Cloudy underwater conditions can be a major problem for robotic vehicles exploring the ocean.
When underwater robots land on the seafloor, inspect shipwrecks, or dig into sandy areas, they often stir up clouds of sediment that block their cameras.
In many cases, operators have no choice but to wait for the water to clear before continuing their work.
Now, researchers from MIT and the Woods Hole Oceanographic Institution (WHOI) have developed a new technology that allows underwater robots to “see” through murky water and safely navigate even when visibility is extremely poor.
The new system, called Sonar-MASt3R, combines information from optical cameras and sonar sensors to create detailed 3D maps of underwater environments in real time.
The idea is similar to combining the strengths of two marine animals.
Sonar works like a dolphin’s echolocation, allowing the robot to detect shapes and distances using sound waves.
Cameras provide detailed visual information, similar to how a sea turtle sees nearby objects. Together, they allow a robot to understand its surroundings even when visibility is limited.
Traditional underwater cameras can capture highly detailed images, but they struggle in cloudy or dark water. Sonar, on the other hand, works well in both clear and murky conditions because it uses sound rather than light.
By sending out acoustic signals and measuring their reflections, sonar can determine the location, size, and shape of objects. However, sonar images often lack fine visual detail.
The new system combines the advantages of both technologies.
The researchers built their approach on an existing image-processing system called MASt3R, which can generate 3D maps from ordinary camera images. One limitation of that system is that it can estimate relative distances but cannot accurately determine the true size and scale of objects.
Sonar solves that problem by providing precise distance measurements. By combining sonar data with camera images, Sonar-MASt3R creates accurate, scaled 3D maps of underwater environments.
To test the technology, the researchers set up a water tank filled with sediment and placed various objects inside, including a coffee mug, a packing crate, and a small rock. A robotic arm equipped with a camera and sonar sensor then scanned the tank.
The sonar first created a rough map of the environment, even when the water was heavily clouded with sediment. This map allowed the robot to identify where objects were located and safely move closer to them. Once near the objects, the camera captured detailed images that improved the map’s resolution.
The team tested the system under eight different levels of water cloudiness. Even in the murkiest conditions, where the camera alone could not see through the sediment, the sonar-guided system successfully located objects and revealed details as small as a few centimeters.
The researchers believe the technology could have many important uses, including underwater scientific exploration, infrastructure inspection, construction work, shipwreck surveys, deep-sea recovery operations, and the safe detection of underwater explosives.
Future testing will take place in natural ocean environments. The team expects the system to perform even better outside the laboratory, where sonar reflections are less complicated than in a confined tank.
Ultimately, the researchers hope the technology will allow underwater robots to carry out missions that are currently too difficult or dangerous because of poor visibility, opening up new possibilities for exploring and working beneath the waves.
Source: MIT.
