Robots created by a team working at the University of California, Santa Barbara are able to look through solid walls using just Wi-Fi signals.
With potential applications in search and rescue, surveillance, detection and archeology, these robots have the capability to identify the position and outline of unseen objects within a scanned structure, and then categorize their composition as metal, timber, or flesh.
Working in pairs, the robots traverse the perimeter of an object or structure and alternately transmit and receive Wi-Fi radio signals between each other through the object being scanned. Exploiting the differences in transmitted and received Wi-Fi signal strengths to show the presence of hidden objects, the system uses a wave-propagation model with a target resolution of around 2 cm (0.8 in). By measuring the received field strengths of these wireless transmissions, the robots are able to produce an accurate map of the structure detailing where solid objects and spaces are located.
Though these aren’t the first robots that have been claimed to be able to see through concrete – the Cougar20-H surveillance robot achieved that some years ago – other systems have relied on a number of GHz-range, high-power radio sensor arrays that were essentially complex radar systems. Similarly, a fixed Wi-Fi system created by MIT was able to detect movement behind walls using Wi-Fi as its transmitter and receiver, but the resolution was too low to do more than detect movement, let alone categorize and identify objects.
The UCSB robots, however, rely solely on interpretations of Wi-Fi radio transmissions which – even though they are still of both lower strength and much lower dynamic range than higher-powered arrays – indicates that the signal processing and post-capture computation must be key to their “X-ray vision” capabilities. This is borne out in the team’s assertions that they use wavelet, total variation, and spatial domain filters and computations in their receiving equipment and processing computers, and the use of a SLAM algorithm in their on-the-fly mapping computations.