Why the accelerator and why not steering or breaking ? It's a matter of failure mode: whilst a failure of the 'gas pedal' is easily handled switching to "limp home mode" (=better slow than towed), there is no such way for steering. Electrical breaking seems to be under evaluation but I do not know what the failure reaction will look like.
Many years ago I designed-in a joystick from Flight Link Controls Ltd (now owned by PWT). The joystick used a differential inductive transducer principle which sounds tricksy but used only a handful of R,C,TR discretes to give a contactless, failsafe design, whatever component or connection failed.
The reason for this post is that if you can get a copy of the schematic you will one of the most cunning, elegant and well-thought-out analog designs you will ever meet.
These joysticks had a very good track record, and were/are often fitted to wheelchairs etc for users who often could not have manipulated a wheel anyway.
"There has to be, first off, a good reason to change something...."
The manufacturers have a good reason - they could design the front so that the controls can be placed either side without rearranging the whold car....
Ref my luddite comment...as I said I am one myself. Until vehicles have as much input as I get while driving, I wouldn't trust a DBW system either.
And I agree ref the steering wheel..Don't reinvent the wheel, it's done a good job for reasons many posters have outlined, and a wheel is as good as anything as a man-machine interface. And some vehicles already have self-steering into parking spaces.
I think (and I kind of hope) that a completely DBW vehicle won't happen anytime soon, but as I said above... anyone care to tell me it will never happen?
When a joystick is incorporated part of that closed loop control system MUST be taken out of the control of the human and put into the car. That has nothing to do with age, experience, what you are familiar with, etc. That is just the reality of replacing what we could call a low gain input, the steering wheel, with a high gain input, i.e. the joystick.
Do I think there are some inherent safety advantages with the current mechanical linkage? Within the confines of a car: cost sensitive, questionable maintenance, etc. that for the foreseeable future, the mechanical linkage is an acceptable and perhaps cost effective redundant system. Is that likely to be the case always? Highly unlikely.
Do I expect that driving over time will be taken somewhat out of the hands of the driver? ... That is an absolute, 100% given, at first on major highways and then perhaps to major secondary roads, though perhaps a long time till it hits the local roads. Don't believe that? Imagine being able to rush along on crowded highways during rush hour at 120+kph (75mph) because your computer controlled vehicle is only sitting 3 metres (10 feet) from the car in front of you ... just like the one directly behind you. Impossible with driver input, viable if computer controlled. 3-5x the number of cars on the same amount of road and less fuel usage due to drafting. Of course, maybe our vehicles will just be pulled along by magnets, etc.
To someone's point, progress is progress. It has its hiccups. Almost impossible to "work" on my car anymore, but since it keeps going and going with minimal maintenance, who cares. I don't use the "DOS" prompt much any more either, but again, I don't have to.
I have been playing video games with joysticks since I was a pre-teen which makes that over 30 years ... Yes, that is how long video games have been around!
Could a joystick be used to effectively "pilot" a vehicle?
Need to consider that steering is a closed loop system. Decide where I want to go, turn the wheel, see if I am going where I want to go, make adjustments. While there is speed sensitive steering that adjusts the turns ratio between the steering wheel and the real wheels, for the most part in any given situation, there is a level of linearity (or at least piecewise linearity). There is also a fairly significant movement arm, i.e. the radius of the steering wheel multiplied the required angle to turn. That allows fairly small corrections with fairly gross movements.
A joystick is not able to easily replicate this relationship between human input and what happens at the wheels.
The advantage of a rotary control for steering (and also for radio volume) is the decoupling of the desired action (changing direction/volume) from the variable inertial forces of a non-uniform driving surface. Consider the radio volume control while driving on a bumpy road: You grab the knob with your hand and rotate to the correct volume, with any road-bump accelerations to your arm being ignored as your hand position is stabilized by the knob supports. Now, consider a vertically mounted slide-pot: You grab the slider and begin to position for the new volume setting, whil a road bump causes your arm to make violent changes to the position.