Camera-IMU-based underwater localization

Abstract

Sensor-based localization, as one of the essential parts for robot navigation, continues to be a major research focus in mobile robotics. In this paper, we consider a more challenging localization scenario where an autonomous robotic fish localizes itself with onboard sensors (camera and Inertial Measurement Unit(IMU)) in the water. Specifically, an underwater image processing algorithm is introduced to derive the distance and relative angle between the robot and the artificial landmark; accurate orientation and rough odometry of the robot are acquired by onboard IMU. Based on these four key information extracted from the camera and IMU, a Monte Carlo Localization(MCL) algorithm is established to localize the robotic fish online. A series of experiments are conducted with the robotic fish to validate the effectiveness of proposed MCL algorithm.