Leiden University Student Repository

Spoor, Lindsay (2024)

Optimising Group Alignment and Cohesion in Autonomous Drone Swarms for Targeted Military Exploration

Master thesis | Physics (MSc)

Military technological advancements have shown that autonomous drone swarms have emerged as a powerful tool for executing complex missions with enhanced resilience and efficiency, offering significant advantages due to their collective behavior; they can operate autonomously, cover large areas, and continue missions even if individual drones fail, making them highly robust in dynamic and hazardous environments. This study uses an approach from reinforcement learning and presents a single-agent framework designed for a central-controller setup within drone swarms, and focuses on optimizing group alignment and cohesion while exploring a designated target area. Results show that smaller swarms (N = 2) achieve better cohesion and higher alignment, while increased alignment tends to reduce exploration efficiency. Performance metrics show that smaller swarms reached 69% ± 3% of optimal exploration while maintaining group alignment and cohesion, whereas larger swarms (N = 4) achieved 47%...Show moreMilitary technological advancements have shown that autonomous drone swarms have emerged as a powerful tool for executing complex missions with enhanced resilience and efficiency, offering significant advantages due to their collective behavior; they can operate autonomously, cover large areas, and continue missions even if individual drones fail, making them highly robust in dynamic and hazardous environments. This study uses an approach from reinforcement learning and presents a single-agent framework designed for a central-controller setup within drone swarms, and focuses on optimizing group alignment and cohesion while exploring a designated target area. Results show that smaller swarms (N = 2) achieve better cohesion and higher alignment, while increased alignment tends to reduce exploration efficiency. Performance metrics show that smaller swarms reached 69% ± 3% of optimal exploration while maintaining group alignment and cohesion, whereas larger swarms (N = 4) achieved 47%±11%, highlighting the trade-off between maintaining group cohesion and optimizing exploration.Show less

Swarming, Drones, Collective Behaviour, Vicsek Model, Reinforcement Learning

Author

Spoor, Lindsay

Faculty

Faculty of Science

Specialisation

Classical and Quantum Information

Supervisors

L. Giomi

ECTS Credits

36

Language

en