A breakthrough system that uses generative AI to “see” objects hidden behind walls or obstacles using wireless signals like Wi-Fi. Instead of relying on cameras, the system interprets how signals bounce off objects and reconstructs what’s out of sight. This innovation represents a major step forward in robotic perception, allowing machines to better understand environments even when visibility is limited.
The key advancement lies in combining wireless sensing with generative AI. Earlier systems could only create rough outlines of hidden objects, but this new method builds partial reconstructions from reflected signals and then uses AI to fill in missing details. As a result, the system can generate much more accurate 3D shapes of concealed items, overcoming long-standing limitations in precision.
Beyond individual objects, the researchers also developed a system capable of reconstructing entire indoor environments. By analyzing how wireless signals reflect off moving people and surfaces, the AI can map rooms, including furniture and layout. Importantly, this approach works with a single stationary sensor and avoids the need for cameras, which helps preserve privacy while still enabling environmental awareness.
The potential applications are wide-ranging. In warehouses, robots could verify package contents without opening boxes, reducing errors and returns. In smart homes, robots could better track human movement and interact safely. Ultimately, this research shows how generative AI can extend perception beyond traditional vision—unlocking a future where machines can understand the physical world even when it is hidden from view.
