Resilient Leader-Follower Drone Swarm Simulation with Failure Recovery and Dynamic Re-election

บทคัดย่อ

Swarm intelligence is a bio-inspired paradigm in which simple agents collectively exhibit complex, coordinated behaviors through local interactions and decentralized control. This approach has been widely applied to robotics, distributed systems, and autonomous aerial networks due to its inherent scalability, robustness, and adaptability. In this work, we apply swarm intelligence principles to the problem of drone communication and coordination under conditions of random failures. We propose a leader-follower drone swarm model in which each drone moves according to local rules, including alignment with nearby drones and attraction toward a dynamically designated leader. The leader guides overall swarm motion, but may fail randomly, triggering an automatic re-election process to select a new active drone as the leader. Drones can also recover from failure probabilistically, enabling the swarm to maintain continuous operation. To improve visualization and analyze swarm connectivity, all active drones are forced to remain visually linked to the leader, while local neighbor links illustrate short-range interactions. Experimental simulations demonstrate that the proposed approach preserves swarm cohesion, adapts to leader failure, and recovers efficiently after multiple drone failures. Quantitative metrics, including the fraction of drones connected to the leader, the number of active drones over time, and swarm cohesion, are used to evaluate performance. The results highlight the effectiveness of leader-follower dynamics and recovery mechanisms in maintaining functional swarm behavior, providing insights for resilient autonomous aerial systems, cooperative robotics, and distributed networked agents.