CORALE is an on-going project I am developing under the mentorship of Professor Jaime Fernández Fisac and Madison Bland at the Princeton University Safe Robotics Lab.
Abstract:
CORALE: Coral Optimized Robotics for Autonomous Learning and Environmental Navigation – is an autonomous robotic system to safely navigate and monitor complex coral reef environments. Coral reefs, critical to marine biodiversity, are increasingly threatened by pollution, climate change, and physical disturbances. Traditional monitoring methods, which rely on divers, are costly, logistically challenging, and risk causing damage to reefs. Recent developments in robotic technology offer potential solutions, but coral reefs’ intricate structures and dynamic ocean flows present unique challenges for autonomous systems. Unlike conventional sensor-based navigation, which lacks adaptability in highly variable environments, CORALE integrates reinforcement learning (RL) and real-time flow prediction to enable adaptive, context-aware navigation.
Using OpenFOAM and HoloOcean, I developed a simulated coral environment to train the RL model in safe obstacle avoidance, incorporating safety filters and cost functions to prevent collisions and prioritize efficient movement. The system is designed to transition onto an underwater remotely operated vehicle (ROV), with initial testing planned in a controlled tank environment followed by field trials in natural reef ecosystems. These tests will validate CORALE’s adaptability and precision in real-world settings, supporting conservation efforts by reducing human impact on sensitive reef habitats.
Besides monitoring, CORALE can contribute to direct restoration by providing services for accurately delivering materials, like CoralGel, for in-situ restoration efforts, reducing human intervention. Future developments will refine the RL model and explore further applications for coral restoration, establishing CORALE as a valuable tool for both ecological monitoring and restoration.
