Edge-to-Cloud Swarm Coordination for autonomous urban air mobility routing under real-time policy constraints

Описание и применение

Архитектура edge‑to‑cloud распределяет управление дрон‑свером по трем уровням, обеспечивая быстрое соблюдение динамических правил полёта и устраняя проблемы централизованного планирования.

Задача

Coordinating thousands of autonomous drones in urban airspace while complying with dynamic, location‑specific policy constraints that can change in milliseconds.

Результат

The edge‑to‑cloud system reduced coordination latency to sub‑second for regional clusters and eliminated cascading failures, enabling safe routing of large drone swarms under rapidly changing airspace policies.

Рабочий процесс

• 1. Deploy a three‑tier architecture: cloud, edge, vehicle.
• 2. Implement Raft consensus on edge nodes to share local constraints.
• 3. Use a policy‑constrained PPO model on each drone to map observations to actions while filtering against real‑time policy constraints.
• 4. Continuously update global policy plans in the cloud and propagate to edge clusters.
• 5. Validate coordination in a simulated swarm of 10,000 drones on a home lab with RTX 4090s and Raspberry Pi 5s.

Доказательства

I watched in my simulations as a single network partition caused 3,000 simulated drones to enter emergency landing procedures simultaneously over a densely populated area.

Источник

Through studying distributed systems theory and applying it to aerial robotics, I developed a hierarchical coordination framework that distributes intelligence across three tiers.

Источник