RGB-D SLAM in Dynamic Environments Using Point Correlations.

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Simultaneous Localization and Mapping (SLAM) is crucial for autonomous systems.
• Dynamic objects in environments pose significant challenges to traditional SLAM algorithms.
• Existing methods often struggle with accuracy and real-time performance in cluttered or moving scenes.

Purpose of the Study:

• To propose a novel SLAM method robust to dynamic objects.
• To enhance the accuracy and reliability of SLAM in challenging environments.
• To achieve real-time performance for practical applications.

Main Methods:

• Utilizes map point correlation to differentiate static and dynamic elements.
• Employs Delaunay triangulation to build a sparse graph representing point relationships.
• Applies point-correlation optimization to remove edges between uncorrelated points, isolating static scene points.
• Performs motion estimation exclusively on identified static map points.

Main Results:

• Successfully separates static scene points from moving objects.
• Demonstrates robust and accurate performance in both slightly and highly dynamic environments.
• Achieves competitive accuracy and good real-time performance compared to state-of-the-art methods.
• Validated on public RGB-D benchmarks and diverse challenging environments.

Conclusions:

• The proposed SLAM method effectively handles dynamic environments by filtering moving objects.
• It offers a reliable solution for accurate localization and mapping in real-world scenarios.
• The method provides a strong foundation for future advancements in autonomous navigation systems.