Project CETI, or Cetacean Translation Initiative, aims to collect vast amounts of high-quality vocalizations from sperm whales to better understand their communication. However, finding the whales and predicting their surfacing locations to gather data has been a challenge for researchers. A team led by Stephanie Gil, Assistant Professor of Computer Science at Harvard’s SEAS, has developed a new reinforcement learning framework using autonomous drones to locate sperm whales and predict their surfacing spots. This research, published in Science Robotics, utilizes various sensing devices, such as VHF signal aerial drones, to estimate the directionality of received pings from CETI’s on-whale tags and predict when and where a whale may surface.
The AVATARS framework, or Autonomous Vehicles for whAle Tracking And Rendezvous by remote Sensing, jointly develops autonomy and sensing components to maximize visual whale encounters and inform decision-making processes. By using measurements from autonomous drones, acoustic AOA from underwater sensors, and whale motion models, the AVATARS algorithm aims to minimize missed rendezvous opportunities with whales. This research is the first to co-develop VHF sensing and reinforcement learning decision-making for maximizing rendezvous of robots and whales at sea, similar to the time-critical rendezvous systems used in rideshare apps to efficiently match drivers with riders.
This study is a significant advancement for Project CETI in its mission to obtain large-scale, high-quality whale vocalization data. By integrating diverse data types, such as location estimates and routing algorithms, the project can improve its efficiency in achieving its goal of collecting millions to billions of whale vocalizations. The use of autonomous systems and advanced sensor integration helps to address key challenges in tracking and studying whales in their natural habitats. Researchers hope that by better understanding whale communication and behavior, they can ultimately translate and interpret what sperm whales are saying.
With the addition of various types of data and the development of new algorithms and decision-making processes, researchers are optimistic about the future of Project CETI and its impact on marine biology and language processing. By combining wireless sensing, artificial intelligence, and marine biology, this interdisciplinary work shows how robotics can play a crucial role in decoding the social behavior of sperm whales. This project provides a unique opportunity for testing algorithms in the field and utilizing robotics and AI to enhance data collection and expedite research in marine biology, language processing, and habitat preservation.
Researchers involved in the project express excitement about the breakthroughs achieved through this research, highlighting the technological advancements and critical steps taken to understand the complex communications and behaviors of sperm whales. By leveraging autonomous systems and advanced sensor integration, they aim to contribute to the broader science of marine biology and language processing while protecting the health and habitat of sperm whales. The continued efforts of Project CETI and collaborations between researchers in various fields demonstrate the potential for robotics and AI to revolutionize the study of marine life and enhance conservation and research efforts in the future.