![]() 07/19/2018 at 08:17 • Filed to: None | ![]() | ![]() |
Obviously, the era of self-driving cars is imminent. One of the biggest questions surrounding that is how it will affect people who can’t afford AVs.
Right now in 2018, I have a car with no cruise control. I can go on eBay and buy an aftermarket unit that I can install (or pay someone else to install, yeah right) and hav e cruise control. I hypothesize that in twenty years, people who want an a utonomous vehicle but drive a shitty twenty year old car will be able to go online and buy a computer/sensor suite that can be installed to make their vehicle autonomous. Most cars you buy today are all throttle/steer by wire, so you wouldn’t even need to put motors on the steering column or throttle.
What are your thoughts, Oppo? How likely is this?
![]() 07/19/2018 at 08:31 |
|
![]() 07/19/2018 at 08:36 |
|
No.
As a software engineer the extreme number of fringe cases that supporting more than one model present and also the liability make it pretty impossible. Like you need more than cameras and hardware to push and turn everything, you’d need to completely rewrite the software for every car because of how differently every car reacts, the different sightlines etc etc. it’s not like cruise control where it just holds a speed.
![]() 07/19/2018 at 08:42 |
|
I agree. It's far more complicated (and therefore expensive, to the point that it might not make sense to spend that much on an old car). I'd also wonder about safety regulations. How would you certify a unit to work with any car?
![]() 07/19/2018 at 08:42 |
|
If such a system were to exist, it would certainly need to have model/year specific controllers (I can’t cram cruise control from a Camry into my Golf, so that was never an expectation). But I really don’t see why it would be impossible. It probably wouldn’t be impossible to have at least low level autonomy (adaptive cruise control, lane keeping) installed in the aftermarket- in a way, that’s kind of what nascent systems are, if you look at the stuff that’s been retrofitted to Volvos by Uber, for example.
07/19/2018 at 08:43 |
|
Obviously, the era of self-driving cars is imminent.
Just like fusion, flying cars, and Half -Life 3.
![]() 07/19/2018 at 08:45 |
|
Yip, exactly this.
Also, installing and calibrating the sensors is a long and finicky process. There would need to be defined install locations for every make/model. It's not something for the aftermarket. The cost for a system that accommodates this would be more than a used car.
![]() 07/19/2018 at 08:46 |
|
Unlike those three, pretty much every automaker is testing vehicles at various levels of autonomy on public roads.
HL3 isn’t gonna happen tho
![]() 07/19/2018 at 08:49 |
|
It already exists; it’s called a chauffeur.
As for self driving things, I think that the kind of modular system you want would be very hard to make economically with the current tech.
![]() 07/19/2018 at 08:49 |
|
This seems to be more of an adaptive CC, but people are already prototyping it.
![]() 07/19/2018 at 08:54 |
|
With current technology, I don’t think it can happen. One major limitation would be the power consumption, but I’m thinking 15, 20 years from now. By then we’ll have much better processors and batteries. A modular system wouldn’t really make a lot of sense, as others have said, but one specialized to a model might.
![]() 07/19/2018 at 08:57 |
|
I think it would be too expensive to make sense. You’d be better off just buying an old can with self driving functionality.
I doubt it would be more than a niche thing until you have roads where self-driving is required (which seems unlikely to happen for at least 20 years after self-driving becomes ubiquitous in low-end cars, so quite a ways out, even if someone has a good system available tomorrow). Once it is a requirement, there would be greater pressure to upgrade, but you might also just see self-driving flatbeds that you can rent if you need to take your classic car through a self-driving only area (self-driving has the most benefits in cities, but I expect you see it first on limited access highways, since there’s usually an alternate route a non-autonomous vehicle could take.)
![]() 07/19/2018 at 08:59 |
|
It’s not impossible, but it’s extremely unlikely. Cost will be the biggest factor.
Assuming that you could by all of the stuff to retrofit self-driving capabilities to an older car, the cost and complexity of install would make it more expensive than just buying a new car.
![]() 07/19/2018 at 09:00 |
|
I mean, in such a timeframe it’d be hard to discuss it. By then self driving tech might even be mandatory in cities, so there could be a space to retrofit older cars, on the other hand, the sheer cost would prove to be difficult if not imposs to work around. Also, if thw average age of cars stays the same, by 2050 average cars might be MY2039; which means we cannot know what they might have installed! Maybe a more advanced OBD, maybe digital steering,very probably battery or hydrogen motor it’s impossible to tell.
![]() 07/19/2018 at 09:02 |
|
I don't think it will be impossible, just cost prohibitive for most people.
![]() 07/19/2018 at 09:13 |
|
Probably true. I think a price point of ~$500 would be about the top of the market.
![]() 07/19/2018 at 09:20 |
|
Will never happen. I work pretty extensively on autonomous vehicles for my day job and it is way more complicated than most people think, and I think the technology is farther off than people think in general. The sheer amount of support systems and wiring you would need to integrate into the vehicle ignoring the physical sensor locations, power consumption and calibration is staggering.
![]() 07/19/2018 at 09:35 |
|
Yup, I thought about saying something about the massive power consumption, but I figure that in 20 years, power generation/storage will be much better, as will the power consumption of the hardware. I don’t think there will ever be an economical level 5 retrofit system, but maybe an adaptive cruise control/lane keeping system?
![]() 07/19/2018 at 10:39 |
|
Adaptive cruise control and basic lane keeping are pretty straightforward and non-model specific so that is very doable . Honestly with an Arduino, a laser distance sensor of some sort and some little motors to tap the controls around I think I could make a non-adaptive cruise control into a simple adaptive one pretty easily.
The hard part of a fully autonomous vehicle is every single possible emergency scenario where it needs to make decisions based on the exact situation(is there oncoming traffic? How far away? How fast is it approaching? Are there cars behind? O ther cars in front? P edestrians on the sidewalk? P edestrians in the crosswalk? P edestrians where they aren’t supposed to be? This creates millions of potential scenarios to be taken into account ) as well the particular vehicle’s capabilities, in the same situation an F150 may need to slam on the brakes and hope for the best while an M3 might be better off taking evasive action. Cruise control/ lane keeping don’t need to make any choices besides ‘object in front is going 2 meters/second slower, we will slow down by 2 m/s. Object in front is now gone, we will resume normal speed ’ or ‘white line getting closer. Edge away from white line.’ The driver is still expected to be paying attention to make the tough decisions .
Also I’m no lawyer but I don’t even want to imagine how much insurance you need for a product like an aftermarket fully autonomous setup.
![]() 07/19/2018 at 11:21 |
|
The hard part is package protecting for the additional harnesses and especially playing nice with whatever controller network you are retrofitting to. One of the big struggles with even level 5 is coming up with sufficient cleaning systems and environmental mitigation for things like rain, snow or roads that get salted. Now we look at air or fluid based cleaning and all that has to be packaged as well.
07/19/2018 at 11:58 |
|
Testing yes , but only at level II, maybe III if I’m being generous. I t’s going to be a long time, decades perhaps, before a true level IV or V car is available.