I agree with your asessment of the steering wheel. I think the main point of this dicussion thread is not joystick VS wheel, but mechanical linkage VS drive-by-wire. Good points raised by all on both sides of the debate.
The biggest reliability hazard by far, and most unreliable link in the chain is the human driver. The list of things that can cause a fatal mechanical or electronic failure pales in comparison to driver causes. An incident that comes to my mind is one that occured some years ago near the community I live in. A heart attack behind the wheel on a rural highway resulted in a horrific multi-fatality accident with a school bus. For those of you that fear the loss of direct mechanical steering control, I assume you also ensure that you are in the best of health, drive strictly by the rules of the road, and only drive roads where you are quite certain all the other drivers are as diligent as yourself. These steps would maybe somewhat minimize the huge disparity in failure probabilities between the vehicle's systems and the driver.
Perhaps this is somthing that drive-by-wire may actually improve. If coupled with smart, speed-context scaled safe auto-limiting to prevent sudden steering actions and intiate throttle release where sensors would predict a collision or loss of vehicle stability could otherwise result, new technology could actually make a significant dent in the failure probability of the driver. I'd personally feel much safer behind the wheel of such a car, particularly if all the other vehicles around me are similarly protected.
The reason for a mechanical backup is, that mechanical
systems degrade in a predictable manner(assuming the
system passes through various inspection and tests).
Automotives are one, where irregular maintenance and
knowledge/skillfulness of the operator is not guaranteed.
Also the economy driven drive by wire systems found in
automotives are not very suitable for realtime duty.
High level design(not to mention the complexities of
realtime OS) coupled with failure in large electronic designs(not to be confused with the reliability of the LSI's itself), makes this even difficult, not to mention the rough operating env.
So, in essence the electronics in automotive that
are used in the critical realtime duty, will
be used in a backup mode to allow for graceful degradation in the event of a catastropic failure,
letting the mechanical sub system handle it in a
predictable but less effective manner.
My apologies, I was wrong: there are apparently no standard cars using drive-by-wire steering, just electric or hydraulic power-assist. But why not steering? If you say "steering and safety and reliability" are the reason--what about that accelerator? That is certainly electronic and software controlled, and can lead to accidents just as steering failure can.
The obvious answer is that the steering wheel was developed to meet the needs of controlling the direction of the vehicle. Nothing else is as intuitive or as refined (yes even a joystick!). The left/right turning is directly associated with the wheel motion whereas how would that translate into a joystick? It seems weird to consider it, although there are many robots that use remote joysticks for control. The faster more car-like robots seem to use steering wheels (USB based of course), while crab/swerve drive robots most often use joysticks as that motion control more directly translates.
Long, long ago, when I was new to driving, I and my friends had to make due with beater cars that we could afford on our meager high school and college incomes. Control problems (brakes, steering, acceleration) were not at all uncommon. Most of these beater cars had between 80,000 and about 150,000 miles on them. That's hardly old at all for a modern car.
More than one of us had an accelerator return spring break. That's bad, but we all managed to survive. I know of at least two incidences of total hydraulic break failure. One in a car of mine and one in a friends car while I was towing him. Scary.
Another friend had an A-arm break on his pickup truck. I once had a needle and seat valve work its way loose. The float bowl filled up with gas and then it ended up spewing out on the engine. How that did not result in a fire, I'll never know.
Why am I droning on about failures past? The message is that the more electronic my vehicles have become, the more reliable they have become. Without a doubt, electronic cars have and will exhibit catastrophic failures. No question about that. However, the old pure mechanical cars did too.
We need to keep a close watch on the designers and regulators for these new systems, certainly. but I don't think we should fear them more or even as much as the old systems. I'd love to see some data comparing the per-capita failures of pure mechanical cars vs new electronic / mechatronic cars. My guess is that we'd find an order of magnitude fewer failures in the new cars. If anyone has access to such data, I'd love to see it.
But we don't hate to disappoint you and your BOGUS claims. Please list which current automobiles and commercial vehicles have this Drive-By-Wire steering you claim so strongly.
Many cars today have electrically-assisted steering, and a few high-end have auto-parking. All of them retain mechanical linkage. Multiple manufacturers like Mazda, Chevy, Toyota have had recalls and advisories regarding the electric motors on their power-steering, but the reports always state that the worst failure is loss of Power Steering, not total loss of steering control. You just manhandle the steering wheel like you always did with your Dad's "classic" car.
Even the DARPA Challenge vehicles had mechanical steering linkages. The teams utilized either their own actuators to control the wheel, or took advantage of the steering assist, but the steering column remained.
A "joystick" (which I presume means something fairly substantial, not a twig sticking out of a small box) would work well for vehicle control if the vehicle was largely autonomous. Indeed, for terminal control such as offroad or parking in non-standard areas a joystick would be fine, and allow much more space in the cabin with the large clunky round steering thing removed.
The classic 1DOF steering wheel has advantages until autonomous vehicles are available though. Either hand can operate the wheel easily, resistant to erroneous body inputs, familiarity (they used these on sailing ships for millennia after all, so there must be something to them), provides leverage and variable steering control...
No has yet to suggest more novel control methods, such as proprioception-- for instance, vehicle by shifting weight and dispensing with the monkey hands as control objects. Most modern monkeys probably can't master this though, at least not easily, but the experience weds the user much more closely with the environment. This method could be attractive if wedded to safe area semi-autonomy.
Yet another potential failure mode, never mind the possible effects of the vehicle beside you with a ham radio operator and a hundred watts of RF.
A nearby atomic bomb explosion will likely kill the solid-state ignition system, the intact steering mechanism will let you coast over onto the shoulder so that farmer in the 1930's pickup truck behind you (breaker points ignition) can get by.
Of course, being that close to the explosion nothing else would matter much anyway...