Work

Minimizing Information Age and Sensor Energy in Ocean Monitoring Networks through Hybrid Acoustic and Optical Data Harvesting by Cooperative Underwater Vehicles

Year: 2026
Type: preprint
Abstract: Ocean monitoring networks must deliver fresh measurements from battery-powered seabed sensors, yet underwater links impose a hard trade-off. Acoustic links reach kilometres but are slow and energy-hungry, whereas optical links are fast and energy-cheap but reach only metres. Autonomous underwater vehicles (AUVs) can harvest data by approaching sensors, but ocean currents perturb their motion. This paper proposes AquaFresh, a cooperative policy that decides, for every AUV and time slot, which sensor to serve and whether to use an acoustic or an optical link, minimising priority-weighted age of information (AoI) and sensor energy. AquaFresh combines Lyapunov drift-plus-penalty scores with a graph attention residual shared by all vehicles and is trained by antithetic evolution strategies on the team objective. The simulation model includes Thorp absorption, Wenz ambient noise, Rayleigh fading, co-channel acoustic interference, residual Doppler, water-type-dependent optical attenuation and a time-varying double-gyre current field. On held-out ocean realisations, AquaFresh reduced weighted AoI by 12.5% and sensor energy by 28.8% relative to a Lyapunov max-weight scheduler, and AoI by 28.0% relative to an AoI-greedy hybrid policy at equal energy. It retained the best AoI in zero-shot tests across water types, current strengths, 40 to 100 sensors and 2 to 5 vehicles, and an energy-weight sweep traced an AoI and energy front that dominates all baselines. Ablations indicate that end-to-end optimisation of the energy-aware objective drives most of the gain.
Institution: Saveetha University
Language: English
Open Access status: green
OpenAlex ID: W7236167596
Cites: 21
Cited by:
Related to: 10