MIT advances wireless sensing with generative AI
Wireless sensing advances could boost logistics and smart homes, providing privacy-friendly alternatives to cameras while improving machine perception of hidden spaces.
Researchers at MIT have developed a new approach that combines generative AI with wireless signals to detect objects hidden behind obstacles. The system uses Wi-Fi-like millimetre wave signals to build partial reconstructions and then completes missing details with AI.
Traditional methods struggled with limited visibility due to how signals reflect off surfaces, often leaving large portions of objects undetected. The new technique, Wave-Former, uses generative AI to fill missing data, improving reconstruction accuracy by nearly 20%.
An extended system, called RISE, takes the concept further by mapping entire indoor environments. By analysing reflected signals from human movement, the system reconstructs room layouts using a single stationary radar, removing the need for mobile sensors.
Applications range from warehouse automation to smart home robotics, where understanding hidden objects and human positions is critical. Unlike camera-based systems, the technology also preserves privacy, marking a significant step forward in wireless sensing capabilities.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!