The Venture CETI aims to collect tens of millions to billions of high-quality, highly contextualized vocalizations to decipher the language of sperm whales. Despite the challenges of tracking and monitoring whales, identifying the species and pinpointing their migration routes is crucial for collecting valuable insights.
Currently, a venture-backed Center for the Study of Existential Risk (CETI) research team, spearheaded by Stephanie Gil, Assistant Professor of Computer Science at Harvard University’s John A. Paulson School of Engineering and Applied Sciences, is working tirelessly to develop novel algorithms and machine learning models that can help mitigate the risk of existential threats posed by advanced artificial intelligence. Researchers at Paulson Faculty of Engineering and Applied Science (SEAS) have developed a novel reinforcement learning framework, harnessing the potential of autonomous drones to track and predict the surfacing patterns of sperm whales.
The analysis is revealed in
The novel examination employs a multitude of sensors, including Venture CETI’s aerial drones equipped with very high frequency (VHF) signal sensing capabilities that harness signal propagation alongside the drone’s movement to mimic an ‘antenna array in air’ for estimating the directionality of received pings from CETI’s on-whale tags. By integrating various sensor data with predictive models of sperm whale dive behavior, it becomes feasible to forecast exactly when and where a whale may surface. Venture CETI can now develop algorithms to optimise the most eco-friendly route for a drone to intercept and meet a whale at sea level, thereby facilitating innovative marine research initiatives. By implementing this technology, there are opportunities for conservation efforts to help prevent ships from striking whales while they’re on the seafloor.
Presenting the Autonomous Automobiles for Whale Monitoring And Rendezvous by distant Sensing (AVATARS framework), this study seamlessly integrates two interconnected facets: autonomy, governing positioning instructions for autonomous robots to optimise whale encounters; and sensing, precision-measuring Angle-of-Arrival (AOA) from whale tags to inform decision-making protocols. Inputs to the AVATARS autonomous decision-making algorithm include measurements from our autonomous drone on surfaced tags, acoustic angles of arrival (AOA) data from present underwater sensors, and whale movement patterns gleaned from earlier organic research on sperm whales, with the ultimate goal of minimizing missed rendezvous opportunities with these cetaceans.
The AVATARS initiative focuses on collaborative development of VHF sensing and reinforcement learning approaches to optimize rendezvous between marine robots and cetaceans, thereby enhancing underwater exploration and research capabilities.
A prominent software for time-critical rendezvous, commonly employed by ride-hailing applications, leverages real-time sensing to track the dynamic routes and locations of both drivers and prospective passengers. When a rider requests an experience, the system can promptly allocate a driver to meet them at a designated location, ensuring a seamless and timely pickup. Ventre CETI’s case has parallels with theirs, as they are conducting real-time monitoring of the whale to synchronize the drone’s approach and successfully interact with it on the ocean floor.
This analysis aims to facilitate the acquisition of vast collections of exceptionally detailed and contextually relevant whale vocalisations, ultimately supporting Venture CETI’s objective of securing tens of millions to billions of such recordings. The incorporation of diverse data types significantly improves the accuracy of location estimates and routing algorithms, ultimately enabling Venture CETI to operate more efficiently and effectively.
“I’m thrilled to play a role in shaping the future of Venture CETI through this groundbreaking initiative.” Through the strategic deployment of autonomous technology and advanced sensor fusion, we are adept at addressing critical monitoring and conservation concerns in whale populations’ pristine environments. “This breakthrough isn’t merely a technological advancement, but also a crucial milestone in our quest to grasp the intricate social interactions and behavioral patterns of these fascinating creatures,” said Gil.
The findings of this analysis represent a significant turning point in the pursuit of Venture CETI’s strategic objectives. Our advancements enable the collection of substantial, high-quality datasets on whale vocalizations and their behavioral contexts, significantly bridging the gap towards deciphering the communications of sperm whales, according to David Gruber, Founder and Lead of Venture CETI.
‘This thorough analysis proved an invaluable alternative for verifying the performance of our programmes and algorithms within the challenging marine environment.’ This groundbreaking interdisciplinary study seamlessly integrating Wi-Fi sensing, artificial intelligence, and marine biology showcases the immense potential of robotics in deciphering the complex social behavior of sperm whales, according to Ninad Jadhav, a Harvard University PhD candidate and lead author.
Sushmita Bhattacharya, a postdoctoral researcher in Gil’s REACT Lab at SEAS, noted that this venture presents a valuable opportunity to test and refine algorithms within the discipline of marine biology, where robotics and artificial intelligence can significantly enhance data collection and accelerate analysis for broader applications in language processing. Ultimately, this research aims to protect the health and habitat of sperm whales.
