INVITED KEYNOTE SPEAKER

Abstract of the talk:  

Operational Perception for Underwater Robots: From Vision and Sonar to Persistent Autonomy 

To realise underwater autonomy, optical and acoutical imaging sensors are essential to achieve understanding of the of the scenery. Backscatter and attenuation reduce the range and quality of optical imagery, often necessitating complementary acoustic sensing and making the transformation from raw sensor data to actionable information a central challenge for underwater autonomy.

From monocular and multi-camera systems to hyperspectral imagers and imaging sonars, different sensors expose different aspects of the underwater scene, from optical texture and spectral signatures to acoustic range and structure. 

But there can be a long step from research concepts to results. This talk will present applications for ship hull inspection, biology, archaeology and intervention.

Across these applications, approaches such as sonar-assisted visual odometry, multimodal SLAM, uncertainty-aware mapping, and cross-modal sonar–visual perception are taken into real underwater environments and adapted to their sensor and operational constraints. The resulting perception pipelines support concrete tasks ranging from autonomous hull inspection and 3D mapping to resident vehicle navigation, docking, and persistent monitoring.

Progress toward persistent underwater autonomy will depend less on finding a single superior sensor or algorithm than on combining complementary sensing modalities and extracting the information each can provide under changing conditions. Advances in AI and robotic sensing will broaden this toolbox, but their value ultimately depends on adaptation and validation in the marine environment.