Autonomous 3D Exploration of Large Structures Using an UAV Equipped with a 2D LIDAR

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Exploring large, unstructured industrial environments poses significant challenges for autonomous systems.
• Existing methods often struggle with real-time mapping and efficient coverage in complex spaces.

Purpose of the Study:

• To develop and evaluate an efficient system for unmanned aerial vehicle (UAV)-based exploration of large, unknown industrial environments.
• To create a 3D map of complex scenarios using a novel approach combining existing techniques and a new maneuver.

Main Methods:

• The system integrates frontier-based exploration, the Lazy Theta* path planner, and a flyby sampling maneuver.
• A low-cost 2D laser is utilized for 3D structure measurement.
• World representation relies on the multi-resolution octomap for all algorithms.
• Hardware-in-the-Loop (HitL) simulation validates real-time, on-board performance.

Main Results:

• The proposed approach achieved exploration of 93% of the search space within 30 minutes.
• The system generated non-repetitive paths that adapt to occupied space.
• Performance was superior to reference heuristics in explored space metrics within the simulation environment.

Conclusions:

• The developed UAV system offers an effective solution for autonomous exploration of complex industrial environments.
• The multi-resolution octomap-based approach enables efficient real-time 3D mapping and path planning.
• The system demonstrates robust performance in diverse indoor locations, irregular structures, and with suspended obstacles.