DEVELOPMENT OF AUTONOMOUS QUADCOPTER WITH VISION-BASED OBSTACLE AVOIDANCE SYSTEM

• Main Author: Rais Ismail, Lutfi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/81536
• Institution: Institut Teknologi Bandung

Unmanned aerial vehicles (UAVs) of the quadrotor type are widely used today for both civilian and military purposes. Quadrotors are popular due to their excellent maneuverability, lower manufacturing costs, and absence of risk to crew. This creates high demand in the fields of control and navigation for autonomous operation. Autonomous systems aim for quadrotors to follow desired paths and avoid obstacles. This study involves designing and building hardware for an autonomous UAV quadrotor capable of detecting and avoiding obstacles. The methodology includes two phases: quadrotor development and the design and implementation of instrumentation and control systems for object detection and avoidance. The constructed quadrotor weighs 1.663 kg, has a lift-to-weight ratio of 2.381:1, and can fly for 6.15 minutes. Testing results show the object detection system can detect objects up to 7 meters away with a mean average precision (mAP) of 89.63%. The position estimation system can estimate object positions up to 6 meters on the y-axis and 1.2 meters on the x-axis with average measurement errors of 0.146 meters on the y-axis and 0.019 meters on the x-axis. The flight path planning system can find the closest flight route to the target while avoiding obstacles. Based on implementation results, the quadrotor can detect objects, estimate object positions, determine the closest flight route, and fly to the destination by following the predetermined route while avoiding obstacles. Flight data shows the quadrotor can follow the predetermined flight route with an average mean absolute error (MAE) of 0.120 meters or 9.246% of the maximum deviation limit. Keywords: unmanned aerial vehicle, quadrotor, autonomous, object detection,