Düşmanca kentsel ortamlarda taktik görevler için uçbirim zekâlı merkeziyetsiz iha sürüleri

This thesis proposes a comprehensive framework for autonomous UAV swarms empowered by decentralized edge intelligence to operate effectively in adversarial urban environments. The study addresses key challenges such as GPS-denied navigation, dynamic obstacle avoidance, and cyber threats. The system adopts a decentralized edge-computing architecture that emulates onboard intelligence for each UAV. Various edge-level algorithms were implemented to enhance autonomy and coordination, including A* for path planning, ORCA for collision avoidance, Boids-based flocking for formation control, Kalman filtering for sensor data fusion, and Reinforcement Learning for adaptive decision-making. High-fidelity simulations confirmed a 95.2% mission success rate, a 3% collision probability, and latency below 25 ms, demonstrating robust and reliable performance in complex virtual urban scenarios. Keywords: Edge intelligence, decentralized control, UAV swarms, anti-jamming, urban navigation SLAM, convolutional neural networks, hybrid path planning, federated learning, cyber security