During the construction robot indoor operation, how to realize the accurate solution of the odometer has vital influence on the subsequent positioning and mapping and precise operation. The traditional simultaneous localization and map building (SLAM) method has the accuracy problem due to loopback detection. To address this issue, a method using building information modeling (BIM) data to correct the cumulative error of laser odometer is proposed to achieve precise positioning. Firstly, the global initial positioning of the robot in the BIM is solved at the initial moment. Secondly, the key points of the 3D point cloud are extracted and converted into 2D data. Then, the data of the wheel odometer are set as the predicted value to solve the inter-frame transformation. Finally, the BIM data are combined to eliminate the cumulative errors and obtain a high-precision odometer positioning. Experimental results show that this method has good stability and accuracy in robot initial positioning, laser point cloud processing and motion solution for eliminating accumulative errors. The initial positioning error is less than 2 mm, and the odometer offset error is controlled within 0. 09% , which provides a strong guarantee for the accurate establishment of the subsequent point cloud map.