MIT robots can now identify hidden objects thanks to a system that can reconstruct their shape with 96% accuracy. This innovation, called mmNorm, uses millimeter-wave (mmWave) signals to generate detailed three-dimensional reconstructions of elements not visible to the naked eye, and is already being integrated into prototype MIT robot prototypes prototype robots to improve their non-contact inspection capabilities.
The system was conceived to operate in environments where visibility is limited, such as factories or warehouses. mmNorm picks up reflections from waves passing through common materials – such as cardboard, plastic or thin walls – and applies a specialized algorithm that reconstructs the shape and orientation of hidden objects.
MIT Robots Calculate Hidden Shapes Using Reflection
Unlike conventional back projection techniques, which offer low resolution for small objects, mmNorm incorporates the concept of surface normal, estimating the direction in which each point of an object reflects the signal. This makes it possible to reconstruct complex, curved details, such as the handle of a tool or the silhouette of a piece of cutlery.
The team used a robotic arm equipped with radar to scan from multiple angles, collecting votes from each antenna on the surface orientation. The result is a three-dimensional is a three-dimensional image that allows the robots to recognize specific shapes without the need for physical contact.
Industrial and domestic applications
The efficiency of the system, which does not require higher bandwidth than current methods, makes it a viable option for multiple applications. From automated assembly lines to healthcare environments, robots could identify utensils hidden in drawers, check the condition of packaged parts, or handle tools without risk of damage.
In addition, the technology shows potential in areas such as augmented reality and augmented realitywhere it could offer real-time vision of occluded objects, and in security systems, improving detection at airport checkpoints or remote inspections.
A change in the robotic perception approach
According to the researchers, this development involves a new way of using mmWave signals, traditionally used in radar. “We wanted to go beyond detecting objects: we sought to understand their shape,” explained those responsible for the project.
In the future, the team plans to extend the system’s capabilities to work with denser materials and less reflective surfaces, as well as to refine the reconstruction resolution.
This technique could redefine the way robots interact with their environment, providing a more complete, accurate and adaptive perception in contexts where direct visibility is not possible.
Follow us on social networks and don’t miss any of our publications!
YouTube LinkedIn Facebook Instagram X (Twitter) TikTok
Source: MIT
Photo: Shutterstock
