In order to effectively manipulate indoor mobile robots, it is important to localize their positions precisely. Although a number of different ways using ultrasonic beacons have been developed, these methods are restricted to a relatively small area because all the beacons should be localized before any actual measurement. In this paper, a new methodology is developed to improve the accuracy of the auto-calibration process in the localization method. The new three-point-extraction algorithm is used in conjunction with existing trilateration and auto-calibration algorithms to compose a new localization method. Validity of the new algorithm is verified through experiments for two and three triangular areas, where the localized path error is confirmed to be significantly small in average. In addition, the path error turns out not to accumulate to a considerable extent. The result enables the robot to autonomously localize its path by recognizing the accurate positions of newly appearing beacons.