SLAM for a small UAV with compensation for unordinary observations and convergence analysis

Abstract

This paper deals with the Simultaneous Localization and Mapping (SLAM) problem for a small Unmanned Aerial Vehicle (UAV) via extended Kalman Filter (EKF) with compensations for unordinary observations. In the SLAM problem, a robot sometimes loses its proper observations then the estimation accuracy deteriorates. In this paper, to remove the effects of unordinary observations, we propose a novel SLAM method considering unordinary observation based on Mahalanobis distance. The proposed method detects the unordinary observations by comparing the observation values with its estimation and determines the weight of these observations. The convergence of the state error covariance matrix is proven. In experimental validation, we show that the UAV state and environment information can be estimated with the proposed method.