- Mmnorm reconstructs complex hidden forms using Wi-Fi frequencies without touching the object
- Robots can now look inside messy drawers using reflected signals from the surrounding antennas
- Mits technique beats current radar accuracy by 18% across more than 60 tested objects
In environments where visibility is hindered, such as interior boxes, behind walls or under other objects, artificial intelligence could soon have a new way of moving on.
Researchers at MIT have developed a technique called MMNORM that uses millimeter wave signals, the same frequency range as Wi-Fi, to reconstruct hidden 3D objects with surprising accuracy.
“We have been interested in this problem for a while, but we have hit a wall because previous methods while they were mathematically elegant did not get us where we had to go,” said Fadel Adib, senior author of the study and director of Signal Kinetics Group at MIT.
Overcome radar restrictions
Previous techniques rely on back projections that produce low resolution images and fail when applied to small, occluded items such as tools or tools.
The researchers found that the error lies in the supervision of a physical trait known as speculity – how millimeter waves reflections behave as mirror images.
Instead of simply measuring where signals jump back, the direction of the mmnorm estimates the direction of the surface, what scientists call the surface normally.
“To rely on Specularity, our idea is to try to estimate not only the location of a reflection in the environment, but also the surface at the time,” explained Laura Dodds, lead author of the paper.
By combining many such estimates from different antenna positions, the system reconstructs the 3D curvature of an object that distinguishes between forms that are nuanced as a mug’s handle or the difference between a knife and a spoon in a box.
Each antenna collects reflections with varying strength depending on the orientation of the hidden object.
“Some antennas may have a very strong vote, some may have a very weak vote, and we can combine all votes together to produce a surface agreed by all antenna locations,” Dodds added.
This new approach achieved a restructuring accuracy of 96% on over 60 objects, exceeding existing methods that reached only 78%.
The system worked well on objects made of wood, plastic, glass and rubber, although it is still struggling with dense metal or thick barriers.
When researchers work to improve the solution and material sensitivity, the potential use cases are growing.
In security scan or military contexts, MMNORM could reconstruct the shape of hidden items without opening bags or boxes.
This capacity may prove to be important for AI-driven robots in warehouse automation, search-and-rescue or even assisted housing environments.
Via Techxplore
