Semantic Mapping for Autonomous Subsea Intervention

Area of Science:

• Robotics
• Computer Vision
• Autonomous Systems

Background:

• Autonomous underwater vehicles (AUVs) require precise navigation and environmental understanding for intervention tasks.
• Mapping submerged man-made structures like pipes is challenging due to sensor limitations and dynamic environments.

Purpose of the Study:

• To develop a method for building a semantic map to aid AUVs in autonomous intervention tasks within submerged pipe structures.
• To enhance the accuracy and robustness of AUV navigation and object recognition in complex underwater environments.

Main Methods:

• Integration of feature-based simultaneous localization and mapping (SLAM) with 3D object recognition.
• Utilizing Doppler velocity log (DVL), pressure, and attitude and heading reference system (AHRS) sensors for navigation.
• Employing laser scanners for real-time 3D point cloud acquisition and Bayesian techniques for object class estimation.

Main Results:

• Generation of a consistent, drift-less map of the pipe structure.
• Accurate identification and localization of pipes, valves, elbows, and tees.
• Improved object recognition through fusion of observations and pipe connectivity information.

Conclusions:

• The proposed semantic mapping method enables autonomous intervention tasks for AUVs in submerged pipe structures.
• The system provides a precise semantic map, crucial for future high-level manipulation commands.
• This approach enhances AUV capabilities for inspection and maintenance in underwater infrastructure.