A workpiece is gripped by a robot hand and an image of the workpiece is captured by a camera. An image processing device detects the position and posture of a characteristic portion of the workpiece. On the basis of the present position of the robot, the relative position and posture between a flange of the robot and the workpiece characteristic portion is detected. The relative position and posture is compared with that observed when the workpiece is gripped correctly, to determine a gripping error. If the gripping error exceeds a permissible error, the robot is stopped. If the gripping error is equal to or less than the permissible error, a taught position where the workpiece is to be released is corrected so as to cancel the adverse effect of the gripping error.