trafficbrowser AI3D visiontiny games
Semantic communications
What if cars shared meaning instead of flooding the road with pixels?
Interactive simulation · no model download
A parked truck blocks the ego's view of a pedestrian darting across. An oncoming car (A1) sees them — does the ego get the warning in time?
Three comm modes · same scenario
A1 has line of sight to the pedestrian; the ego doesn't. Raw V2X could share A1's video — but a 4 MB frame takes seconds to arrive, and the ego would still have to do its own perception on the pixels. Semantic V2X transmits only what matters — '{ entity: pedestrian, pos: (x, y) }' — in 64 B at 80 ms latency. That's the difference.
Ego (you) Agent vehicle Emergency vehicle Pedestrian OBJECT packet BRAKE intent EMERGENCY broadcast
How to play: the demo starts in NO COMMS so you see the crash first. Then try RAW V2X — pixels stream between cars at huge bandwidth but arrive too late, and the ego still crashes. Finally switch to SEMANTIC V2X — every car sends 64-byte structured packets and the ego brakes in time. Same scenario, three radically different outcomes.
swipe to move through the room