i'd guess adding an abstraction layer (between a general purpose microcontroller and the custom interfaces it needs to support) is likely not worth the added complexity (and bugs), since cars are on development cycles of many years anyway (for the benefit of simpler software development, and the risk reduction of being able to swap out the microcontroller if need be).
They actually theoretically have a thing for this already. It's called AUTOSAR (now AUTOSAR Classic). It's supposed to be a component architecture to write down the software against a framework that can be backed by different MCUs. Of course, in practice much of the underlying esoteria of any given MCU/board configuration bled up through the abstractions and/or produced custom extensions.
Indeed, it's a mess and the surrounding tooling ecosystem is a pile of weird Windows GUI tools of about the quality you'd expect for very expensive tools that do very little provided to a captive customer base. It's all very circa 1990s RAD-tool mania-esque still in many ways.
That said, we're working with some groups in some automotive companies who are breaking this mold rapidly.
I can't upset the NDA Gods on this one, unfortunately. I can say one is a traditional OEM that you wouldn't necessarily expect to be the sort to take vanguard steps like this and the other is a newer OEM that's not as encumbered by tradition (though due to the supply-chain also can't rid themselves of it).
Shameless plug WARNING: we (https://auxon.io) are hiring for engineering, marketing, and BD if industrial & operational technology is your thing.
I never signed the NDA, although I follow the space distantly, and they went public on their own website a couple years ago with a list:
BMW, Bosch, Continental, Daimler AG, Ford, General Motors, PSA Peugeot Citroën, Toyota, and Volkswagen.
I always rooted for an older competitor GENIVI which (very handwavy in a general sense) boiled down to "make dbus great again". I just always have a soft spot for anything that replaces CORBA and DCOP. More or less GENIVI had the same players as AUTOSAR.
I believe AUTOSAR suffers from trying to do too much; GENIVI's "all you need is a compatible bus" is the minimum you need so simplicate and add lightness and that's what should win; doesn't really matter if that guy runs QNX and that other guy runs FreeRTOS as long as the busses talk; whereas AUTOSAR tries to own and control everything top to bottom which just ends up stifling any productivity.
If you look through the "Adaptive AUTOSAR" literature and documentation that's out there you can find that enormous chunks of it are lifted straight from GENIVI's stack/specs.
However, GENIVI/Adaptive-AUTOSAR tends to serve a different function in the vehicle architecture. It's mostly cockpit and less control/platform. AUTOSAR Classic is still the champ on the MCU side of the house.
