Multi-Agent RL for Drone Fleet Deconfliction in Urban Airspace
A new study from arXiv (2605.01041) explores tactical deconfliction for heterogeneous fleets of small unmanned aerial systems (sUAS) in dense urban airspace. Using multi-agent reinforcement learning, researchers simulated package delivery missions over Dallas, Texas, with multiple companies operating fleets of homogeneous aircraft. The study asks whether policies can converge to a conflict-free equilibrium and whether weaker-equipped fleets face discrimination. An attention-enhanced Proximal Policy Optimization algorithm was employed. The research addresses the complexity of managing diverse aircraft configurations in future urban air mobility.
Key facts
- Study uses multi-agent reinforcement learning for sUAS deconfliction
- Simulated over Dallas, Texas, USA
- Heterogeneous fleets with homogeneous aircraft per fleet
- Attention-enhanced Proximal Policy Optimization algorithm used
- Addresses policy convergence and potential discrimination
- Published on arXiv with ID 2605.01041
- Focuses on package delivery missions
- Considers future dense urban airspace scenarios
Entities
Institutions
- arXiv
Locations
- Dallas
- Texas
- USA