Imagine if a self-driving car could spot a pedestrian before they step into view from behind a building.
Engineers at the University of Pennsylvania have developed a new system that brings this idea closer to reality.
The technology, called HoloRadar, allows robots to “see” around corners using radio waves and artificial intelligence.
Most robots and autonomous vehicles rely on cameras and LiDAR, which use light or lasers to detect objects.
These tools work well when something is directly in front of them.
But they struggle with objects hidden behind walls or around corners.
Some past research has tried to solve this problem using visible light reflections or shadows, but those methods depend heavily on good lighting conditions and often require controlled environments.
HoloRadar takes a different approach. Instead of using visible light, it uses radio waves. Radio waves have much longer wavelengths than light waves.
In traditional imaging, this has been considered a disadvantage because it lowers image resolution. However, the Penn research team realized that for seeing around corners, longer wavelengths are actually helpful.
Because radio waves are much larger than the tiny bumps and roughness on walls, flat surfaces such as walls, floors, and ceilings act like mirrors for radio signals.
When a robot sends out a radio pulse, the signal can bounce off these surfaces and travel into hidden spaces. The reflections then bounce back to the robot, carrying information about what is out of sight. In this way, the environment itself becomes a network of invisible mirrors.
However, interpreting these reflections is not simple. A single radio pulse can bounce multiple times before returning, creating a complex mix of signals. To handle this, the researchers developed a custom AI system.
The first part of the system improves the quality of the raw radio data and separates different reflected signals.
The second part uses a physics-based model to trace the paths of those signals backward. This allows the system to figure out where objects actually are, rather than where their reflections appear to be.
The team tested HoloRadar on a mobile robot in real indoor environments, such as hallways and building corners. The system successfully reconstructed three-dimensional scenes, including hidden walls and people standing outside the robot’s direct line of sight. Importantly, the system works in darkness and under changing lighting conditions.
HoloRadar is designed to complement, not replace, existing sensors like LiDAR. By adding another layer of perception, it can give autonomous vehicles and robots more time to react to potential dangers. In the future, the researchers plan to test the system outdoors at street intersections and in other complex environments.
This technology represents an important step toward safer, smarter machines that can better understand the world around them—even the parts they cannot directly see.
