Solution to the SLAM problem in low dynamic environments using a pose graph and an RGB-D sensor.

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Simultaneous Localization and Mapping (SLAM) is crucial for robot navigation.
• Low dynamic environments, where object positions change slowly, pose challenges for conventional SLAM algorithms.
• False loop closures due to repositioned objects lead to inaccurate robot localization and mapping.

Purpose of the Study:

• To develop a novel SLAM method robust to low dynamic environments.
• To address the issue of false loop closures caused by slowly changing objects.
• To improve the accuracy of robot localization and mapping in such environments.

Main Methods:

• Utilized a pose graph structure combined with an RGB-D sensor.
• Grouped graph nodes representing robot poses based on noise covariances.
• Pruned false constraints using an error metric on grouped nodes.
• Reoptimized the pose graph after removing erroneous information.

Main Results:

• Successfully pruned falsely grouped constraints in the pose graph.
• Eliminated false information, leading to corrected localization and mapping.
• Demonstrated improved performance through real-world experiments with a mobile robot.

Conclusions:

• The proposed SLAM method effectively handles low dynamic environments.
• The novel approach enhances robot navigation accuracy by mitigating false loop closures.
• Validated effectiveness through practical mobile robot experiments.