Water/wastewater
Engineers at MIT and the Woods Hole Oceanographic Institution (WHOI) have developed an underwater mapping technique that allows vehicles to navigate through murky, low-visibility water.
The method, named Sonar-MASt3R, fuses visual images from optical cameras with acoustic data from sonar sensors.
Sonar is used to quickly map the general shape of the surrounding environment, even where visibility is poor.
Once a vehicle has identified a shape of interest in that sonar map, it can move close enough for its optical cameras to resolve the object in visual detail.
Remotely operated underwater vehicles often lose the use of onboard cameras when they disturb sediment on the seafloor, or when they operate in naturally cloudy water.
Historically, the only option has been to wait for sediment to settle before continuing.
Optical cameras alone provide detailed imagery, but only in clear, well-lit water.
Sonar sensors work regardless of visibility, determining the shape, distance and depth of objects by measuring the timing and angle of returned acoustic waves, but sonar data on its own lacks visual detail.
Combining the two, an approach known as opti-acoustic fusion, has been attempted before.
Amy Phung, the MIT graduate student who led the new work, said prior techniques have mostly been geared toward object recognition rather than high-resolution mapping, and few work in real time.
Sonar-MASt3R builds on an existing image-matching algorithm, MASt3R, developed by researchers in France, which estimates the relative depth of each pixel in a scene from ordinary camera images.
That approach can generate a 3D map in real time, but has no absolute sense of scale.
"It will say 'this pixel is five units closer than this pixel,' but it can't say whether that's five metres or five feet," Phung said.
Sonar corrects for this. The timing of sonar reflections translates directly into a specific depth and distance for the objects that a signal bounces off, giving the mapping system a fixed scale to work from.
Phung and her co-author, WHOI senior scientist Richard Camilli, used sonar data to correct the algorithm's scaling, producing precise 3D maps even in murky water.
The team tested Sonar-MASt3R using a tank fitted with a robotic arm carrying an underwater camera and a sonar sensor.
Objects including a small boulder, a coffee mug and a packing crate were placed in the tank alongside sediment.
Each test began with a sweep trajectory, in which the arm moved across the tank to build a coarse sonar-based map of the surroundings.
That map then guided close-up camera images, which were used to add resolution.
A keyframe approach compared each new image to the previous one, adding it to the map only if it contained new information, allowing the system to fill in visual detail in real time.
Across eight levels of turbidity, Sonar-MASt3R produced more accurate 3D maps than other opti-acoustic fusion approaches tested, resolving centimetre-scale detail even in the cloudiest conditions.
In the murkiest water, where the cameras alone could not see through the sediment, sonar still produced a usable rough map that let the arm navigate safely toward specific objects.
"An analogy would be if you were to go into a china shop in the dark, and try to pick your way around to find a specific coffee mug without knocking things over," Camilli said.
"This would allow you to do that."
Camilli said the work was motivated partly by the challenge of safely recovering unexploded underwater ordnance, much of which sits in surf-zone environments where poor visibility compounds the difficulty of safe removal by robotic vehicles.
"There can be old explosives in areas that make it unsafe for ships to be in, and the ability to get rid of those safely is best done by robotics," Camilli said.
"A lot of these explosives are set in surf zone environments where visibility adds to the challenge of doing this safely. That's one of many applications that our technique can be used for."
Beyond hazard clearance, the researchers see potential uses in scientific exploration and underwater construction and maintenance, where knowing the precise location of submerged infrastructure or hazards before moving equipment closer would reduce risk.
Phung presented the work at the IEEE International Conference on Robotics and Automation (ICRA); the research was supported, in part, by NASA and the National Science Foundation.
The team plans to test the approach next in open water, where they expect the mapping task to be more straightforward than in a tank, where acoustic reflections off the tank walls complicate processing.
IET 36.3 May
Product directory
.jpg)
