Implementation of monocular visual SLAM with ARCog-NET for aerial robot swarm indoor mapping.

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Operating Unmanned Aerial Vehicle (UAV) swarms in GPS-denied indoor environments presents significant localization and coordination challenges.
• Existing frameworks often lack the decentralized decision-making and adaptive capabilities required for complex, real-time operations.

Purpose of the Study:

• To present a real-time distributed framework for UAV swarms in GPS-denied indoor environments.
• To enable autonomous navigation and collaborative 3D mapping through a cognitive architecture and visual SLAM.

Main Methods:

• The framework utilizes the Aerial Robot Cognitive Network Architecture (ARCog-NET) with a multi-layered Edge-Fog-Cloud (EFC) hierarchy.
• Monocular visual Simultaneous Localization and Mapping (SLAM) is integrated for joint trajectory estimation and environment reconstruction.
• Reinforcement learning dynamically adapts navigation paths based on coverage metrics and historical data.

Main Results:

• A six-UAV swarm successfully demonstrated autonomous navigation and collaborative 3D mapping in a controlled lab environment.
• The system achieved accurate trajectory estimation and high-fidelity point cloud reconstruction.
• Key performance metrics including trajectory error, decision convergence, and knowledge reuse rate were evaluated.

Conclusions:

• The proposed method enables scalable, autonomous SLAM and planning for real-world UAV networks.
• It highlights the cognitive synergy between navigation and perception in distributed aerial robotics.
• The framework offers a robust solution for complex tasks in challenging indoor environments.