거리 초음파센서를 이용한 수중 로봇 SLAM 용 직선 특징 검출 알고리즘

Abstract

This paper presents a sonar feature extraction method for building a map and localization of intelligent underwater robots in structured environments. Many mobile robots use ranging sonar(ultrasonic) sensors to measure the distance to the nearest objects. An elaborate ranging sonar sensor can detect obstacles only when the sonar beam hits the target at an incident angle of nearly 0°. Using this property of the ranging sonar sensor, this paper proposes the method which generates a line feature normal to the sonar beam at the hit point. The line feature extraction algorithm is incorporated into EKF-SLAM to reduce the uncertainty of the line features and robots pose simultaneously. As the robot moves and gathers distance measurements from the ranging sonar sensors, EKF-SLAM algorithm extends the existing line features or generates a new feature depending on the data association results. The simulation shows that EKF-SLAM with this extraction method gives stable line landmarks. Preliminary experiments in the test pool show the feasibility of the proposed approach.