TY  - JOUR
T1  - Development of a Stereo Camera-Based Autonomous Obstacle Avoidance Drone System
AU  - Song, Si-woon 
AU  - Ji, Chang-hun 
AU  - Kim, Min-kyu 
AU  - Lim, Gyeong-hun 
AU  - Joe, Tae-hyeon 
AU  - Moon, Sung-tae 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2025
DA  - 2025/1/1
DO  - 10.7840/kics.2025.50.11.1760
KW  - UAV (Unmanned Aerial Vehicle)
KW  - Stereo Camera
KW  - Range Finder
KW  - Obstacle Avoidance
KW  - 
KW  - Autonomous Flight
KW  - PX4-ROS2
KW  - Ground Segmentation
AB  - Recently, research on utilizing drones for reconnaissance, search, and rescue operations has been on the rise.
The existing studies primarily rely on 3D LiDAR and precise SLAM-based systems to achieve high accuracy. However, However, they face limitations due to the weight and cost of these sensors, making them unsuitable for small drones. Additionally, these systems are often designed and validated on flat terrains, restricting their performance in complex environments. This paper proposes a cost-effective autonomous obstacle avoidance drone system that integrates stereo cameras and range sensors. As a practical alternative to expensive 3D LiDAR, the proposed system reduces sensor configuration and system complexity while enabling stable flight in challenging terrains such as slopes and forests. Additionally, by applying an obstacle inflation technique, the system achieves obstacle avoidance performance comparable to that of 3D LiDAR-based systems. The key contributions of this study are as follows. First, the system achieves 3D LiDAR-level obstacle detection and terrain tracking performance using only stereo cameras and range sensors. Second, unlike prior studies focused on flat terrains, the proposed algorithms are validated in complex environments, including steep slopes and forested areas, demonstrating robustness. Third, the system is designed based on PX4 and ROS2, ensuring high scalability and reusability by employing a consistent system across both simulations and real-world flight environments. This paper begins by analyzing the limitations of existing research, followed by a detailed explanation of the proposed obstacle detection and autonomous avoidance system's algorithms and architecture.
The system’s performance is validated through both simulation and real-world flight experiments.