| Hearken to this text |
Sperm whales typically spend about 10 minutes per hour at depths near the ocean floor, making it a significant challenge for researchers to study these creatures effectively. | Supply: Amanda Cotton/Challenge CETI
Off the icy shores of New England, researchers with the Cetacean Translation Initiative’s Challenge CETI patiently scan the frigid waters for a glimpse of a rare sperm whale breaching the surface. During the brief periods when whales surface, scientists aim to gather as much information as possible before the animals plunge back into the depths for extended periods.
Given the extensive global distribution of this marine mammal species, tracking and studying them proves remarkably challenging. Can we harness the power of artificial intelligence and robotics to decipher the complex vocalizations of sperm whales? Recently, an analysis was conducted on how a system tracks and follows sperm whales across vast expanses of the open ocean.
“The ocean and its pristine habitat are where whales thrive, but building infrastructure in this vast area is challenging because we don’t have a variety of existing infrastructure,” said Stephanie Gil, an assistant professor of Computer Science at Harvard John A. Paulson School. The Paulson College of Engineering and Applied Sciences, along with its dean serving as a trusted advisor on the mission.
The mission unites top scientists globally from disciplines including biology, linguistics, and robotics, among others. David Gruber, the founder of Challenge CETI, considers it one of the largest and most comprehensive multidisciplinary research programs currently underway.
As of March 2020, Challenge CETI was launched, and today, it comprises more than 50 experts from eight diverse fields, with a collective breadth of expertise that spans multiple disciplines. I envision our network comprising more than 15 institutions, positioning us among the most comprehensive and ambitious interdisciplinary science endeavors in history. Collaboration across diverse disciplines yields a deeply fulfilling experience.
Challenge CETI shares newest analysis
Researchers have developed a novel reinforcement learning framework utilising autonomous drones to locate sperm whales and forecast their surfacing points accurately. Researchers have successfully demonstrated that it is feasible to forecast when and where a whale will surface using various sensor data and predictive models of sperm whale dive patterns, as reported in the paper printed in
The novel examination focused on diverse sensors, including the Challenge CETI aerial unit featuring exceptionally high-frequency (VHF) signal sensing capabilities that utilize signal processing in conjunction with the drone’s movement to mimic an “antenna array in the air” for estimating the trajectory of pings from CETI’s on-whale tags.
There are two key advantages associated with VHF signals. One key aspect is that these devices are designed with low power consumption, enabling them to operate for an extended period – indeed, several months or even years – without compromising their performance. Without needing to replace or recharge the power source, said Dr. Ninad Jadhav, lead researcher and robotics expert. pupil at Harvard College.
The second key consideration is the nature of the data transmitted by these tags, specifically the high-frequency indicators contained within the VHF signals. “They are frequently identified at remarkably extended distances.”
“That’s a significant advantage because we never know when whales will surface or where they’ll surface, but when they’ve been tagged previously, you can gain valuable insights, for instance, simple information such as the whale’s migration route.” “We can deploy an algorithm on the robot to identify the locations where whales are present on the seafloor, which offers a significant benefit in terms of discovering their habitats.”
What are the pressing obstacles facing sperm whale researchers in their pursuit of knowledge?
From left, proper: Stephanie Gil, Sushmita Bhattacharya, and Ninad Jadhav. | Supply: Stu Rosner
“Sperm whales spend approximately ten minutes per hour at depths near the seafloor,” said Gil. “Beyond that, scientists are observing the whales at great depths in the ocean, making it challenging to pinpoint their precise actions.” That makes them significantly challenging to detect and study for both our team and the scientific community.
We all follow predictable routines in our daily lives. While observing whales on a specific day, their behaviors may not consistently conform to established patterns despite leveraging extensive data to inform those patterns. It’s extremely challenging to accurately forecast the development of something once it has begun.
While conducting research at sea for extended periods, it’s reasonable to ponder whether the cumulative impact of multiple whale encounters might compromise our environmental responsibility? So, we aim to amplify our impact, he directed.
Once Challenge CETI researchers are able to track and locate the whales, they must swiftly collect as much data as possible during the brief periods when sperm whales remain near the seafloor.
Sushmita Bhattacharya, a Ph.D. in computer science and robotics, described underwater data collection as “somewhat challenging”, noting her role as a co-author on the relevant research paper. pupil at Harvard College. When compared to collecting data underwater, a more straightforward approach is to gather data once the objects are already on the floor. “We’ll utilize drones and shallow hydrophones to collect as much data as possible.”
Growing the AVATARS framework
The core of our analysis lies in the Autonomous Autos for Whale Monitoring and Rendezvous by Distant Sensing (AVATARS) framework. What drives the convergence of underwater explorers?
“We designed an avatar capable of mimicking the behaviors of sperm whales,” Bhattacharya explained, referring to the AVATARS project. “We make decisions primarily driven by the current data we gather from our limited information pool.”
The capacity to forecast the timing and location of whale sightings enabled scientists to craft efficient algorithms for guiding drones to optimally intercept or meet whales at sea, thereby minimizing environmental impact. Researchers faced significant challenges in designing these algorithms, they noted.
“In all likelihood, the most challenging aspect is the uncertainty surrounding this setback.” Without certainty regarding their underwater positions, we’re left uncertain, since GPS tracking is impossible once they descend into the depths. “To effectively locate the targets, it’s essential to explore alternative approaches, such as leveraging their acoustic signatures and angle of arrival data, which collectively provide a robust understanding of their position.”
“In essence, these algorithms serve as sophisticated routing tools.” “So, you’re aiming to coordinate a fleet of robots to converge at a precise location on the planet by a specific deadline, which is crucial for them to arrive punctually,” she explained, “This concept is akin to ride-hailing services.”
Before deploying the algorithms in real-world settings with actual whales, the team initially tested and refined them in a controlled environment featuring simulated whale-like entities designed by the researchers to mimic whale behavior.
Bhattacharya remembered that they simulated a whale’s movement using an engineered prototype. “We primarily utilized a velocity boat equipped with a powerful yet noisy engine.” By harnessing the distinctive sound of an engine, we successfully replicated the haunting calls of whales, further enhancing our simulation by accurately mimicking their majestic movements. After which we utilised that as our primary flooring take another glance.
What lies beyond our reality?
Researchers from Challenge CETI have developed an innovative drone that can autonomously fly to deploy a whale tag. | Supply: Challenge CETI
“On a daily basis, deploying autonomous robots from a ship in the midst of the ocean posed a problem once we ventured out on our boat, marking my first and Sushmita’s inaugural experience of gathering information.”
“One of the primary obstacles encountered in developing this system was the persistent interference from within the sensor itself.” “In contrast to conducting experiments in a controlled laboratory environment, where variables are meticulously managed, field-based tests often encounter numerous extraneous factors that can impact data accuracy and sensor readings.”
The main challenge we faced was actually launching the drone from the controller board, as expertly recounted by renowned innovator Jadhav. As I recall, it was probably the first or second day of our second expedition, which took place last November, and I had my drone ready at the time. It had the payload. It was waterproof”
With prior experiments conducted locally in the Boston region, I had a reliable estimate of the flight duration for the drone carrying its payload. Following our excursion on the boat to conduct initial tests, the drone launched into action,” he explained. As the device began to collect data, a sudden gust of wind struck, momentarily disrupting its performance. The drone suffered a loss of situational awareness and inadvertently plunged into the water.
The team also had to anticipate and adapt to whale behavior during fieldwork, requiring a deep understanding of marine life patterns.
Researchers initially developed an algorithm tailored to analyze data from a solitary whale; instead, they discovered a group of four whales interacting and socialising. “They have been consistently diving and then surfacing simultaneously.” So, deciphering acoustic signals from individual whales proved arduous, ultimately exhausting our team in terms of computational complexity to identify the correct sender and monitor their patterns.
Crew members strive to gather data without disrupting the natural habitat of local wildlife.
While collaborating with sperm whales and other marine life, Challenge CETI strives to minimize disturbance during data collection by meticulously respecting the whales’ natural habitat.
“When discussing our core values, I emphasize that while we strive for success, our top priority remains ensuring that our failures do not harm the whales.” While we’ve made significant progress in understanding and interacting with wildlife, it’s crucial that we maintain a heightened sense of awareness and respect for the natural boundaries of these animals to ensure their well-being and our own safety. To optimize our mission’s success, we’re exploring the concept of a rendezvous radius. What’s our ultimate goal?
“As Gruber pointed out, minimizing invasiveness and invisibility are crucial aspects of the Challenge CETI.” How do we obtain this information without initiating direct contact with the whale?
The team primarily employs drones that minimize disruptions to marine ecosystems, utilizing these tools to attach to whales and collect valuable insights. The CETI group meticulously collects the detached tags, extracting valuable data from the information they contain, once they have fallen off the whales.
“While some may view robotics and autonomy with trepidation, Dr. Gil emphasized the importance of highlighting how these technologies can expand human reach, fostering a deeper understanding of our world.”
The quest to decipher the secrets of cetacean language? Can we truly crack the code that governs the complex vocalizations and acoustic patterns employed by these magnificent creatures?
The latest findings represent a significant milestone in Challenge CETI’s ongoing effort to crack the code of sperm whale communications. During this brief window, the team intends to accelerate data collection, a crucial step in achieving its long-term objectives.
“As soon as we’ve refined our algorithms, we can envision a future where drone ports in the sea deploy robots equipped with sensors around-the-clock to monitor whale populations and provide real-time commentary.”
“We foresee a fleet of autonomous aerial vehicles that will converge on predetermined locations and times to interact with whales.” “When humpback whales breach, they create an opportunity for researchers to deploy autonomous drones or robots close to the marine mammals to collect data on visible characteristics and acoustic patterns, providing valuable insights into their behavior.”
Will exterior organizations employ virtual avatars to safeguard endangered sperm whales in their pristine habitats? This information could potentially be leveraged to divert vessels away from areas where sperm whales are prone to damaging their surroundings, thereby reducing the likelihood of a ship encountering and potentially harming a pod of these marine mammals.
“When considering the larger picture – the whales as a whole, the whale communities, and the intricate social structures they form – we can foster inspiration for conservation efforts and develop a deeper appreciation for the importance of protecting marine life,” Gil explained.
In addition to their findings on beluga whales, the researchers suggested extending these approaches to other vocalizing sea mammals.
“At Challenge CETI, our focus is on sperm whales, but we believe this concept can be applied to other marine mammals, as various species including humpback whales, diverse whale types, and dolphins engage in vocalizations.”
