Collaborative Research: Open Source AI Acoustic Buoys and Drifters

最近的进展表明，使用配备被动声传感器的移动的平台，在绘制大空间尺度上的海洋生物分布图方面大有希望。海洋被动声学（收听工具）可用于识别和提供发声动物的位置，发声动物包括捕虾、许多鱼类（包括具有商业重要性的鱼类）、海洋哺乳动物和船只。然而，这些移动的平台具有若干限制，阻止它们用于收集非常大规模的天气数据，包括机械噪声、需要支持船只以及通常高成本。这项工作将力求开发和测试比目前的移动的平台便宜得多的创新性声学浮标和漂移器，以便能够进行全球现场生物观测。 我们的目标是广泛使用声学浮标和漂流器，以便小型研究小组能够进行海洋规模的研究。低成本的声学浮标和漂流器将使部署的传感器数量达到使用单个自主车辆可行的100倍。漂移的主要挑战是，感兴趣的数据必须通过低带宽的卫星链路发送。这种带宽限制需要处理机载数据并发回关键信息以回答特定的生物学问题。这项研究将在漂移器中集成先进的卷积神经网络AI处理引擎，并开发工具，使研究人员能够轻松训练自己的检测和分类网络，并将其加载到漂移器上。漂流者还将存储高分辨率的原始数据，如果漂流者被回收，则可以访问这些数据。该开源设计将使其他研究人员能够定义自己的信号处理方案，以检测感兴趣的信号。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Recent advances using mobile platforms equipped with passive acoustic sensors show great promise to map the distribution of marine organisms over large spatial scales. Ocean passive acoustics (listening tools) can be used to identify and provide the locations of sound producing animals including snapping shrimp, many fishes including those that are commercially important, marine mammals, and boats However, these mobile platforms have several limitations preventing their use to collect synoptic data on a very large scale including mechanical noise, requiring support vessels, and generally high costs. This work will seek to develop and test innovative acoustic buoys and drifters that are substantially less expensive than current mobile platforms to enable global in situ biological observations. Our goal is to make acoustic buoys and drifters widely available so that it is possible for small research groups to perform ocean-scale studies.A low-cost acoustic buoy and drifter will enable deployments of 100x as many sensors as would be feasible using a single autonomous vehicle. The main challenge with drifters is that the data of interest must be sent over low-bandwidth satellite links. This bandwidth limitation necessitates processing data onboard and sending back the key bits of information to answer specific biological questions. This research will integrate an advanced convolutional neural network AI processing engine in the drifters and develop tools to make it easy for researchers to train their own detection and classification networks and load them onto a drifter. The drifters will also store high resolution raw data that can be accessed if the drifter is recovered. The open source design will enable other researchers to define their own signal processing schemes to detect signals of interest.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.