Voice control of cars?

More and more places are passing laws forbidding handheld telephone use while driving, a good thing, as far as it goes. Studies have shown that it only helps somewhat — that just having a conversation with someone who isn’t there is sufficiently distracting to be a problem. What we really should be forbidding is doing anything that distracts you from driving, while you’re driving... but that’s too vague for U.S. law. Anyway, what we have is at least of some use.

And so lots of people have installed hands-free setups in their cars. All they have to do is start the call, a small extra distraction, and then they’re off. Add voice calling to that, and we’ve shrunk it to a single button for saying, “I’m going to make a voice call now.”

But it seems to me that we’re missing a major point with all of this, because having controls that are hands-free is only a part of it. We do lots of other things in a car, shifting and signalling and turning on the wipers, and we use our hands for them. The real key is making things eyes-free.

I was thinking, last night as I sat in my car, about the various controls that I use all the time. We’ve put a bunch of them on the steering column for a reason (the headlight and wiper controls weren’t always there) — so we can get at them without moving our hands very far, but more importantly, so we can operate them without looking at them. I can do the following in my car, entirely without having to look in order to do them:

• Shift gears (yes, I have a manual transmission).
• Signal a turn.
• Sound the horn.
• Operate the lights.
• Operate the wipers.
• Operate the cruise control.
• Open and close the windows.
• Lock and unlock the doors.
• Switch the stereo on and off, and change the volume.

That’s pretty good; I was pleasantly surprised to find that most things that need to be done while I’m driving are feasible with such minimal distraction. Then there’s the next tier, things that I need to look at, at least momentarily, in order to do:

• Other operations on the stereo (select a pre-set radio station, switch between radio and CD, and so on).
• Operate the climate controls.
• Reset the trip odometer.
• Operate the emergency flashers.

Now, what I don’t understand is why cars haven’t been at the forefront of voice-control applications. It seems to me that the car is an ideal place to want to control things by voice, and that the number of things you need to do, and, thus the vocabulary the system has to get right to accomplish them, is small.

Why can’t I say things like these, and have them work?:

• “Switch to FM pre-set 2.”
• “Switch to CD number 3.”
• “Turn the air conditioner on.”
• “Warmer... warmer... OK.”
• “Turn the wipers on.”
• “Reset trip odometer 1.”
• “Make a phone call.”

Of course, there are different ways we might say those things, in human language — “Turn the air conditioner on,” might also be, “Run the A-C.” — but I’d be happy to learn the standard phrases that it wants me to use, in order to avoid having to fiddle with things by eye and hand while I’m driving. I’d also be happy to prefix everything with “Car”, or some longer phrase that would tell the system that I’m not just singing along with the radio, talking to my passenger, or calling some other driver a moron.

The biggest problems I can see would be these:

1. Making sure the car knows when I’m talking to it, as above.
2. Making sure the car doesn’t respond to some “unauthorized” voice, such as, say, a mini-human imp in the back seat.
3. Handling very noisy situations.
4. Dealing with “I don’t understand,” situations.
5. Worse, dealing with misunderstandings, like having “Turn the wipers on,” result in the trip odometer’s being reset.

Because of the limited vocabulary, 4 and 5 aren’t serious problems. Even an untrained system can be very accurate in such an environment, and there could be an optional training step, to accommodate drivers with sufficiently strong or strange accents. 3 — and, to some extent, 2 — can be mitigated with a highly directional microphone placed where it only picks up the driver’s voice. And I’ve already mentioned 1. Still, all that said, I think we’d want to limit the voice control to things that wouldn’t crash the car, or some such, if they went awry (so, no steering by voice, for instance).

In the end, though, this hasn’t come out the way I thought it would when I started to write it. There really aren’t that many things that I need to do that require my taking my eyes off the road. They’ve done a very good job of arranging almost everything so that I can operate it blind, leaving nearly all my attention on my driving. It’d still be nice to handle those few remaining things, like changing the radio station or fiddling with the climate control, by voice... but it’s not as big a deal as I’d originally thought it would be.

“Open the pod bay doors, HAL.”

“I’m sorry, Dave. I’m afraid I can’t do that.”