Optimal Frontier-Based Autonomous Exploration in Unconstructed Environment Using RGB-D Sensor

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Aerial robots offer advantages in search and rescue due to maneuverability.
• Autonomous exploration on Unmanned Aerial Vehicles (UAVs) is challenging due to payload and computational limits.

Purpose of the Study:

• To present an improved autonomous exploration algorithm for aerial robots in search and rescue operations.
• To address the limitations of onboard computing and payload capacity in UAVs.

Main Methods:

• Utilized an RGB-D sensor for environmental data acquisition.
• Employed OctoMap for environment representation (obstacles, free, unknown spaces).
• Applied clustering and information gain-based cost function for frontier selection, with A* path planner and safe corridor generation for path planning.

Main Results:

• The proposed algorithm demonstrated a shorter exploration path compared to baseline methods.
• The algorithm achieved significant savings in exploration time.
• Validated through experiments in diverse environments and real flight tests.

Conclusions:

• The developed autonomous exploration algorithm is effective for search and rescue tasks.
• The algorithm offers improved efficiency in terms of path length and time.
• Successful real-world flight validation confirms its practical applicability.