Autonomous underwater glider passively follows sperm whales by their voices
Key Takeaways
- ▸AI-powered underwater glider system can autonomously track sperm whales using their echolocation calls with minimal environmental impact
- ▸The 'backseat driver' uses acoustic detection, source separation, and angle-of-arrival estimation to follow whales in real time while remaining completely passive
- ▸Unlike existing tracking methods, this system enables long-term monitoring while maintaining 100+ meter distance to avoid disrupting whale behavior
Summary
This research describes a novel 'backseat driver' system for underwater gliders that uses artificial intelligence to autonomously track sperm whales. The system employs a four-element hydrophone array mounted on a Project CETI-SeaExplorer glider and uses advanced machine learning techniques—acoustic detection, source separation, and angle-of-arrival estimation—to process sperm whale echolocation clicks in real time. By analyzing these acoustic signals, the AI system automatically adjusts the glider's bearing to follow the whales, creating a completely autonomous tracking system.
Unlike traditional whale monitoring approaches that rely on boats, satellite tags, or fixed moorings, this system is entirely passive, emitting no radiated noise except brief buoyancy adjustments (roughly once per hour). The glider maintains a safe distance of over 100 meters from the whales, ensuring minimal impact on their behavior while still collecting detailed acoustic data about their communication and movement patterns. This approach enables sustained, long-term monitoring of marine mammals in their natural habitat without intrusion.
The researchers validated the system through controlled sea experiments and field trials off Dominica, demonstrating reliable source separation, accurate angle-of-arrival estimation, and acceptable command response latencies. By releasing the code and data as open source, the team is making this technology available to marine conservation researchers worldwide, potentially advancing non-invasive whale monitoring across the globe.
- Code and data are open-sourced, advancing the field of marine mammal monitoring and conservation research
Editorial Opinion
This elegant application of AI to marine conservation fills a genuine gap in whale-tracking technology—most existing methods either require invasive tags or involve vessel intrusion that disrupts ecosystems. The passive acoustic approach is genuinely innovative, and by open-sourcing the code and data, the team enables researchers worldwide to adopt and build on these methods. This represents exactly how AI should be deployed in conservation: with minimal intervention, maximum ecological insight, and shared knowledge.
