Perpustakaan Digital ITB

With the growing applications of drone swarms across industries, demand for innovative and intuitive control systems has also increased as traditional control mechanisms often face challenges in scalability and adaptability, particularly in dynamic environments. This research addresses these challenges by developing a body gesture-based control system for drone swarms, inspired by standardized air marshalling signals. The primary objective of this study is to interpret human body gestures into drone swarm commands through a high-level control pipeline. The developed system integrates a real-time gesture recognition module utilizing Mediapipe with a modular swarm control algorithm. The system recognizes thirteen distinct gestures as swarm control commands, from takeoff and landing, return to base, linear movements, and formation changes. The drones are modeled as point mass and controlled via a PID-based velocity controller. To enhance fidelity, a control delay is simulated based on empirical data. Evaluation of the system includes performance metrics such as gesture recognition accuracy, swarm control latency, and scalability. Experiments conducted with varying swarm sizes revealed an average gesture recognition accuracy of 94% with a standard deviation of 6% and latency of 15ms, while the swarm control latency showed an average of under 100ms for up to 16 drones. This research contributes to human-drone interaction by demonstrating an intuitive control system based on body gestures to serve as high-level commands for drone swarm applications. Future works may extend this framework to support 3-dimensional visualizations or apply it to physical drone for real-world deployment.