무인 운반체의 항법을 위한 위치, 자세 및 지도 추정 방법

Abstract

This study implements estimation of location and attitude of unmanned vehicles and builds map of the environment. Extended Kalman filter is used for estimation of location and attitude, and graph-based optimization is utilized for the map building. The extended Kalman filter is applied using bias-calibrated angular rate. The method estimates bias in angular rate measured by an Attitude Heading Reference System(AHRS). Also, it fuses multiple sensor measurements for localization of an underwater vehicle. The sensors used for the fusion are Inertial Measurement Unit(IMU), Short Baseline(SBL), Doppler Velocity Log(DVL), and depth meter. Two methods are implemented for map building : Invariant Extended Kalman filter(IEKF) SLAM and Graph-based SLAM. IEKF SLAM is an on-line SLAM which results in robuster performance than EKF SLAM. The IEKF SLAM estimates the location of a robot as well as the location of landmarks. The Graph-based SLAM is an off-line algorithm which is a optimization approach based on graph representation of constraints. These methods are verified using open measurement data set and test-tank experiments.