PIRATE Autonomous Marine Robotics
Monitoring marine ecosystems and tracking underwater animals are fundamental tasks in marine science and conservation. Traditional observation methods rely heavily on manual surveys, diver-based monitoring, or stationary sensor infrastructures, which can be expensive, labor-intensive, and spatially limited. Recent advances in robotics, artificial intelligence, and embedded computer vision are enabling a new generation of autonomous marine platforms capable of collecting environmental data with minimal human intervention.
The PIRATE project (Precision Imaging Real-time Autonomous Tracker and Explorer) explores the integration of autonomous surface vehicles, acoustic localization, and onboard visual perception to enable persistent observation of marine environments. The system is designed to autonomously navigate predefined routes, localize acoustically tagged underwater targets, and capture visual observations using onboard cameras and real-time computer vision algorithms.
The PIRATE Platform
PIRATE is an autonomous unmanned surface vehicle (USV) platform developed for marine research applications. The system integrates multiple sensing and processing components that work together to enable fully autonomous operation in coastal environments.
The platform combines an acoustic telemetry receiver capable of detecting signals from tagged marine animals with onboard visual perception modules that analyze the surrounding environment in real time. Visual perception is performed using deep learning–based object detection and tracking algorithms operating directly on embedded computing hardware.
The PIRATE architecture also includes autonomous navigation modules that allow the vehicle to follow predefined routes, perform listening patterns for acoustic localization, and reposition itself dynamically based on incoming sensor observations. By integrating acoustic sensing with visual perception, the platform enables coordinated localization and observation of marine targets.
Autonomous Perception and Tracking
A key objective of the PIRATE project is the development of a real-time perception pipeline capable of detecting and tracking objects in marine environments under computational constraints. The system employs a hybrid detector–tracker architecture that combines deep neural network–based object detection with lightweight visual tracking algorithms. This design allows the platform to maintain continuous observation of targets while minimizing computational overhead.
The perception system runs entirely onboard the vehicle using embedded GPU hardware, enabling real-time processing without reliance on remote computation. This capability allows the platform to autonomously respond to visual observations and adjust its navigation behavior accordingly.
Resources
The PIRATE project includes a growing collection of datasets and experimental results generated during field deployments and controlled experiments. These resources include visual datasets collected during marine surveys, laboratory experiments, and acoustic telemetry logs.
Published Papers
- D. Zlotnikov et al., PIRATE —Precision Imaging Real-Time Autonomous Tracker & Explorer, Journal of Marine Science and Engineering.
Acknowledgments
The PIRATE research project was funded in part by the Israel Science Foundation under Grants No.~501/25 and 504/25.