Odometry-Vision-Based Ground Vehicle Motion Estimation With SE(2)-Constrained SE(3) Poses.

Area of Science:

• Robotics
• Computer Vision
• Motion Estimation

Background:

• Keyframe-based batch optimization is standard for mobile vehicle localization and mapping.
• Vision-based methods typically use SE(3) for keyframe poses, while range-scanner methods use SE(2).
• Ground vehicles have specific motion constraints not fully captured by existing methods.

Purpose of the Study:

• To propose a new SE(2)-constrained SE(3) pose parameterization for ground vehicle motion estimation.
• To develop a robust odometry-vision-based motion estimation system using this parameterization.
• To validate the system's performance and accuracy in real-world industrial environments.

Main Methods:

• Developed a novel SE(2)-constrained SE(3) parameterization for ground vehicle poses.
• Integrated this parameterization into a graph optimization framework for motion estimation.
• Formulated specialized edges within the batch optimization process to enforce SE(2) constraints on SE(3) poses.

Main Results:

• The proposed SE(2)-constrained SE(3) parameterization effectively models ground vehicle motion.
• The developed odometry-vision system demonstrated superior accuracy in real-world experiments.
• The system's modular graph optimization structure allows for flexibility and adaptation.

Conclusions:

• The novel SE(2)-constrained SE(3) parameterization offers significant improvements for ground vehicle motion estimation.
• This approach enhances the accuracy of visual-odometry based localization and mapping systems.
• The validated system is suitable for practical applications in industrial indoor environments.