Nearly half of Americans don't want a self-driving car

[Bill]

View attachment 67777

Computer, 1965:

Computer, 50 years later:

It will happen.

 

[tspear]

@flyingcheesehead

Not saying it is impossible, but IT has trouble supporting stuff from five years ago. Ten is a stretch. This is a fundamental change where you need to either support it for twenty years or require people to upgrade the HW on a regular basis.

Sent from my SM-J737T using Tapatalk

 

[azblackbird]

Interstate Highways in particular have a very well fleshed out set of standards, which makes them relatively easy in the grand scheme of things, which is why Tesla et al have gotten systems working there first for the most part.

Yep... for the most part our interstate system is already somewhat autonomous friendly. It's the individual cities and suburbs that are going to require the upgrades in a uniform and universally adoptable infrastructure. This is where it can get prohibitively expensive.

 

[Juliet Hotel]

@flyingcheesehead

Not saying it is impossible, but IT has trouble supporting stuff from five years ago. Ten is a stretch. This is a fundamental change where you need to either support it for twenty years or require people to upgrade the HW on a regular basis.

Sent from my SM-J737T using Tapatalk

Personal computers and their software are not the same as the systems you'd find in proprietary hardware like a car. There are Boeings flying just fine today with 20 year old processors. There are trucking fleets still rolling down the highways with 20 year old telemetrics hardware. I would bet you can still find ATM's out there with antique hardware.

 

[tspear]

Personal computers and their software are not the same as the systems you'd find in proprietary hardware like a car. There are Boeings flying just fine today with 20 year old processors. There are trucking fleets still rolling down the highways with 20 year old telemetrics hardware. I would bet you can still find ATM's out there with antique hardware.

That Boeing is not connected. Neither is that truck.
As part of my career I developed software which managed building control systems. The top of the line stuff which was network aware not only had crappy security it was also generally on a five year replacement plan. Basically the HVAC, access control and other systems which last for twenty to thirty years in theory are very stupid and have zero security. You then create a module with the smarts which sits in front of the system.
The problem is the module gets slower as the software requirements grow, as cybersecurity gets more complicated. These smart boxes are now lasting about five years.
For the AV you need much longer.

Sent from my SM-J737T using Tapatalk

 

[Palmpilot]

Interesting article from MIT Technology Review:

The Seven Deadly Sins of AI Predictions

 

[Palmpilot]

15 Predictions for the Future From 2005 That Turned out to be Totally Wrong

 

[Juliet Hotel]

That Boeing is not connected. Neither is that truck.

Umm... The truck most definitely is connected. That's exactly what telemetrics are and exactly what they do. Its how I know exactly where all my trucks are and what they're doing at all times. They connect directly to the vehicles J1939 buss or 1708 or 1587 on older trucks and transmit that data over the cellular network to a server.

 

[tspear]

Umm... The truck most definitely is connected. That's exactly what telemetrics are and exactly what they do. Its how I know exactly where all my trucks are and what they're doing at all times. They connect directly to the vehicles J1939 buss or 1708 or 1587 on older trucks and transmit that data over the cellular network to a server.

I should have been more specific. Tesla directly connects the core computer to the network.
All the truck systems I dealt with for an energy management study (looking at hybrids and EV trucks), the telematics were like the smart systems we used in buildings. The telematics are upgradable and replaceable units which interface to the trucks systems. They are also smaller, limited HW and cheaper.

For the EV, none of the rather limited information available looks like they are addressing this. Instead like Tesla they have a single monolithic system.

Sent from my SM-J737T using Tapatalk

 

[patrick wentworth]

As for core computer being connected to the network. If there is no connection (you can turn it off) the car still operates under normal control. So the question is the mother ship directing the car, or is it the onboard data and onboard logic? I believe it is on board but actually not sure if that will be the case at Level 4-5.

Understand that the updates are coming from a monolithic source and are of course at risk. The feedback loop through the onboard "bug report" is quite robust and spend a day on their forums --- they do not hold back

As for upgrades to hardware, the car is a big phone with a motherboard that is engineered to be replaceable. I would fully expect to have that replaced if needed.
Am I a Fanboy (way to old to be a fanboy) no, but this car and the concept is heading toward easy maintenance, unlike other cars I have owned.

It is an experiment and fun to be a part of it. Kind of like Ipads in the airplane and Ads-b showing me who is in the neighborhood. Tailbeacon yes, Zaon old and a good first cut, TCAS too expensive for us little birds.

We are in for a wild ride -- did you happen to notice how many satellites SpaceX put up as add ons to regular launches.. Communication limitations will be less and less an issue and the car to car data flow may well handled in ways we have yet to see.

 

[flyingcheesehead]

Yep... for the most part our interstate system is already somewhat autonomous friendly. It's the individual cities and suburbs that are going to require the upgrades in a uniform and universally adoptable infrastructure. This is where it can get prohibitively expensive.

It's really no different. Yes, some lanes are narrower, but a double yellow line always means the same thing, a stop sign always looks the same, etc.

Really, the hard part is going to be getting a self-driving car to use a dirt "road" in the middle of nowhere, or driving with a fresh snowfall where it's difficult to delineate where the road is. City streets aren't that hard.

As for core computer being connected to the network. If there is no connection (you can turn it off) the car still operates under normal control. So the question is the mother ship directing the car, or is it the onboard data and onboard logic? I believe it is on board but actually not sure if that will be the case at Level 4-5.

I think it'll still have to be on board at level 4/5, with additional "smarts" being downloaded as able. Otherwise, well, "Signal lost"... splat.

 

[Dav8or]

You're partially right. What we will see in the future are entire newly built suburbs and neighborhoods that are autonomous friendly. As far as re-building entire inner city infrastructures. Not going to happen! Way to expensive and the taxpayers won't stand for it.

I'm not sure what you think needs to be rebuilt? The taxpayers are almost irrelevant now. Trillions get spent every year and nobody gets too upset. We spend all kinds of money in the name of "public good".

 

[UngaWunga]

It'll be fun, when everyone else is in a self-driving car, to whip through traffic on a motorcycle or a 1960's Alfa. It won't be terribly difficult to make a transmitter that will trigger the sensors on the self-driving cars so you can make them bake to cut them off in traffic.

 

[azblackbird]

Really, the hard part is going to be getting a self-driving car to use a dirt "road" in the middle of nowhere, or driving with a fresh snowfall where it's difficult to delineate where the road is. City streets aren't that hard.

So it never snows on city streets? I guarantee there's going to have to be some sort of magnetically activated lane striping or imbedded sensors that the cars can follow or they'll be over the road after a fresh snow fall. Same goes for the interstates. GPS systems only so accurate and can vary by as much as 10 ft...so that technology gets thrown out the window right away if you expect a "smart car" to stay in it's own lane after a fresh snow fall.

 

[tspear]

So it never snows on city streets? I guarantee there's going to have to be some sort of magnetically activated lane striping or imbedded sensors that the cars can follow or they'll be over the road after a fresh snow fall. Same goes for the interstates. GPS systems only so accurate and can vary by as much as 10 ft...so that technology gets thrown out the window right away if you expect a "smart car" to stay in it's own lane after a fresh snow fall.

Actually camera sensors, radar, street maps, GPS.... all of these tools exceed the human eye and brain in general.
Where computers fail is combing the information and determining the cues required to make the determination of where the edge of the road is, or where a rough approximation of the road is. Considering the number of miles of roads in the USA alone, it is cheaper to solve this problem than to change the infrastructure.

Tim

 

[Salty]

Computer, 1965:

Computer, 50 years later:

It will happen.

Most popular mousetrap 50 years ago.

Most popular mousetrap today

It won’t happen.


Seriously though. Just because one thing happens it is not logical to assert that every other thing will eventually happen.

 

[Juliet Hotel]

Seriously though. Just because one thing happens it is not logical to assert that every other thing will eventually happen.

Just because you disagree, doesn't mean we're automatically asserting that every other thing will eventually happen. But yes, autonomous vehicles will happen. Soon.

 

[Brad Smith]

We have a 2017 SUBARU with "eyesight" technology which does not operate better than the human eye. For example, it will will apply FULL braking if it's "sees" a plastic shopping bar floating in the path of the car. It does not care if somebody is following closely. Same reaction to a small tumbleweed. In responding to one situation it a creates a more dangerous situation. No, it is not better than the human eye.
Also, somebody mentioned that GPS has a 10' accuracy. Not going to work very well where I live where some dirt roads aren't much more than 10' with steep drop-offs. With no sensors or lines to guide it, will it veer back and forth 5' in each direction looking for the GPS boundaries? Not yet ready for prime time.

 

[Salty]

Just because you disagree, doesn't mean we're automatically asserting that every other thing will eventually happen. But yes, autonomous vehicles will happen. Soon.

Autonomous vehicles already exist. So do better mouse traps.

 

[tspear]

We have a 2017 SUBARU with "eyesight" technology which does not operate better than the human eye. For example, it will will apply FULL braking if it's "sees" a plastic shopping bar floating in the path of the car. It does not care if somebody is following closely. Same reaction to a small tumbleweed. In responding to one situation it a creates a more dangerous situation. No, it is not better than the human eye.
Also, somebody mentioned that GPS has a 10' accuracy. Not going to work very well where I live where some dirt roads aren't much more than 10' with steep drop-offs. With no sensors or lines to guide it, will it veer back and forth 5' in each direction looking for the GPS boundaries? Not yet ready for prime time.

Actually, all your examples are not sight problems. They are decision problems.
GPS would be just one of many data sources, and I doubt they would accept such an in-precise version of GPS as the only data point.

Tim

 

[Palmpilot]

Just because you disagree, doesn't mean we're automatically asserting that every other thing will eventually happen. But yes, autonomous vehicles will happen. Soon.

Just because you say it doesn't make it so.

I don't think anyone knows whether autonomous vehicles will turn out to be practical as the main means of transportation.

 

[flyingcheesehead]

So it never snows on city streets? I guarantee there's going to have to be some sort of magnetically activated lane striping or imbedded sensors that the cars can follow or they'll be over the road after a fresh snow fall. Same goes for the interstates. GPS systems only so accurate and can vary by as much as 10 ft...so that technology gets thrown out the window right away if you expect a "smart car" to stay in it's own lane after a fresh snow fall.

It does snow on city streets, and it will still snow on them after you rebuild them. But it will NEVER be economically viable to rebuild the street I live on to accommodate an autonomous car. They built it once and they'll chipseal it until the end of time.

So, there needs to be some other way of detecting where the road is. GPS-like systems with better accuracy will help, as will HD mapping (which will sense things like mailboxes, signs, etc on the side of roads) to allow the car to drive where the road is, even if it can't be seen. But, you have to keep in mind that humans have just our five senses and our memory, and we're only really using a couple of those senses in this scenario. We can see, though the road may be much more difficult to see we can process the previously-known "shape" of the road relative to things that are not the road surface, like those mailboxes, signs, etc. A computer can do that, too... And that is how autonomous cars are being designed now. And they'll have the advantage of GPS (more likely, Galileo or another GNSS that's more accurate than GPS).

The last time I know of anyone talking seriously about having autonomous vehicles that required special roadways was in the 1990s when I was an engineering student and very interested in the idea. Now, it's laughable.

 

[azblackbird]

Where computers fail is combing the information and determining the cues required to make the determination of where the edge of the road is, or where a rough approximation of the road is.

Exactly what I've been saying! That's why for full autonomy you will have to have so-called "magic rails" to keep all these cars on track. In adverse weather, or bad road conditions where the lane striping has the potential to be hidden/covered up, that solution is embedded sensors or magnetic lane striping.

Here's a good article on the subject.

 

[Palmpilot]

Based on the analysis below of data from Waymo and Uber, the number of autonomously driven miles would have to increase to about 14 times what it is now, without a single fatality, in order to equal the record of human drivers. And the technology is yet to be tested in the full range of weather and road conditions that human drivers currently face.

That's why I say that so far, no one knows whether this technology will deliver on the promises that are being made for it.

When the statistics for the variety of driving conditions encountered in the real world turn in favor of autonomous vehicles, then we will know. Until then:

Here’s the data (from https://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_statistics/html/table_01_35.html)

Self driving cars have been on the roads, what, maybe 2-3 years? I’ll give it the benefit of the doubt and use data for 2 years only from the above. I’ll include the total highway miles for 2014 and 2015. I doubt this includes in town driving or short trips, but it will serve to illustrate the issue.

I’ll use scientific notation NeM which means N x 10 raised to the M power.

In 2014 and 2015, total highway miles driven = 6e12 (6 million million). In those 2 years, there were approximately 70,000 (7e4) vehicle fatalities. ( https://en.m.wikipedia.org/wiki/Motor_vehicle_fatality_rate_in_U.S._by_year)

So, deaths per mile for non automonous vehicles = 7e4/6e12 = 1.2e-8.

This reference https://medium.com/waymo/waymo-reaches-5-million-self-driven-miles-61fba590fafe has Waymo’s numbers for miles driven. I’ve seen estimates for uber of 1 million. So Waymo (Google) + Uber = 6e6

Deaths per mile for autonomous vehicles = 1 / 6e6 = 1.7e-7

Odds of death in self-driving / non-self-driving = 1.7e-7/1.2e-8 = 14.

So, the data to date shows that there is a chance of death 14 times higher with the self driving vehicles. That is significant.

My strong suspicion is that since the number of total miles driven by self-driving cars is so low, the death rate will likely go up and not down as these cars are exposed to more real world situations. I strongly doubt these things have had to deal with icy roads, heavy fog, storms, etc. yet. What happens when the sensors are degraded and damaged due to dirt, rocks thrown up by the vehicle in front, hail, heavy rain, mud, etc?

And the driver behind the wheel didn’t even touch the controls, so that certainly doesn’t make the case for autonomy.

My concern isn’t whether Uber thought their cars should go on the road or not. It’s clear they thought so. They were wrong. I’m usually not a fan of a lot of heavy regulation, but I don’t want me or my family to be the guinea pigs for these folks’ software quality control testing. Get these things out of the public and back into the controlled testing environment until they can be _proven_ to be at least as safe as the normal vehicles. And the heavy burden of proof is on the vendors.



Sent from my iPad using Tapatalk

 

[flyingcheesehead]

Exactly what I've been saying! That's why for full autonomy you will have to have so-called "magic rails" to keep all these cars on track. In adverse weather, or bad road conditions where the lane striping has the potential to be hidden/covered up, that solution is embedded sensors or magnetic lane striping.

Here's a good article on the subject.

So what that article basically says is that 3M is betting against the automakers and hoping to cash in when they "give up".

I wouldn't hold my breath.

 

[Palmpilot]

Where computers fail is combing the information and determining the cues required to make the determination of where the edge of the road is, or where a rough approximation of the road is.

Exactly what I've been saying! That's why for full autonomy you will have to have so-called "magic rails" to keep all these cars on track. In adverse weather, or bad road conditions where the lane striping has the potential to be hidden/covered up, that solution is embedded sensors or magnetic lane striping.

Here's a good article on the subject.

Here's an example of how difficult that problem can become even in perfect weather. This came up in a discussion of the Tesla on "autopilot" that smashed into the divider between the main highway and the left-side off ramp. I'm familiar with this interchange because it's only a couple of miles from my house.

 

[Juliet Hotel]

Just because you say it doesn't make it so.

I don't think anyone knows whether autonomous vehicles will turn out to be practical as the main means of transportation.

True enough. No one can really know anything until it actually happens. But what we're talking about is not impossible. With the technology that already exists, its far from impossible in fact. I see it being very similar to heavier than air powered flight. By the time December 17, 1903 rolled around, many different entities were developing and working on finding a way to make flight happen. And although no one knew quite how to do it until it actually happened, it was largely accepted by most in the scientific community that powered flight was possible based simply on what was already known and the technology that already existed. Fully autonomous vehicles are very similar in this respect.


And we're not talking about some form of dryer lint based unlimited free energy that one guy is trying to invent in his garage here. This technology is being pursued and developed by every major auto maker as well as several large and well funded independent entities. It really isn't a question of if this happen, its only a question of when. I happen to be of the opinion that it will happen sooner than many people expect.

 

[Palmpilot]

But what we're talking about is not impossible.

I'm not convinced that anyone knows whether it's possible or not.

It really isn't a question of if this happen, its only a question of when.

This is where your argument goes off the rails, in my opinion, because you're supporting it with plausibility arguments, and plausibility arguments do not justify certainty. Predicting the course of technological advancement is just not that easy.

 

[azblackbird]

So what that article basically says is that 3M is betting against the automakers and hoping to cash in when they "give up".

I'd take that bet in a heartbeat! I don't care how high-tech the car is, there is just way too much "noise" out in the real world for an autonomous car to operate safely. You will have to employ some sort of guidance systems whether it be embedded sensors, smart road signs, or magnetic lane stripping to keep these cars on the proper track. Even then it's still a crap shoot. Right now it would take me about 2 minutes to disable even the most sophisticated autonomous car on the road and make it go bonkers. We have a very long ways to go yet before any of these cars are ready for prime time.

 

[flyingcheesehead]

I'd take that bet in a heartbeat!

I'd suggest you go and buy some 3M stock then.

 

[tspear]

For those that say it will not happen.
Have you tested the limits of parking assist?
Now this is a very limited situation, however some of these systems can deal with snow, fog, night, low/nil lighting...

I have seen how much parking assist has changed in the past five years in a few cars. The ability of cars to deal with adverse weather in this limited situation has improved very fast. I think the ability of cars to determine where the road is, has also been advancing just as rapidly.

Sent from my SM-J737T using Tapatalk

 

[1RTK1]

Nothing better than an old mouse trap to snap the little buggers neck and stop them in there tracks......Now if you have multitudes poison works better.....

So do better mouse traps.

 

[Salty]

Just imagine how great life is going to be in the brave new world when you encounter these scenarios in your full autonomous hominid cargo vehicle. Either the roads will completely shut down, or the hominids are going to have a really bad day.

 

[Juliet Hotel]

Just imagine how great life is going to be in the brave new world when you encounter these scenarios in your full autonomous hominid cargo vehicle. Either the roads will completely shut down, or the hominids are going to have a really bad day.

In all of those situations, the roads would be completely shut down now. And in some of them, the dust storm for instance, we may very well find that the sensors on autonomous vehicles can see through the cloud just fine and keep right on rolling safely while Arlo T Luddite drives his '72 Dart right into a pole because he can't see squat and doesn't have enough sense to slow down or stop. If you're going to mock autonomous tech, at least come up with something that's at least slightly valid.

 

[Dav8or]

And we're not talking about some form of dryer lint based unlimited free energy that one guy is trying to invent in his garage here.

I have dryer lint. Should I be working on that? Kickstarter, or just go straight to IPO??

 

[flyingcheesehead]

Just imagine how great life is going to be in the brave new world when you encounter these scenarios in your full autonomous hominid cargo vehicle. Either the roads will completely shut down, or the hominids are going to have a really bad day.

Ironically, the camera-based autonomous driving solutions will handle those a heckuva lot better than any magic magnetic roadway systems...

 

[Salty]

Ironically, the camera-based autonomous driving solutions will handle those a heckuva lot better than any magic magnetic roadway systems...

Not remotely as good as a human would.

 

[Juliet Hotel]

Not remotely as good as a human would.

Agree to disagree.

 

[1RTK1]

Not remotely as good as a human would.

You don’t drive in kalifornia much do ya...

 

[Palmpilot]

Elon Musk says "Your Tesla will soon be able to go from your garage at home to parking at work with no driver input at all."

https://arstechnica.com/cars/2018/1...t-update-to-allow-for-no-driver-input-at-all/

I thought this was interesting:

"A study from the RAND Corporation in 2016 demonstrated that a vehicle would need to be driven for 275 million failure-free miles to demonstrate a fatality rate equal to that of a human driver."
​