Robust MAV State estimation using an m-estimator augmented sensor fusion graph

The recent surge in the MAV industry has brought many new prospective commercial applications for MAVs. Due to their versatile flight capabilities, MAVs are capable of operating in urban and human centric environments. However navigating in these environments becomes an increasingly difficult task as they are commonly GPS challenged environments as well. Low satellite visibility, multipath, and Non-Line of Sight (NLOS) errors degrade the GPS-derived navigation solution preventing fully autonomous flight. Before widespread commercial adoption of MAVs can occur, navigation within GPS-challenged environments needs to be safe and reliable. In this paper we present a sensor fusion graph augmented by m-estimators for MAV trajectory estimation in a GPS- challenged environment. We take a probabilistic approach to tight coupling of GPS and IMU sensor data. We then apply m-estimation to our sensor fusion graph in order to mitigate the multipath and NLOS errors that would otherwise skew our navigation solution. Finally we demonstrate the effectiveness of our algorithm by flight test of an As- cTec Firefly MAV, navigating from an open-skied environment into a GPS-challenged environment and estimating its trajectory.