Hacker News Reader

Comment by deevious | Hacker News Reader

I'd say no to that, aside from the reliability and KISS aspects of this, there is one thing cameras won't be able to do properly: parallax. In tight parking spaces or near curbs I will often move around in the car so that I can see obstacles without having to adjust the mirrors.

View on HN

I think that parallax aside, the benefit of cameras is that a much wider angle can be represented. Much like current reverse-cameras in vehicles, the camera can capture wide/ultra-wide angles of 120-170 degrees. This lets you see everything without craning for parallax. Guidelines can be overlaid for curb/parking situations on demand. Additionally, many auto companies are already augmenting their backup cameras with 360-degree stitched top-down views, which are excellent for curb-centering, etc.

You can also curve mirrors for a wider angle, see it on trucks all the time.

Or on the passenger side mirror of any average passenger car made over the past half century or so:

https://en.wikipedia.org/wiki/Objects_in_mirror_are_closer_t...

The point is, switching to cameras has aerodynamic efficiencies. Mirrors are currently a necessary cost that cameras could replace to improve drag coefficients, improve fuel economy/range, etc. OP's concern was that the cameras would not provide sufficient substitution for the mirror's features. I'm arguing that wide field of view solves that problem.

So could another camera in the car that watches the drivers head and adjusts the view based on movements. It could also be better too as you can now be comfortable and not contorted to get a proper view(even better than adjusting the mirror). It could also be disabled to provide a stead view. With a touch screen we could stay focused on a target too.

Camera's provide more options

If you are removing mirrors to improve aerodynamic efficiency that's the same as making a 1 gallon gas tank and saying the car will have better performance because it has a better power to weight ratio.

True in theory, but actual practicality near zero.

We aren't racing formula1 cars on the highway, so a mirror isn't really slowing you down that much and creating drag. Especially when you think about how much surface area it occupies versus the actual car itself.

On the contrary:

”They can account for as much as three to six percent of the total aerodynamic drag of a car. That may not sound like a lot, but it adds up. Over the course of 200,000 miles, that could cost you as much as $2,000. A study of big trucks (with really humongous mirrors) found they can add a whopping 10 percent to drag.“

(0) https://www.cartalk.com/blogs/staff-blog/what-drag-time-get-...

That's accounting for the higher end of the 3-6% drag. Also speed is critical, because the aerodynamic effect is exponential with speed so unless you are traveling above 80mph it's barely impacting anything. Sure if you are on the Autobahn it will be higher at 100mph+ but this is a bit of an optimistic guess on their part.

Also, these car mirrors after 200,000 miles will still work. I don't know if the car "video" mirrors will be operational after 200,000 miles so the cost of replacing them can easily out run the "fuel savings" that you would expect.

Lastly, there is something that car manufacturers lose when they move away from the physical world. The parallax effect is one that others have pointed out as a key benefit that would be lost, the fact that I can get a rain spot on my mirror and still see clearly, but if there is water on my rear parking camera I see nothing is something else that needs to be considered.

Progress is great, but at some point you realize you are actually degrading the experience not improving it. The Tesla has a great big iPad to control the car, but guess what, it is super distracting, because there are no physical knobs to touch so anytime you change a setting you are forced to look down. Because there are no physical knobs your finger can not rest on anything so as you are driving on the road any bump or jolt moves your finger which makes it much harder to hit the key you want. This leads to more distraction.

Just because you can doesn't always mean you should.

This isn't well informed -- Mirrors are a substantial source of increased drag. Something like 5% of the air resistance of a vehicle comes from the outside mirrors -- over the life of a car, that's worth several hundred gallons of gas, and across the population of cars, several billions of dollars in savings to consumers. I fret about the longevity of the technical solutions (mirrors on cars from the 60s work as well today as they did when they were installed) but it could be a huge boon.

Wouldn’t removing the front mirrors cause helmholtz resonance[1], aka annoying “window drums”, when the front window is rolled down? If you’ve ever driven with only the rear windows down you’ll know what I’m talking about.

[1] https://jalopnik.com/why-do-slightly-opened-car-windows-make...

If you're doing that, you have introduced a far larger source of turbulence than the mirrors ever could produce.

If you're doing what? I don't follow

Driving with the windows down at freeway speeds.

> You can also curve mirrors for a wider angle, see it on trucks all the time.

Yes, but with a wide-angle lens camera you can convert to a less distorted view in software. Or stitch together with other views.

I drive buses for a living. Believe me, when reversing some 15 meter monster into a parking slot, I trust and use my mirrors more than I do the very fancy rear-view camera.

A mirror I have a lifetime experience worth of knowing what is, no perceivable extra layer of processing required. A camera, not quite so, although of course it's a nice enough extra to have.

> A mirror I have a lifetime experience worth

Which nicely explains why young people don’t have the hang ups you have.

Sure, mastering the basic controls should be the first step, before ceding control to augmented systems.

However, I have yet to see a mirror that will see behind the back bumper: this has already prevented a host of dangerous situations for me ("person was there, then went somewhere, where the heck are they, oh: for some reason admiring my back bumper and the pretty light that the reverse lights give out").

If "I have a lifetime of experience with this so thats how it should stay" was humanities criteria for accepting new technology we'd still be bashing the neighboring tribes heads in with rocks tied to sticks.

I recently rented a Peugeot SUV in France and the reverse top-down view was awesome - I had to back out of a 50ft (er, 15m) narrow winding driveway and the top-down helped me to do it in a minute instead of 15 min.

A fancy enough camera system can give you a wide FOV and accurate distance indicators to everything on screen.

Parallax is one way we estimate distances, technology can do a lot better than "move head side to side to get a rough idea."

Now getting people used to wide FOV, that may be challenging. Find an FPS that lets you set the FOV to anything you want, set it to 180. After a couple of minutes you can get used to it. Beyond 180 it is hard! But even 190 or so is doable after a bit of practice.

Not sure how the general public would do with that though! (At what point has a high enough % of the general public played FPSs that such a UI could be gotten away with?)

They could parallax though, so that doesn't seem like the best argument.

They could also use complex optics like "plain lens" with a "fish-eye border", or highlight obstacles by boosting colour contrast or adding in non-visual sensor data.

If cameras are that much better we should be getting vehicles that use both (which we do for rear-view but not side-mirrors AFAIK).

> They could parallax though, so that doesn't seem like the best argument.

I've never seen parallax emulated with head tracking that didn't have piss-poor latency issues. Not to mention poorly configured kalman filters that seem to always do too little or too much smoothing. These systems are fine for cute tech demos but don't belong in safety-critical applications involving multiple tons of steel moving at 100+km/h.

Before replacing mirrors in cars, why not start with making a TV "window" that actually works. Prove the technology in a real world application that isn't safety critical, then we can talk. If it's ready for use on highways, then it should be easy to mount a TV on a wall and provide a convincing experience of actually looking out a window.

This. A “transparent” A-pillar (the bit separating your windshield from driver’s side window) would be a good start, and if it worked well, it’s a feature I’d pay extra for.

Absolutely. Big thick A pillars have increased survivability in crashes (embedded airbags, stronger structure) but that doesn't come for free - they cause some accidents (hiding bikers, also cars when pulling out of junctions).

I'd love to see a transparent A pillar but I can imagine the parallax issues (talked about in other comments here) would make them a bit weird.

Volvo made a concept car with glass A pillars and no B pillars.

https://en.wikipedia.org/wiki/Volvo_SCC

That was interesting, thanks. Good to see how much of its safety features are now common (blind spot assist, adaptive cruise control etc.) And, given this thread, amusing that the wikipedia article doesn't pick up on the A pillars.

I've never seen head tracking applications at all, and I'm not surprised that they're underwhelming, but I hope the state of practice gets better because it would be really cool to have an AR setup where metadata is projected onto the windshield and windows such that it lines up with the driver's perspective.

Imagine a thousand foot tall semi-transparent google maps pin, resting lightly on the head of the person you're picking up, visible for hundreds of miles but only to those who are interested.

The sci-fi tech I've wanted for years is a liquid crystal windshield that blacks out a small circle over the sun from the driver's perspective while leaving the rest of the windshield clear. Driving west in the evening or east in the morning, I wager a lot of people have had this idea at some point.

I have no idea if it will ever become feasible, but I sure hope it does.

I've had the exact same thought. I expect it to happen eventually, or some other tech will render the issue moot.

This. Not just the sun, but more annoyingly, opposing headlights in the night. Technologies to avoid blinding opposing drivers are not adequate. We could do with technology to assist those who are on the affected side.

Material sciences could bring a solution:

Transparent material that becomes opaque to bright light (something like gamma < 1, gain < 1) in the direction of the light. So each point of the material is transparent or opaque depending on the viewing angle and respective directed light flow through it.

I think such a material could be engineered.

>If cameras are that much better we should be getting vehicles that use both (which we do for rear-view but not side-mirrors AFAIK).

The 2015 Honda CR-V has a camera under the right mirror. When you turn your right blinker on, the screen switches to that view with an overlay for car lengths. It's kinda handy, but I probably wouldn't pay much more marginally for that feature.

It's about being more likely to choose the CR-V because of the features, rather than paying more, I think.

There's a new Honda 'e', that doesn't have side mirrors at all, and just has two screens inside.

https://www.youtube.com/watch?v=pEKq8jmckz0

Honda has a video specifically on the camera mirror system, and mentions using a wider fov: https://www.youtube.com/watch?v=Urf9JK5Szn8

Honda is awesome! The only other car manufacturer I would buy from beside them is probably Toyota. That being said... Please don't let them touch ANY of the software... The head unit in my car is a UI/UX disaster, if you try to use it in any capacity while driving its definitely worse then just texting and driving. The only saving grace of the Honda head unit is that they finally started implementing Android Auto in them. Also, the Lane Keep Assist and Anti-Collision features leave a lot to be desired.

Yikes. Terribly distracting controls and sounds like significant challenges with safety features (lane keep and anti-collision). What makes Honda "awesome", in your opinion? I'm wondering what is ranking higher than these safety concerns you mention?

I've owned a Plymouth Voyager, Toyota Tercel, Mercedes c280, Mercedes ml320, a Ford Focus, and a Jeep Wrangler in the past. Hands down the Honda is the most reliable and cheap to maintain car of any I have owned(other then perhaps the Toyota Tercel). My Honda Fit was $18,000 new and had features that were only standard on $25,000+ cars. Things like android auto, CVT transmission with flappy paddle shifters, moon roof, lane keep assist and anti-collision. Sure those systems have some problems, but I like still having them. I fully expect this car to last me 250,000+ miles.

I'm just over 100k miles on my 2015 fit, and I see it going about 100k more without any thing more than a tune up and new brakes.

Disappointing that Honda isn't bringing the E to the United States.

I say it is time for a forward sideways-looking camera for exiting corners or sharp crossings. I do not know if this exists but I have missed it a number of times.

That reminds me of a small project where the British Army gave a budget to some soldiers to add "accessories" to a main battle tank - one of the things they did was add sideways looking cameras at the end of the main gun so they could use it to see round corners.

Audi A8 has it: https://youtu.be/TchqhDUbP8w?t=342

It's gonna be a long time before mainstream car adopt this though.

Oh dear. That must be my next car.

Unfortunately, it just cannot be...

Glad it is at least being researched. Thanks for the (very precise) link!

For my eye, Bose's suspension seemed smoother: https://www.youtube.com/watch?v=3KPYIaks1UY

And it had Turbo-Boost (same video with cue-in): https://youtu.be/3KPYIaks1UY?t=81

> a long time

The Ford Edge has had the option since 2015. Looks like the Explorer got it the next year. The Subaru Ascent has it for 2019. I'm sure there are others that didn't come up in the first couple pages of my search results.

Or camera feeds on the road that the car can access, to allow the driver to see around corners ahead of them.

So many problems in such a small package! Security and privacy aside (sic!), I'm not convinced this would help safety: people are notoriously bad at interpreting images from multiple points of view in realtime.

Get a BMW, here's a video from 2012: https://www.youtube.com/watch?v=VfX1plunhAo

My 2013 Volvo can be fitted with exactly these as an original part by the dealer (it seems they were not available with the original order); the module is placed into the Volvo logo, and the driver sees a split image in the infotainment display.

But like the Audi A8 it's not exactly a cheap car; maybe it will be available in more cars in a few years (or once some patent expires?).

Edit: Here is a 52s video/ad for the feature: https://www.youtube.com/watch?v=OkF9Txbeaxg

I would say, what is also lacking are resolution, dynamic range, and latency. Regarding the latter, even if it's a closed circuit analog system with no digital buffering, there is still frame refresh to consider.

At night, if the display isn't dark enough, it will act as a light source, interfering with the driver's night vision. Black has to be black, and dark objects that would be visible with the naked eye via mirror had better be visible with the camera.

I'd assume this isn't really a problem these days? Aren't OLED screen capable of high resolution, dynamic range (include true black) and low latency?

The main issue would probably be cost, but at a technical level having a suitable screen for car mirrors shouldn't be a problem.

Yeah, throwing money at it could realistically bring this from just-around-the-corner FM to Actual Machines. Not quite there for COTS equipment yet, for cost-saving reasons (fast processing needs beefy HW, which is expensive in itself, needs higher and smoother power, cooling, etc.).

The point being that a camera could just film the entire surroundings and show you that picture, instead of exactly what you are able to see through the mirror reflection- i.e. you're constantly seeing your entire blind spot, not just the part the mirror is showing you for your head location.

Most new cars have parking sensors of some kind. Synthetic 360 vision camera systems are starting to become common on luxury cars and even available on some more down-market models...and those provide way more visibility than any mirror ever will.

which also gives you a much better sense of depth too as you move your head.

A mirror reflects a 3D image. An LCD screen isn't 3D.

Not only parallax, mirrors reflections are also 3D.

Reply on news.ycombinator.com