A Lightweight, Centralized, Collaborative, Truncated Signed Distance Function-Based Dense Simultaneous Localization...

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Simultaneous Localization And Mapping (SLAM) is crucial for mobile robot autonomy in unknown environments.
• Resource constraints on mobile robots limit their capability for long-term, large-scale exploration tasks.
• Existing SLAM systems often struggle with the computational and memory demands of dense mapping.

Purpose of the Study:

• To propose an efficient multi-robot dense SLAM system to address the limitations of individual mobile robots.
• To reduce the computational and memory burdens on individual robots through a centralized structure.
• To enable real-time dense mapping and accurate localization for enhanced autonomous exploration.

Main Methods:

• A centralized multi-robot system architecture to distribute computational load.
• A lightweight and accurate dense mapping method for real-time agent-level mapping.
• A novel loop closure detection method using combined visual and dense geometric information.
• Multi-robot pose graph optimization integrating dense geometric and trajectory data.

Main Results:

• Demonstrated system efficiency and accuracy through experiments on pre-recorded datasets.
• Achieved a real-world dense mapping update rate of approximately 14 frames per second.
• Reported low onboard mapping RAM usage (~3.4%) and bandwidth consumption (~302 KB/s) on a Jetson Xavier NX.

Conclusions:

• The proposed efficient multi-robot dense SLAM system effectively alleviates computational and memory burdens on mobile robots.
• The system enables real-time dense mapping and accurate multi-robot localization for large-scale autonomous exploration.
• The approach shows significant promise for practical deployment in hazardous or inaccessible environments.