Validation of Underwater Sensor Package Using Feature Based SLAM

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Robotic vehicles require accurate self-localization and environmental mapping for navigation in unknown areas.
• Acoustic range sensors face performance limitations in enclosed underwater environments due to reflections and are costly for low-cost vehicles.
• Existing solutions are inadequate for low-cost, autonomous underwater navigation and obstacle avoidance.

Purpose of the Study:

• To propose and validate a novel, cost-effective sensor package for underwater unmanned vehicles.
• To enable robust simultaneous localization and mapping (SLAM) for robotic navigation in challenging aquatic environments.
• To address the limitations of traditional underwater sensing technologies.

Main Methods:

• Development of a sensor package featuring a downward-facing camera for visual odometry and a custom 2D vision-based rangefinder.
• Integration and testing of the sensor package within a SLAM framework.
• Experimental validation using an unmanned ground vehicle and two feature-based SLAM algorithms: Extended Kalman Filter (EKF) and Rao-Blackwellized Particle Filter (RBPF).

Main Results:

• The proposed sensor package demonstrates feasibility for use in SLAM.
• Feature-based visual odometry and 2D rangefinding provide essential data for robotic localization and mapping.
• Experimental results validate the sensor package's performance within established SLAM algorithms.

Conclusions:

• The developed sensor package offers a viable, low-cost solution for autonomous underwater navigation.
• This approach overcomes the limitations of acoustic sensors in enclosed underwater spaces.
• The sensor package enhances the capabilities of underwater unmanned vehicles for exploration and task completion.