Inertial sensor self-calibration in a visually-aided navigation approach for a micro-AUV

Area of Science:

• Robotics
• Underwater Exploration
• Computer Vision

Background:

• Accurate underwater navigation and mapping are crucial for marine research and exploration.
• Existing platforms often face challenges with sensor data integration and real-time processing.

Purpose of the Study:

• To develop and test a novel underwater robotic platform for enhanced observation, image recording, mapping, and 3D reconstruction in shallow waters.
• To improve the accuracy of robot pose and orientation estimation through advanced sensor fusion techniques.

Main Methods:

• A small underwater robot equipped with a Micro-Electro-Mechanical Systems (MEMS)-based Inertial Measurement Unit (IMU), stereo cameras, and a pressure sensor was utilized.
• Sensor data was fused, adjusted, and corrected using a multiplicative error state Kalman filter (MESKF).
• The MESKF integrated inertial sensor biases (accelerometer and gyroscope) into the state vector for self-calibration.

Main Results:

• The MESKF effectively estimated the vehicle's pose and twist, along with inertial sensor biases.
• Self-calibration of sensor biases significantly improved the accuracy of the robot's orientation estimates.
• Experiments demonstrated satisfactory real-time performance in both controlled and real-world underwater environments.

Conclusions:

• The developed underwater robotic platform offers a robust solution for shallow water observation and 3D reconstruction.
• The integrated sensor fusion approach enhances navigation accuracy and operational stability.
• The system is capable of real-time data processing, enabling practical applications in marine research.