IMU Position Tracking Improvement Through Filtering

Abstract

When GPS, GNSS, or other position reporting systems cannot be used, position can be estimated through a process called dead reckoning. The position is calculated by taking a previously known position and using acceleration and velocity to determine the current position. Inertial Measurement Units (IMU) can be used for such applications. IMUs report the acceleration acting on the sensor. This measurement is then integrated once to give velocity and then again to give position. Due to the multiple layers of integration to get position data out of the sensor, noise and offset cause compounding error in the estimated position[2]. These adverse effects on the output of the IMU can be mitigated through filtering. This project implements three different types of filters and compares the position tracking results. The filters implemented are low/band-pass filtering, linear fitting, and Kalman filter as described in Robust M-M unscented Kalman filtering for GPS/IMU navigation [1].