GNSS-antenna lever arm compensation in aided inertial navigation of UAVs

Abstract

This paper considers aided inertial navigation of unmanned aerial vehicles aided with position measurements from one or more global navigation satellite system antennas, where the exact positions of the antennas are assumed unknown. This reflects that the antennas' location relative to the inertial sensor, i.e. the lever arms, might be difficult to measure accurately in the coordinate frame of the inertial sensor. Using inaccurate lever arm values will deteriorate both the position and attitude estimates of the vehicle. It is easier to manually and accurately measure the distances from each antenna to the inertial sensor and between antennas, and this information can be used constructively to estimate the lever arms online. In this paper, the distance information is used to reduce the representation of one or more lever arms to two or three states, respectively. An error-state extended Kalman filter is derived and compared to two other similar filters in simulations: one filter in which the lever arms are known and one which represents all lever arms as body-fixed positions. The suggested filter is shown to perform as well as the latter, but with a significantly smaller state space representation.