Automatic Shark Detection via Underwater Acoustic Sensing

Shark attacks are a rare but ever-present danger for swimmers and surfers in some regions of the world, and the threat of sharks is embedded in popular culture. Traditionally, shark attack mitigation involved the culling of massive numbers of sharks, which has significant environmental and ethical downsides. More recent systems for mitigating the risk of shark attacks involve the manual or automated detection of sharks close to the shore, alerting water users to the potential danger when it occurs and evacuating the water if the shark gets too close. In this work, we present the design of a Shark Warning Acoustic Network (SWAN) that exploits underwater acoustic sensing and communication to automate the spotting, providing a highly accurate and relatively low-cost alternative to visual spotting. We analyze the performance of the SWAN in terms of communication performance and accuracy in alerting water users to dangerous situations, and compare different medium access schemes to identify the most effective network design.