해저 지형 정보를 사용한 Bayes 필터 기반 수중 로봇 위치추정

Abstract

In this paper, underwater robot localization method based on Bayes filter using seabed terrain information is proposed. This method is used to predict underwater robot pose using proprioceptive information and undergoes a process of updating the position using exteroceptive information. The advantages of Bayes filter method is estimated considering uncertainty of proprioceptive, exteroceptive information. In this paper, proprioceptive information for predicting the underwater robot position was only used robot command(truster speed, the angular velocity elevator, rudder angular speed). Exteroceptive information was only used altimeters. Estimation algorithms were used Extended Kalman filter, unscented Kalman filter and particle filter methods based on Bayes filter. Especially, UKF localization is a feature-based robot localization algorithm using the unscented kalman filter. The UKF uses the unscented transform to linearize the motion and measurement model. Instead of computing derivatives of these model, the unscented transform represents Gaussians by sigma points and passes these through the model. This paper was verified localization of the proposed EKF, UKF, PF methods through the simulator and the actual experiments.