Perpustakaan Digital ITB

This research explores the use of the Reciprocal Velocity Obstacle (RVO) method as a decentralized collision avoidance strategy for swarm drones during dynamic formation changes. Quadrotor UAVs were modeled in a two-dimensional kinematic framework and tested in six scenarios, including structured formation transitions and randomized initial positions. Simulations without avoidance showed frequent collisions, particularly at the formation center. With RVO implemented, all scenarios achieved safe formation changes with zero collisions, confirmed by minimum drone separations consistently above the threshold. Path deviation analysis revealed that smaller swarms and simpler transitions produced minimal detours, while denser and more complex transformations led to larger deviations due to congestion. Randomized scenarios further demonstrated the robustness and consistency of RVO across repeated trials. These results validate RVO as an effective and reliable framework for swarm drone collision avoidance, while highlighting the need for future studies on scalability, three-dimensional modeling, and real-world sensor integration.