Self-localization Research Of Underwater Robot

Underwater robot localization refers to the real-time acquisition of the robot poserelative to its aquatic working environment. It is the base for the robot to realizeautonomous navigation and accomplish complex practical tasks. This paper proposesa self-localization algorithm for a kind of autonomous underwater robot-thebiomimetic robotic fish. The proposed localization algorithm utilizes the datainformation from onboard camera and Inertial Measurement Unit (IMU), andintegrates the Monte Carlo Localization and Inertial Navigation System to localize theunderwater robot autonomously.First, this paper introduces the Linux-based embedded system design for therobotic fish. The hardware system of the robotic fish mainly includes the mainprocessor, sensing unit, monitoring unit, locomotion control unit and powermanagement unit. Software system of the robotic fish is built on the base of Linuxoperating system, which uses multi-threading technology, has the high stability, ittakes up less resources.Second, this paper proposes an underwater image processing algorithm. Due toinsufficient light environment, non-uniform illumination, refraction and reflectionunder the water, the underwater image quality is very poor, which makes theunderwater image processing more difficult. In order to eliminate the underwaternoise, automatic color balance (ACE) filtering algorithm is first used to preprocess theobtained image, and then the image is converted to HSV color space. Next, by colorthreshold segmentation, the artificial landmarks are identified. In the end, the distancebetween the robot and the landmark and the relative angle between the robot and thelandmark are acquired to serve as the criteria of the localization algorithm. Moreover,due to the calculated distance and angle contain the measurement errors, Kalman filteris adopted to smooth the data.Next, this paper realize the pose evaluation for the autonomous robotic fish in athree-dimensional water space. According to the distance and angle from theunderwater image processing, the depth information from the pressure sensor, and therobot orientation and odometry from the IMU, the proposed localization algorithmwhich fuses the Monte Carlo Localization and Inertial Navigation System, is able tolocalize the robot in the three-dimensional underwater environment.Finally, the underwater image processing and autonomous localizationexperiments are conducted with the robotic fish. The extensively experimental resultsverifies the effectiveness, accuracy and robustness of the proposed localizationalgorithm for the underwater robots.