AI-powered sensing
The team faced significant challenges in developing algorithms for PanoRadar, aiming to maintain high-resolution imaging while the robot moves.
Achieving LiDAR-level detail with radio signals required measurements from multiple positions at sub-millimeter accuracy, demanding precise control even as the robot moved. Another hurdle involved teaching the system to interpret complex indoor environments, which they addressed by training the AI to recognize patterns in radar signals, much like human perception.
Using LiDAR for initial verification, the system refined its understanding, demonstrating superior tracking in smoke-filled areas and mapping spaces with glass walls, which traditional sensors often struggle to detect.