I recently enjoyed Jacky Li’s (from Connected Lab) relevant summary of what is wrong with touchscreen-only interaction in cars. Read article here.
Having tested more than 50 car interfaces with several hundred participants, I would say that the key to good in-car usability is not focusing on one input methodology but to offer true multi-modality. Users might struggle with touchscreen input while driving but it is still superior in speed to jog-dial input while standing.
Also, while offline voice control is constantly improving (e.g. as in the 2018 Mercedes A-class), it often still does not provide the same accuracy as manual data input (especially for setting navigation destination). In addition, many participants in our studies are still not comfortable with voice input or state that they would not want to use it when riding with (potentially noisy) passengers.
I would add two more aspects to the discussion of touchscreen usability in cars: