Abstract: At present, the pose of drilling robots is mostly adjusted by manual remote control and manual retest in combination. Full automatic adjustment has not been achieved. Besides, due to the low precision of open-loop control and inadequate automation capability, the accurate hole positioning and automatic construction of hole groups cannot be realized by drilling robots with water exploration and discharge, outburst prevention and punching prevention in coal mines. Herein, the kinematic model of drilling arm was established by analyzing the structure and action of the drilling robot’s drilling arm. Meanwhile, the dip angle factors such as machining error, body deformation and assembly clearance were analyzed. It is found that the dip angle and azimuth error will be enlarged because of the structure clearance of worm gear of the rotary reducer, which may be up to 0.85° in the body. Firstly, in order to eliminate the errors, the static and dynamic error compensation model of the drilling arm was established using the traditional kinematics error compensation method of the drilling arm. Besides, the joint clearance and deformation were measured by a total station. Then, the error compensation was obtained by inverse solution based on the compensation model and RBF neural network method. The average error between the expected and the actual pose of drilling arm is 9.6 mm in x direction, 18.2 mm in y direction, and 16 mm in z direction, which meets the requirements of engineering application. Secondly, in order to solve the complexity and non-real-time problems of the traditional test measurement method by a total station, a method to detect the pose error with the laser rangefinder and the high-precision open-hole orienteer in combination was proposed. Thereby, the difference between the actual and theoretical dip angle and azimuth were calculated in real time through ranging, which was used as the control input of the new dip angle and azimuth after error compensation for the real-time error detection and compensation of drilling arm. Finally, the pose error compensation model of laser ranging combination was verified using the traditional total station accuracy detection method. The test shows that: The maximum error difference and the maximum azimuth error obtained by the positioning error detection method of laser ranging combination are within ±0.5° and within ±0.5°, respectively, which are 41.1% and 37.5% higher than that before error compensation, meeting the requirements of drilling robot opening hole positioning error. In this paper, research was firstly carried out on the online accurate positioning and error compensation of the drilling robot’s drilling arm in coal industry, which has important reference and guiding significance for the accurate automatic hole positioning of drilling robots and the automatic construction of hole group.