CNN-Driven Seabed Exploration: Reactive Path Adjustment for Species Detection and Control

This work develops and tests an adaptive nav-igation framework for Autonomous Underwater Vehicles (AUVs) to improve Nephrops norvegicus population density measurements. The framework employs a Convolutional Neural Network (CNN) to detect individuals of this species and their burrows, triggering adaptive AUV reactions around biological targets. This approach enhances mission coverage and population density estimates by temporarily altering the vehicle's trajectory. The framework is inte-grated as a high-level layer in a ROS-based architecture and it first tested through simulations, with plans for real-world deployment during the PLOME project.