Abstract: Aiming at the poor positioning reference, complicated calibration process, incomplete detection parameters and other difficulties in the relative position sensing of the joint unit of digging, supporting and anchoring robots under the complex environment of high dust and fog and low illumination in the face of coal mine excavation, an infrared thermal imaging-based relative position detection method for the joint unit of digging, supporting and anchoring robots is proposed. Firstly, based on the self-developed joint unit of digging and supporting anchor robot, a relative position measurement system consisting of infrared thermal camera, heat source target, and stepping anchor supporting equipment was constructed; the infrared thermal imaging camera captures the signal source of the heat source target mounted on the combined unit of digging and supporting anchor robots and carries out the aberration correction and image enhancement processing on the output thermal imaging image; Secondly, the UNet segmentation algorithm is used to segment the region of the heat source target signal source, and then the image binarisation and morphological processing are used to refine the features of the segmented region of the signal source and extract the pixel coordinates of the centroid of the signal source by combining with the OpenCV library function; Again, the P4P method is used to solve the spatial 3D coordinates of the centre points of the four groups of signal sources, and the relative position parameters of the combined unit of the digging and supporting anchor robots are solved by the coordinate system transformation; Then, the relative positional parameters measured by the Realsense D435i depth camera combined with the singular value decomposition algorithm are used as the true values, and the measurement errors of the proposed method are further calculated to verify the feasibility and effectiveness of the proposed method; Finally, an experimental platform for relative position detection of the combined unit of digging, supporting and anchoring robot based on infrared thermal imaging was built, and the experimental results show that, in the measuring range of 2.40～4.80 m, under the coordinate system of the stepping anchor supporting equipment, the absolute mean values of relative errors of the digging machine's translations along the three axes of X, Y and Z are 0.0216 m, 0.0198 m and 0.0269 m, and the absolute mean values of relative errors of the rotational angles of the digging machine around the three axes of X, Y and Z are 1.45°, 1.07° and 1.27°, which can meet the needs of relative position sensing in the complex environment of the coal mine.