An mmWave + UWB victim-location system that helps first responders find people through drywall, smoke, and fire — and renders them in the responder's own AR frame.
A wall-mounted hub sweeps a 60 GHz mmWave radar through 360° to locate people inside a building — distinguishing active from stationary/unconscious — then publishes their positions to a responder's phone. The phone fuses UWB ranging with ARKit visual-inertial odometry so each detection lands in the responder's own frame of reference, even from far away with no line of sight.
First responders entering an unfamiliar building have no good way to know where the people inside are — especially if those people are stationary or unconscious and can't call out. A persistently installed sensor hub plus a responder phone app can deliver that, cheaply, even through walls.
A TI IWR6843AOPEVM running the Vital-Signs & People-Tracking binary uses a 120° 60 GHz scan; the onboard DSP identifies people by detected vitals and movement.
An ESP32 configures the radar over USB host, steps a stepper 90°→360° for full coverage with overlap, and runs ESP-NN to classify tracks as active / unconscious / phantom.
Detections are transformed sensor → hub via SolidWorks-derived transforms + stepper angle, then the app pairs UWB ranging with ARKit VIO (Gauss-Jordan) to place them in the responder's frame.
No single sensor solves it: radar sees through walls but not absolute position; UWB + VIO give the responder-relative geometry. Fusing them is what wins the "find the victim" problem.