An Improved Localization Method for the Transition between Autonomous Underwater Vehicle Homing and Docking

Area of Science:

• Robotics
• Ocean Engineering
• Computer Vision

Background:

• Autonomous underwater vehicles (AUVs) require reliable navigation for docking, crucial for extending operational endurance.
• Existing methods like ultra-short baseline (USBL) and optical navigation face challenges with outliers, slow updates, and ambient light variations, impacting docking success rates.

Purpose of the Study:

• To enhance the accuracy and reliability of AUV localization during the homing and docking transition phase.
• To address limitations of current navigation systems, particularly under dynamic motion and varying light conditions.

Main Methods:

• Development of an improved underwater simultaneous localization and mapping algorithm based on ORB features (IU-ORBSLAM).
• Implementation of a multi-sensor information fusion approach for robust localization.
• Application of nonlinear optimization to refine monocular visual odometry scale and AUV pose estimation.

Main Results:

• Localization tests demonstrated significant improvements in both accuracy and update rate compared to existing methods.
• The proposed IU-ORBSLAM algorithm proved effective under motion mutation and light variation conditions.
• Five docking missions were executed, achieving a 100% successful docking rate.

Conclusions:

• The developed IU-ORBSLAM method offers a feasible and effective solution for improving AUV localization accuracy.
• The multi-sensor fusion approach enhances navigation reliability, crucial for successful autonomous underwater docking operations.