When remotely operated vehicles (ROVs) settle on the seafloor or dig through sand, they often kick up sediment that clouds their cameras, forcing them to wait for the water to clear. A new system from the Woods Hole Oceanographic Institution (WHOI) solves this problem. Developed by Amy Phung and her advisor Richard Camilli, the approach combines sonar with an AI algorithm to create real-time 3D maps even in zero visibility.

The vehicle first uses sonar to quickly map its surroundings, as sonar works equally well in cloudy or clear water. However, sonar alone lacks detail. To overcome this, the researchers paired it with an image-matching algorithm from French researchers that estimates the relative depth of each pixel in a 2D scene. This fusion allows the ROV to navigate safely and get close enough to objects for its cameras to capture clear, detailed visuals.

Camilli likens the technology to navigating a china shop in the dark: you can find a specific coffee mug without knocking anything over. The system enables ROVs to operate in turbid conditions where they previously had to wait or risk collision.

Potential applications are broad, including scientific exploration of deep-sea ecosystems, underwater construction and maintenance, and safely handling unexploded undersea mines. By making ROVs more autonomous and effective in murky environments, this innovation could significantly expand underwater robotics capabilities.