Multi-agent reinforcement learning framework enables heterogeneous drone fleets to maintain safe separation in shared airspace

Researchers developed an attention-enhanced Proximal Policy Optimization-based Advantage Actor-Critic (PPOA2C) framework to resolve conflicts between homogeneous aircraft operating within heterogeneous fleets, with each fleet independently training its own policy while preserving privacy.

In experiments over Dallas, Texas, two fleets with distinct, shared PPOA2C policies reached an equilibrium to maintain safe separation; PPOA2C policies outperformed rule-based baselines in conflict resolution and exhibited safer interaction with rule-based policies.

Policy-configuration evaluations revealed that equilibria between similar policy types tend to favor fleets with stronger configurations, and equilibrium can favor one heterogeneous policy over another even under similar configurations but different policy types, indicating the need for fairness-aware conflict management in heterogeneous unmanned aerial system operations.