CoSLAM: collaborative visual SLAM in dynamic environments.

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Simultaneous Localization and Mapping (SLAM) is crucial for autonomous systems.
• Dynamic environments pose significant challenges for traditional SLAM algorithms.
• Multi-camera systems offer richer environmental perception.

Purpose of the Study:

• To develop a robust vision-based SLAM system for dynamic environments using multiple, independently moving cameras.
• To accurately map both static background elements and dynamic foreground objects.
• To improve localization and mapping accuracy and robustness.

Main Methods:

• Introduced inter-camera pose estimation and inter-camera mapping techniques.
• Maintained map point position uncertainty for enhanced robustness.
• Clustered cameras into groups based on view overlap for efficient inter-camera operations.
• Managed real-time splitting and merging of camera groups.

Main Results:

• Demonstrated robust performance in highly dynamic environments.
• Achieved more accurate results in static environments compared to existing methods.
• Successfully mapped static background points and dynamic foreground object trajectories.

Conclusions:

• The proposed multi-camera SLAM system effectively handles dynamic environments.
• Inter-camera cooperation and uncertainty management enhance system robustness and accuracy.
• The system provides a reliable solution for real-time localization and mapping.