Enhanced Altitude Estimation for Unmanned Aerial Vehicles in a GNSS-denied Environment

GNSS-denied environments represent challenging environments for autonomous drones. Ensuring an accurate altitude estimate plays a crucial role in guaranteeing mission efficiency in such scenarios. The literature on autonomous drone localization in GNSS-denied environments relies on visual inertial odometry-based algorithms. However, the reliability of this technology cannot be guaranteed in poor features environments (homogeneous floor, white walls) or in high and low brightness conditions. Even applications adopting loop closure algorithms in combination with visual inertial odometry localization suffer from considerable position estimation drift in challenging environments. This paper aims to propose a methodology to address the altitude estimation problem for drones operating in GNSS-denied environments by combining a V-SLAM algorithm with altitude measurements from a range finder. To account for ground inconsistencies due to varying terrain or obstacles, we have developed an Adaptive Kalman Filter. The Mahalanobis distance evaluation accomplishes the task of detecting these inconsistencies, enabling the filter to adapt and update states properly even in the presence of inconsistent range finder measurements. Experimental results demonstrate the effectiveness of the proposed solution in mitigating the drift accumulated by a purely V-SLAM algorithm.