3b238e271970461e6db59a5 20250526095403934 beie:beie director@blueeyesintelligence.org WEB-FORM International Journal of Recent Technology and Engineering (IJRTE) IJRTE 22773878 10.35940/ijrte.2277-3878 https://www.ijrte.org/ 05 03 2025 14 1 Department of Information Science and Technology, National Institute of Technology (KOSEN), Yuge College, Ehime Prefecture, Japan. Prof. Hirofumi Maeda https://orcid.org/0000-0002-7145-3889 In Japan, the number of sewer culverts with a service life of 50 years has increased, and aging facilities have become apparent. In recent years, therefore, robots have come to be used for pipe inspections. Thus, the robot must grasp the situation and perform actions to prevent tipping over in the pipe. Consequently, we have decided to explore a software approach to prevent tipping using driving control and aim to realize highly accurate selfposition estimation, which is necessary for this approach. Although we have established a basic localization method based on our previous research, there is a small estimation error due to the influence of tire shape. Therefore, we propose a highly accurate localization method using a neural network to compensate for this estimation error. However, a large amount of teacher data was required to achieve this. For this reason, this research develops a three-dimensional position-measuring instrument that can quickly perform many measurements with high accuracy using two commercially available manipulators. This paper states a three-dimensional position-measuring machine that can quickly perform high-precision and highvolume measurements using two commercially available manipulators. The paper shows how to connect two manipulators utilising a ball joint and proposes a highly accurate measurement theory using an IMU. Furthermore, we confirm that the accuracy of the contact-type three-dimensional position-measuring is within 1 mm on average for translation error, within 1 deg on average for angular error, within 1 mm for the standard deviation of translation error, and 1 deg for the standard deviation of angular error except for 0 γp3, even in the initial state without calibration. This result shows that the method can be used in a neural network to correct estimation errors. 05 03 2025 05 30 2025 1 8 CC-BY-NC-ND 4.0 10.35940/BEIESP.CrossMarkPolicy www.ijrte.org true International Journal of Recent Technology and Engineering (IJRTE) Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP) Based on my understanding, this article has no conflicts of interest. This article has not been funded by any organizations or agencies. This independence ensures that the research is conducted with objectivity and without any external influence. The content of this article does not necessitate ethical approval or consent to participate with supporting documentation. The adequate resources of this article are publicly accessible. The authorship of this article is contributed solely. 10.35940/ijrte.F8214.14010525 https://www.ijrte.org/portfolio-item/F821413060325/