IIRC from greentheonly’s twitter, there’s evidence of a brief experiment to do early fusion of vision and radar in the perception model