Correcting Current-Induced Magnetometer Errors on UAVs: An Online Model-Based Approach

Abstract

The yaw attitude of an unmanned aerial vehicle (UAV) is important for navigation applications. Magnetometers are attractive because they can directly resolve the yaw attitude of the UAV. Roboticists, however, often dismiss magnetometers because of the instrument’s susceptibility to unwanted magnetic sources, such as hard irons, soft irons and electric currents generated by the UAV’s powertrain. Dynamic current-induced magnetometer biases are especially hard to fix because of their time dependency. A hardware fix is to isolate or shield the current carrying wires of the powertrain from the magnetometer. However, for UAVs with weight and space restrictions, this solution may not be feasible. An alternative is to fix the errors using software. This work takes the software approach. Specifically, a model for the current-induced magnetometer bias as a function of the throttle command is established. Based on this model, an adaptive estimator is developed that determines the model parameters in realtime. The advantage of our method over existing techniques is the ability to handle rapid changes in throttle command, thanks to the inclusion of the bias model. Experiments show the estimator can compensate for the current-induced magnetometer bias across all throttle settings and yaw angles.