The Global Navigation Satellite System (GNSS) is widely utilized to accurately determine the position of a vehicle, but at present, the accuracy may be degraded depending on the radio wave reception conditions from the satellite. There are also growing concerns about cyber-attacks on GNSS. We have been developing a method for estimating the position of a traveling vehicle using MEMS sensor data obtained from acceleration, gyro, and geomagnetic sensors. In the current work, we set up an experimental environment to collect data at 50 Hz and implemented a running test on an actual road. Optimizing the weighting of each item of sensor information enabled us to obtain characteristics of the road that could not be captured in the past. The difference of velocity between the reference vehicle and the evaluation vehicle is significant, the localization error increases. We therefore developed an algorithm to improve the accuracy even when the velocity difference of the two vehicles is significant and demonstrated its effectiveness through a field test.