Monte Carlo Localization of Underwater Robot Using Internal and External Information

Abstract

This paper proposes a method for localization of an underwater robot. The method uses Monte Carlo algorithm called the particle filter. It predicts the pose of the robot using the internal sensor information from thrusters, inertial sensors, and electronic compass. A correction procedure follows the prediction. The correction uses external sensor information, that is, the distance from landmarks whose locations are known a priori. The prediction and correction process use samples of robot pose in stochastic and probabilistic approach. Though the external information available from the sensors could include depth, angle and angle rates of yaw, pitch, and roll, the proposed method uses only the distance from some beacons. In contrast to the classical methods which usually use either trilateration principle or dead reckoning to calculate the pose, the proposed approach fuses motion and internal sensor information with the external sensor information. The simulation shows that localization is possible even if only one or two beacons are available for range measurement. The experiments which uses two beacons in a tank suggest that the proposed method can be effective where the number of beacons is limited due to geographical features of the robot work area.