Square Root Unscented Kalman Filter based ceiling vision SLAM

This paper proposes a new approach of monocular ceiling vision based simultaneous localization and mapping (SLAM) by utilizing an improved Square Root Unscented Kalman Filter (SRUKF). With a monocular camera mounted on the top of a mobile robot and looking upward to the ceiling, the robot only needs to process salient features, which greatly reduce the computational complexity and have a high accuracy. SRUKF is used instead of the standard Extended Kalman Filter (EKF) to improve the linearization problem in both motion and perception models. To address the numerical instability problems in the standard SRUKF, several optimization methods are utilized in this paper. Experiments are performed to illustrate the effectiveness of the proposed approach.