This would be great if it were reliable, but there are several reasons FSD can't use this approach
- Many (I think most) departments don't have such a system installed on their emergency vehicles, especially in rural areas where volunteer fire departments struggle to afford even the most basic equipment
- Some of these systems don't use visible light as the means to trigger the light change. Some are IR, some are radio, and I bet some rely on city infrastructure and GPS on the apparatus (i.e. central servers monitor en-route vehicles and trigger light changes ahead of their positions)
- Even the optical ones would be forward-facing only
- Even if there were a standard for this optical system (it would be in NFPA 1901 if anyone wants to do some reading, but I don't think it's there), it would only work in the US, and while some other countries might have similar standards, it's likely some countries wouldn't have such systems at all, so FSD couldn't work there without other means of identifying emergency vehicles
I've served in a volunteer fire department where I know for a fact we had no such system on our makeshift pickup truck response vehicle, although we did have flashing lights on top.
I've long thought identifying emergency vehicles would be one of the most difficult challenges to solve to get to true full-self driving. There are lots of different styles/colors/placements of lights on lots of different base vehicle chassis that are emergency vehicles, and there are also numerous vehicles with flashing lights you don't pull over for (snow plows, garbage trucks, random pick-up trucks used by construction companies that leave the lights flashing for no apparent reason, and there are also places (Korea is one example) where actual police cars have their lights slowly flashing at all times but switch to a much more aggressive flashing pattern when responding to an emergency)