The unfortunate truth is that a touchscreen alone will never be the ideal carpc interface. I actually really like the ford Sync system. MOST things don't need the touchscreen to be acomplished, but all things CAN be done from the touchscreen. of course, the UI itself is a little convoluted, but the interface system is a very good balance of speech, physical controls, and touch screen. (I also like the mini-modes that each function has, so you can have, say, two functions on the screen at once (split screen nav | music), or zoom in to a single fullscreen (big 'ol map).
Well, whatever you guys decide about a UI design, we all know that the number one thing it must be in a car, is realtime fast, and no stuttering or delays, because whenever you DO ever happen to take your eye off the road for a second to hit a button and go back to looking at the road, then when you go back to hit the next button, you can't have a missed click, or a stalling frozen screen with an hour-glass hanging in the middle because that is what makes carputers have the potential to be dangerous. You need the screen to be where you expected it should be based on what buttons you have previously pressed. So it needs to be lean, and mean, and preferably with the hardware matched to only the required software and kernel modules that the 'puter will actually ever use in real life, so it maintains that light 'feel' we all love. I had the idea a while back to try out Windows 7 Embedded Eval version (like what they advertise is for kiosks and bank machines, all the way to semi normal systems like tablet PCs etc), because the embedded OS would allow you to trim down the kernel and all the OS components to run only what is necessary to make the system run optimally. Then with a nice UI that was built around that embedded OS, it would run super fast kinda like Android does. Because in my opinion, carputers are right up there with medical equipment when it comes to design perfection. They can't be buggy, non-responsive, or slow, because someone could die.
I'm sure you guys could still just build this UI to act more like Android does (taking the positive points from theirs and use those) but most specifically I think you should build your UI to have more voice command input implementation, right off the get-go. I would like to see the UI or a supporting app be built around a feature that would let me map any voice command to any single function that could be ran on the carputer for example, so I could literally do anything without needing to press any button. There is an open source project for Windows called AutoHotKey ( http://www.autohotkey.com/ ) and it allows you to set up any kind of complex command or mouse click or anything to be ran with a hotkey, so if we had something like that, but was triggered with voice commands as well, then you could build your UI to be run by touch or by voice...
There is no ideal.
There is no single answer.
EG - I reject interfaces that need sighting for "common & essential" functions.
As to GUI - I recall being blown away by the design & intuitiveness etc of "DoubleKiller" (Jan Schlüter; Big Band Enterprises), but that was back in 2006 amidst software that belonged back before windoze-2 days...
The problem with standardising GUI & UI is NOT preventing better methods (hence why rules are often not what you MAY to, but what you may NOT do).
Then there's culture, platform familiarity....