Interface Puzzler #1 Answer

Puzzler

Thanks to everyone who wrote in and gave solutions to the first Humanized Puzzler. There have been many more responses that we anticipated! I loved the discussion of the problem. Although I had meant for people to email the solutions privately instead of discussing publicly, in retrospect, the discussion was more valuable than the secrecy. The next puzzler will all be discussion.

I’d also like to apologize for taking so long in posting the solution. With the release of Enso, we’ve been very busy.

In short, the puzzler asked, “Can you design a car that isn’t forward/reverse modal?” For those who didn’t read the original post, check out the full question.

Few people were fooled by my implication that a solution was impossible. With modern automatic cars, almost any conceivable behavior is possible for shifting because the gear selector is simply an electronic switch physically decoupled from the transmission. The trick is choosing a good behavior.

Here’s an overview of some of the refreshingly varied solutions.

Two Pedals

Rather than having a gear selector and two pedals, there would be only two pedals: one for forward; and one for reverse. The car is assumed to have an automatic gearbox.Pressing the pedal for the opposite direction, or both pedals together would act as the brake. Touching neither would mean that the car continues to coast, in the same way that cars do now. We would assume that the driver gets enough feedback from the motion of the car that she can determine which direction the car is moving in, at any point in time. — John SutherlandThis is an inventive solution, one that hadn’t occurred to us at Humanized. It solves the problem handily. On the other hand, to use it properly would require more than a modicum of retraining. This presents a rather large hurdle: marketing it would be as hard as selling basashi to a vegan.

Another problem arises: there are many inputs that give the same output. Because the speed of acceleration and braking is determined by the relative position of the two pedals, moving the pedals in tandem yields many positions for the same amount of accelerating/braking. Much more importantly, having both pedals act as brake requires much more dexterity than a single pedal (and introduces accessibility issues). And, a great deal of care would have to be taken when releasing the two pedals, otherwise the car would go skittering off accidentally.

All in all, an inventive but probably not practical solution.

Quasimodal

I have one suggestion: a quasimodal steering wheel. Two opposite portions of the steering wheel could be [pulled] back, or have two buttons that need to be pushed to make the car go backwards. — PianohackerThis is a slightly more elaborate version of the the simplest solution to the modality problem: a single button that, while held, causes the car to move backwards when gas is applied via the standard pedal. Once the car is rolling backwards, there is no need for the button to continue being pressed.

Why doesn’t this constitute a mode? We’ve created a state where, after the car is in reverse and the button has been released, there is no direct kinesthetic feedback. Shouldn’t it be considered a mode? Jef’s The Humane Interface has the answer:

Modes are, in a first approximation, situations where a particular gesture is given different interpretations when the system being used is in different states. When you are unaware of system state (i.e . your locus of attention is not the system state) your use of a gesture may have an unintended effect. This is a mode error. After consciously putting the car into reverse, your locus of attention is on the system state: you are actively driving the car backwards. The system state is guaranteed to be your locus of attention because of the quasimodal entry and then your subsequent concentration on the act of driving.

This solution obviates the need for a “Reverse” entry on the gear-shift.

The two-button coincidence proposed by Pianohacker is a bit overkill. The only time reverse is needed is when the car is already stopped, and it is enough for one button to be pushed. And Henry Blum pointed out a problem with even a single quasimode button:

If this button is on the wheel (rather than on a stick shifter behind the wheel seen on some automatics, or some other location) then you come across the difficulty of holding the button while turning the wheel around.Pianohacker, however, had a solution. He mentioned that the steering wheel could be pulled back to engage the reverse quasimode. It can be used with one hand or two hands, and doesn’t keep (or force) the driver from turning to look back. It’s a nice natural mapping that would yield a wonderful feel to reversing. And, to top it off, it would be faster to use than a gear-shift. I could easily see this solution catching on in sportier cars.

Joystick

Ever since I was a teenager struggling to learn to drive, I have always wished for a car which you controlled with a self-centering analog joystick. Just push with varying pressure to go faster in any direction. Centering the joystick stops.
In fact, I never really got the hang of it, and now I only ride motorcycles. I never realized that it’s because other than the gear shift (which on a motorcycle with no reverse gear, are all really sub-selections of the same mode), there are no modes! — David Young
Like David, I have also always wished for a joystick control for my car. Push forward to go forward. Pull back to slow down and stop. Release the stick and pull back again puts the car into reverse (this way you can stop without worrying about accelerating into the car behind you). Left and right turns left and right. Learning would be almost instantaneous: the control scheme is an elegant natural mapping. No need to shift the seat far forward for those of blessed with shortness, or far back for those cursed with long legs. To top it off, it allows for sleekly clean design of the cabin as well as the removal of the steering column (one of the more dangerous parts of the cabin).

Andrew Clarke makes a valid complaint:
One of the reasons the accelerator in a car is at the driver’s feet is that it’s by far the safest place for it. If you’ve ever driven with a small dog and it jumped down under your legs while you were moving at 60 mph, just remember the sudden panic you felt. At that moment you realize that little alcove is sacred. Bringing that space further up into the cabin allows for a whole new array of potential disasters. When the difference between 10 mph and 100 mph is a slight amount of pressure, you simply can’t allow anything near that accelerator but the body part you use to control it.

Safety comes first. But the current car control scheme already has serious drawbacks: anyone who has had a fast-food drink roll underneath their feet knows that the foot-alcove isn’t as safe as it could be. It’s hard to fix, but with a joystick a simple clear protective dome with a hand-entry port would do the trick.

Unfortunately, this is radical enough that it will probably not get adopted any time soon. But for now, as Darrel pointed out, you can see it and an omni-directional driving system here. Its complicated behavior is controlled with a simple joystick.

Head Turn Interface

A couple people gave solutions that involved using a swivel chair to have the car simply go in the direction you are facing. The solution, while imaginative, is not practical without a major redesign of cars. And even then, unless the swivel was very quick (which might be physically discomforting) it would be a laborious process to do the forward-backward-forward dance of parallel park.

If you’re in neutral and you hit the gas pedal while looking backwards, you go backwards. If you were looking forward, then you go forward. Once you’re in motion, the direction you look has no effect. — Paul Ingemi Although Paul said the example is modal, it isn’t: the kinesthetic feedback of looking means that you will be conscious of the system state. In particular, it will be the same state as your head direction.

This would be a marketable solution, but like many marketable solutions, it has some inhibitive problems. First, when parallel parking, it would be trying to constantly shift your head. Second, where you are looking is not always consciously controllable: a flash, loud noise, or other stimulus can cause you to involuntary your head towards the source. Third, you may not want to look at something else (like an approaching car) while you begin reversing. An interface that is too clever often does the wrong thing. An interface that is too clever can be quite dumb.

Teeter-tot Pedal

Keep the steering wheel. But change the gas [pedal] to swivel in forward and reverse (like a teeter totter). Tilt it forward to go forward. Reverse to go reverse. Have two gears (Park and Go). Keep the break peddle too. I had a Toro riding lawnmower like this once and it was great. — David JustThis is a good solution. Pianohacker mentioned that he, “[remembers] something like [that] from a human factors class [he] took in design school. They [said] that the gas pedal sometimes is confused with the break pedal, especially with elders. A solution was one pedal; tilting forward (your toes) moved the car forward, tilting it back (your heel) was the brake.”

Eddie was quick to give an ergonomic counter. He asked, what “about the ’sudden reverse’ that could spring up with a teeter?”. Andrew Clarke drove the point home, “The tilting gas pedal model has a …problem. Unless you’re a woman used to wearing high heels, pushing the pedal down at a forward angle would be agony on a long trip. Currently the weight of your leg does all the work. Anything that doesn’t allow the driver to relax for most of the trip decreases enjoyment and increases fatigue and risk.”

These problems are not insurmountable. With enough testing and tweaking, a bet a good ergonomic compromise could be reached.

Push-Pull Pedal

Roy Osherove proposed my favorite solution. It solves the accidental rocking problem, the fatigue problem, and the retraining problem. Unfortunately, it was after the deadline.

In this solution, the brake pedal remains the same, and the acceleration pedal remains mostly the same. Pressing down causes the car to go forward, but hooking your foot underneath the pedal and pulling causes the car to go backward. Natural mapping is retained while not breaking current habits. No new controls or buttons are needed, there are no new edge cases to deal with, its as fast as a gear-shift to use (or even faster), and can be explained in an instant. On top of that, reverse will never accidentally get engaged. It’s the sort of evolutionary change that might actually get adopted.

A couple people proposed adding an extra, smaller pedal that just accelerates backwards. While this certainly works, I think a push-pull pedal is a more elegant solution to the problem. Push-pull is naturally mapped and doesn’t add extra clutter to the cab.

The Winner

Going over all of the solutions, it is very hard to figure out to whom we should send the dubious honor of a Humanized shirt. David Young gave an excellent description of the wonderful joystick solution, so he is the winner of the Humanized Puzzler #1!

Thanks again for everyone that took part in the puzzler. Puzzler #2 will be out soon, so stay tuned.

Modes are, in a first approximation, situations where a particular gesture is given different interpretations when the system being used is in different states. When you are unaware of system state (i.e . your locus of attention is not the system state) your use of a gesture may have an unintended effect.

How is pushing and releasing a button not a mode? There is no guarantee that the driver will be focused on system state the entire time (kids, dogs, pedestrians, etc).

Additionally, as written, mode can switch without intentional user interaction. For example, I can be backing out of my driveway and have to stop for traffic; Car switches modes. I can be backing up a hill and take my foot off the gas; car switches modes. Maybe the system is smart enough to avoid the second case. In either case, unintentional mode switching is the true evil of modes.

Otherwise, enjoyed the writeup. I’ll definitely participate the next go ’round. Any of you humanized gents making it to PyCon?

“Release the stick and pull back again puts the car into reverse (this way you can stop without worrying about accelerating into the car behind you).”

Er … this is a mode, then? So not a non-modal solution.

What’s more, the push-pull pedal is downright unsafe — if the driver’s foot is tucked under the drive pedal and he needs to suddenly brake — most common in parallel parking — his foot has to negotiate out and over to the brake, which will take appreciably longer.

All of the solutions here are serious over-engineering, and either don’t solve the “problem” (joystick is either still modal, or applying the full brakes will send you straight into reverse after stopping) or are dangerous in practice.

Thank you so much for writing a full, well thought out response to my comment! Reading this post made my day.
Can’t wait until the next Puzzler.

First, while the automatic transmission control is physically decoupled from the transmission, that’s kind of misleading. I DARE you to shift into reverse at 60 mph. Yes, cars usually have locks so you don’t do this, and you can set it so that the car doesn’t switch into reverse unless it is under a certain speed (cars do this already, I think), but it’s worth remembering.

Personally, I think that none of the solutions will work as well as cars currently do. People just don’t confuse modes in their car that often. Here’s some problems that you didn’t mention:

Quasimodal: The pull-back steering wheel has a drawback: think about how you turn your car - hand over hand. That means that some drivers are going to let go of the wheel every so slightly when turning, and that means the quasimode goes away. The other problem is that it could prevent gangsta-style crusin’. Any quasimodal car interface causes some loss of steering control.

Head Turn Interface: Doesn’t he ever check his blind spot? Or yell at his kids in the back seat?

Joystick/Teeter-tot: A problem with these two is the lack of emergency stopping response that isn’t necessary in things like lawn mowers or forklifts, but is necessary at 60 mph. One could fix the joystick design by making “back” “stop”, and “back while holding the button on top” “reverse”. You could fix both by adding a brake pedal.

But that brings me to the problem with the joystick, teeter-tot, and push/pull design, and why the modal system works well. When was the last time you parallel parked? Did you hit the gas much? When a car goes into drive or reverse mode, it sets the default speed to +/- epsilon mph. This allows you to navigate easily in tight quarters — and 80% of the time you use “reverse” in tight quarters. The other solutions require you to be a master of subtlety when driving close to other cars, and this is not easy. In fact, video games often require the user to be this delicate as a *challenge*.

A very excellent thought experiment, though. I still think that there are places where modes work the best, but it’s definitely worth questioning our basic assumptions about them. Nice work!

I’m a little skeptical about the push/pull model as well for the reason Michael gave (lag time in moving the foot to the brake), but I think I might not be picturing it clearly.

If the pedal was lifted with just the toe area, rather than the foot being inserted entirely, it might work beautifully.

The high-heeled women might like the rocking pedal but I wonder if they’d take issue with the pulling pedal. It seems like they’d have significantly less leverage for such a maneuver, though I’ve never worn heels so I could be totally off. :)

The underside of the pedal would also have to be very soft and rubbery or people with expensive dress shoes would be angered to find them all scraped up after a month of driving. People wearing sandals would have it even worse. Currently we only drive with the part of the shoe designed for protection and abuse so this is an easy detail to overlook. Drive barefoot sometime and you’ll see what I mean.

This has been a very interesting exercise and I’m eager to read the next one. Kudos, everyone.

The clearest and cleanest way to do this would be to make the car never have to go in reverse at all. Allow the car to rotate in place by having the wheels all be able to rotate 360 degrees around the up axis.
Admittedly, for maximum flexibility, you’d want the car to take up the same amount of room from front to back as from side to side, in a probably circular shape so that it’s assured of being able to turn around regardless of what is around it.
Driving the car could then be very much the same as it is today, only when you want to go in reverse, you just reduce your speed, preferably stop rotate and then go forward in the new direction. This could also help greatly with parallel parking at which I’m terrible.

This may go beyond the original intent of challenging the interface of the car, but sometimes instead of patching a bad system with a clever interface, it’s best to change the system.
~Israel

May I be so bold to chime in and state the obvious — that ALL of these solutions are equally poor.

Maybe I’m just old fashioned. Vroom!

The “two pedals” solution is clearly the best. Joysticks work very well for steering, but I wouldn’t want to have to brake (which pushes yourself _forwards_ in relation to the car) by pulling a stick _backwards_. This seems disturbingly hazardous. Also, the analysis provided by the sender-inner of the joystick suggestion is meh.

Anyways, if anyone here really deserves a t-shirt, it’s Israel. Keeping such a straight face while suggesting round cars — genious!

The braking issue with regard to hooking your foot under the accelerator and pulling it back for reverse is not an issue as this is done with your right foot leaving the left free for the break. An excellent solution.

Hey now, give credit where credit is due :P I submitted the push/pull solution a couple days after you posted the puzzler, Aza.

Regardless, I mentioned in that email that they already exist. Our tractor uses it actually.

There’s no lag time in this method because you have to brake to a stop first before you can go in reverse! So when you’re done braking, simply move your foot back to the right, only slip it under the pedal this time to go backwards. Same happens to go forward again. No lost time, very slight movements, quick, and almost completely fool-proof.

Cheers.

Cori,
I also submitted a solution that did not get a mention. I hope they got the emails. Still I guess if they got more than they were expecting maybe they could only mention the best.

Back to the drawing board

I don’t think the Joystick is a good idea. There were already some Concept Cars with this kind of control and from what I understand they made clear that it is a bad idea.

It’s not so much about acceleration and breaking but about steering. The Steering Wheel is a very good interface for controlling a car because it allows input at a great resolution and wide range: I can turn the wheel only slightly on the freeway or turn in a few times on a sharp corner in the city. A joystick isn’t suitable for this kind of task.

There is a simple test: get a racing game and pretend to do a tree point parking with the analouge sticks of a normal game controler. You will see that it is impossible but you don’t notice it since this kind of input isn’t needed in the high speed driving normally required in a driving game.

Also, there is one more thing: a steering wheel has a mechanical connection to the wheels so if the electronic system of the car breaks down, you are still able to control it. A joystick couldn’t have this kind of connection and would require and drive-by-wire system which is quite dangerous - in case of a malfuction you loose complete control over you car.

As an intresting fact, racing cars are eqipped with a sequential gearbox, which is even more modal then the “normal” gearbox. In a sequential gearbox, you can only “shift up” and “shift down”. Racing car drivers have to constantly keep track of the gear they are currently in.

Regarding the head turn interface, I considered potential distractions like a flash, but ultimately if you’re distracted and decide to accelerate at the same time, you’re already screwed. At least you’re looking in the direction you’re going. What I didn’t consider is the difficulty of parallel parking. That’s a great point. Oops.

Regarding the problem of finer steering movement being required at higher speeds for the joystick solution, an on board computer chip could monitor the vehicles speed and adjust the amount of steering accordingly. Fly by wire systems work perfectly well for modern aircraft.

Also, there is one more thing: a steering wheel has a mechanical connection to the wheels so if the electronic system of the car breaks down, you are still able to control it. A joystick couldn’t have this kind of connection and would require and drive-by-wire system which is quite dangerous - in case of a malfuction you loose complete control over you car.

I was a bit startled that this was the only post to mention this: It was on my mind the entire time I read through the various joystick-style suggestions.

It’s extremely important that a car provide a direct, mechanical interface, since there will always be cases where the electrical systems fail.

There will always be cases where the mechanical systems fail, too, but hiding them from the user doesn’t really solve that problem. :)

Good morning, I am new to this site. I have just learned about this site. I am going to read on and it’s very interesting to know

oc91cz449i8enhc0