Humanoid robots could soon see through walls with MIT’s imaging tech

MIT researchers have developed an advanced imaging technique called mmNorm that enables robots to see through walls and other obstructions like cardboard boxes by using millimeter wave (mmWave) signals. These signals, similar to those used in Wi-Fi, can penetrate materials such as plastic, interior walls, and cardboard, reflecting off hidden objects to allow sensors to capture detailed 3D reconstructions. In testing, mmNorm achieved a 96 percent accuracy rate in reconstructing complex objects like silverware and power drills, significantly outperforming current state-of-the-art systems that reach about 78 percent accuracy. The key innovation behind mmNorm lies in its ability to estimate the surface normal—or the direction a surface reflects signals—by leveraging the property of specularity, where mmWave signals reflect like light off a mirror. Traditional radar systems often miss such reflections when surfaces are angled away from the sensor. By incorporating this directional information into its algorithm, mmNorm produces highly accurate 3D models of hidden objects. Potential applications include