AI Drones Collect Ocean Data 106% Better with Queue-Aware RL
New graph reinforcement learning coordinates UAVs to fetch buoy data efficiently.
A team of researchers from Beijing, Surrey, and George Washington University published a paper on arXiv proposing a queue-aware graph reinforcement learning framework for UAV-ISAC-assisted maritime data collection. The system leverages high-altitude platforms (HAPs) and rotary-wing UAVs to perform integrated sensing and communication (ISAC) over drifting buoys in the ocean. Each buoy maintains a buffer of observations, and the objective is to minimize queue-weighted backlog rather than maximize instantaneous throughput. The problem is formulated as a mixed discrete-continuous optimization with constraints on sensing, communication, mobility, safety, and onboard energy.
To solve this, the team designed a structured feasible-association graph-MARL framework. A heterogeneous graph encoder produces candidate edge logits, and a masked sequential b-matching policy selects legal UAV-buoy associations while respecting load and clustering constraints. They used a MAPPO-style training procedure with independent queue-state value critics and a consistency-verification protocol. Simulation results show that the proposed policy improves cumulative queue-weighted collection utility by about 106% over a deterministic rate-driven decoder, maintains a large margin across sea-state and traffic loads, and transfers zero-shot to larger networks. This work could significantly enhance real-time ocean monitoring for climate research, shipping, and defense.
- Queue-aware graph MARL framework achieves 106% better cumulative utility than rate-driven deterministic decoders.
- System explicitly models spatially correlated sea patches, buoy dynamics, RCS/clutter-aware echo sensing, and propulsion energy limits.
- Policy transfers to larger networks without fine-tuning, maintaining performance across sea-state and traffic loads.
Why It Matters
Enables scalable, energy-efficient drone swarms to collect ocean data in real time, improving climate and maritime surveillance.