Pharos: AI system cuts drone-related human fear by 52.72%
New multi-UAV airspace management reduces fear 52% while boosting space use 70%
Lin Sun and five co-authors from (presumably) Chinese institutions have published a paper on arXiv proposing Pharos, a collaborative multi-UAV airspace management system designed to handle the complex safety challenges of urban low-altitude economies. Pharos sits between decentralized local perception and centralized fine-grained control paradigms, coordinating safe parallel execution of drones in shared airspace while explicitly accounting for human fear—a novel consideration for drone management systems.
The system uses the MAPPO (Multi-Agent Proximal Policy Optimization) algorithm, chosen for its faster convergence and higher rewards compared to other MARL methods like HAPPO and HATRPO. Evaluation on a 3D simulation built from real urban data showed impressive results: Pharos reduced human fear by 52.72% and improved spatial entropy (a new metric for space utilization) by 70.82% compared to the Ipopt benchmark. The source code is available in an anonymized repository. This work directly addresses a key barrier to drone adoption in cities—public anxiety about safety—while simultaneously improving airspace efficiency.
- Pharos uses the MAPPO algorithm for faster convergence than HAPPO and HATRPO
- Cuts human fear by 52.72% compared to the Ipopt benchmark
- Spatial entropy (space utilization) improved by 70.82% vs Ipopt and 2.03% vs A-star
Why It Matters
Enables safer drone operations in crowded cities, critical for the low-altitude economy and public acceptance.