Over-horizon Awareness of Seafloor Imaging Surveys from Long Range AUVs

Long-range Autonomous Underwater Vehicles (AUV) enable persistent seabed surveys over spatial and temporal scales that are impractical for conventional ship-based operations. However, the large volume of optical imagery collected during these missions prevents operators from obtaining timely situational awareness when communication is limited to low-bandwidth satellite links. This paper proposes an
over-horizon awareness framework that uses onboard artificial intelligence to transform high-resolution seabed imagery into compact survey summaries suitable for satellite transmission. The framework enables operators to assess survey progress and seabed characteristics without recovering the vehicle or transmitting complete image datasets. Offshore validation over 41 AUV dives demonstrated the feasibility of semantic over-horizon awareness by successfully transmitting semantic survey summaries generated from 37,285 seabed images with an average latency of 8 minutes. This work provides a foundation for persistent ship-free seabed monitoring and future adaptive mission execution based on onboard interpretation of the surrounding environment.