In order to solve the problems that modeling was complex and a large computation was costed for calibration and compensation of aviation drilling robot, a calibration and compensation method based on extreme learning machine was proposed. The aviation drilling robot was regarded as a black-box system in this method which ignored the influence of geometric factors and non-geometric factors of robot. Then, according to robot positional errors measured by a high-accuracy laser tracker, a robot error prediction model based on extreme learning machine was trained and established. Next, the positional error in desired position could be predicted by robot error prediction model and the robot position was compensated to achieve the robot calibration. Final, experimental studies were carried on an aviation drilling robot. The experimental results showed that the mean and maximum positional error of robot was reduced by 75.69% and 78.16%, respectively.