configured to acquire vibration data of a control target portion and a robot control device configured to control an operation of the robot mechanism unit according to an operation program, comprising

• a learning control unit configured to perform learning for calculating a learning correction amount for bringing a position of the control target portion detected by the sensor toward a target position while the robot mechanism unit operates according to the operation program, and
• a robot control unit configured to correct an operation command regarding the target position of the control target portion by using the learning correction amount calculated by the learning control unit and to control the operation of the robot mechanism unit according to the operation program,
• wherein the learning control unit includes
• a power spectrum calculating unit configured to calculate a power spectrum by Fourier transforming vibration data detected by the sensor for each learning cycle by the learning control unit,
• a comparison unit configured to compare for each frequency component a power spectrum at the time of the current learning with a power spectrum at the time of the immediately preceding learning, which are calculated by the power spectrum calculating unit, and
• a learning correction amount updating unit, in a case that as a result of comparison by the comparison unit, there exists a frequency component for which the power spectrum at the time of the current learning is greater than the power spectrum at the time of the immediately preceding learning, configured to adjust at least one of a phase and a gain of the learning correction amount used for correcting the operation command at the time of the current learning such that the power spectrum at the time of the current learning becomes less than the power spectrum at the time of the immediately preceding learning and to set the adjusted learning correction amount as a new learning correction amount used for correcting an operation command at the time of next learning.