iCORE Laboratory Louisiana State University

Perception for marine environments: RGB, sonar, and event-based

Advancing marine autonomous vehicle capability

(a) Photograph of an underwater wreck that we surveyed to build a three-dimensional map; (b) reconstructed 3D map developed by our team.
(a) Photograph of an underwater wreck that we surveyed to build a three-dimensional map; (b) reconstructed 3D map developed by our team.

Autonomous marine robots face significant perception challenges due to poor underwater visibility, changing illumination, and complex environmental conditions. Our research investigates multimodal perception and vision-based navigation architectures that leverage RGB cameras, sonar, and event-based sensors to enhance environmental awareness, autonomous operations, and improved environmental knowledge. We explore attention-driven perception, target tracking and interaction, underwater anomaly detection, and environmental observation focused on spiking neural networks to process event-based data. Through the development and evaluation of these approaches on marine robotic platforms, we aim to improve the robustness and reliability of perception and navigation in challenging aquatic environments.

We are using and developing novel algorithms for event-based imaging, which uses a bioinspired approach to sensing that leverages the dynamics of the image (change over time) for sparse, neuroscience-inspired sensing.
We are using and developing novel algorithms for event-based imaging, which uses a bioinspired approach to sensing that leverages the dynamics of the image (change over time) for sparse, neuroscience-inspired sensing.

These projects have been funded by the National Science Foundation, National Oceanographic Atmospheric Administration, and Coastal Protection and Restoration Agency