Physics-informed Data-driven Communication Performance Prediction for Underwater Vehicles

Underwater vehicles usually rely on acoustics for communication and navigation. Reliable communication and accurate navigation require the vehicle to plan a path through areas with good acoustic coverage from communication gateways and beacons. Planning such a path can be challenging in areas with complex acoustic propagation, especially when the signal strength does not monotonically reduce as a function of distance from a transmitter. When the environmental parameters are not fully known, traditional acoustic propagation models are unable to provide accurate predictions. We develop an online physics-informed data-driven method to predict acoustic signal quality in a region ahead of the underwater vehicle to inform the vehicle's path-planning algorithm.