Will cars be autonomous in three years?

Questions on how we spend our money and our time - consumer goods and services, home and vehicle, leisure and recreational activities
bubbly
Posts: 68
Joined: Tue Jun 16, 2015 4:21 pm

Re: Will cars be autonomous in three years?

Post by bubbly » Mon Apr 24, 2017 1:50 pm

flamesabers wrote:
bubbly wrote:I'm curious, does this fear of automated navigation extend to air travel? It seems like humans have an elevated sense of their personal driving ability when it comes to automobiles. However, I don't see anyone saying they could fly an airplane better than the onboard autopilot software. I would also contend that air navigation is much more difficult and carries much more risk in the event of failure but most people have no problem flying. I'm just curious why this is.
Most people have never piloted a commercial jet so I don't think it's a fair comparison. I think a more accurate question is whether trained pilots think they can fly a plane better than onboard autopilot software.

I would much rather be on a plane that's controlled by a human pilot then one that is controlled by software. I don't mind the use of autopilot software so long as there is a human pilot ready to take control in case there is a glitch in the software.

When it comes to operating vehicles that can result in fatal if not deadly accidents, there's an important difference between humans and computers. Humans have an awareness and instinctive fear of death and injury, computers don't.
Are you sure you would much rather be on a plane controlled by a person? You do realize that when there's low visibility, certain airlines require autopilot to land planes as opposed to a human operator. And it's not whether a person "thinks" they can fly a plane better. It must be demonstrated with scientific data. The fact that the airline industry requires autopilot to land planes during low visibility speaks volumes to the level of perception that humans cannot achieve. You say you don't mind software as long as there is human intervention. You will have this option for autonomous vehicles. Presumably, there will always be a human operator on hand to override the system in case of malfunction.

User avatar
vitaflo
Posts: 1202
Joined: Sat Sep 03, 2011 3:02 pm

Re: Will cars be autonomous in three years?

Post by vitaflo » Mon Apr 24, 2017 1:50 pm

flamesabers wrote:
I think this is a very good point. I also am a person who would be afraid of riding in a self-driving car. There are already alternatives modes of transportation for people who are unable/unwilling to drive.
People also used to be deathly afraid of elevators. They automated them in 1900 but nobody wanted to dare use them because they didn't have operators. People got over it when they added big red Stop buttons to them. Now nobody thinks twice about elevators.

People don't like having control taken away, even if the alternative is much safer. Everyone thinks they're a good driver but everyone has made stupid driving mistakes. If there's one things humans are bad at it's not understanding their own limitations.

Jimmie
Posts: 431
Joined: Tue Mar 07, 2017 7:05 pm

Re: Will cars be autonomous in three years?

Post by Jimmie » Mon Apr 24, 2017 1:52 pm

bubbly wrote:For any out of control driver, there are ways to mitigate this issue. One would be to transmit the GPS positions of that driver. So if it was driving erratically, the trail of GPS positions from that car would alert the system to take caution. You could also do this with a vision system by analyzing the vehicle versus their projected path of travel, which is what Tesla does currently. To get the GPS positions, you would need cars to be equipped with gps receivers and for that receiver to talk to some sort of cloud infrastructure (there are privacy issues with this approach). Going 1 step further, the non-autonomous vehicle could also transmit wheel angle and speed to provide a complete projected path of travel.

As for GPS dead zones. GPS currently is not required for lane centering, as Tesla as currently demonstrated. They are not using GPS at all. But to your point, what if you wanted to go from point A to B, then yes, you will need GPS. However for dead zones, you have inertial measurement units that can very accurately navigate the vehicle. Think of GPS as a grid. Let's say you are at (0,0) meters and you want to get to (10,10) but you lost GPS at (0,0). If you didn't have absolute gps, you can still guide by me telling you to go straight 10 meters, turn left and go straight 10 more meters (depending on how you were originally oriented). The inertial measurement unit is just a more precise way of calculating these positions from accelerometers and/or gyroscopes.

You can also locate your position with vision systems, known as perception based localization. Let's say you don't know where you are because of a GPS dead spot, but you see a stop sign that is known to be at (5,5). You also know that you are 3 meters east and 2 meters south of the stop sign from LIDAR scans, so you can deduce your position that way.
Thanks for the detailed and well thought-out explanations. This is very interesting, especially the way you explained it all.

likegarden
Posts: 2932
Joined: Mon Feb 26, 2007 5:33 pm

Re: Will cars be autonomous in three years?

Post by likegarden » Mon Apr 24, 2017 1:56 pm

Look at all the variations in drivers, cars, roads, weather, day/night, construction going on, road signs and markings, accidents by others, etc. there will be a lot of law suits and happy lawyers in case some company introduces those too fast. There should be big signs on any of those cars, so human drivers can safely stay away from them. I am a retired engineer who worked in automatic controls.

bubbly
Posts: 68
Joined: Tue Jun 16, 2015 4:21 pm

Re: Will cars be autonomous in three years?

Post by bubbly » Mon Apr 24, 2017 1:56 pm

Jimmie wrote:
bubbly wrote:For any out of control driver, there are ways to mitigate this issue. One would be to transmit the GPS positions of that driver. So if it was driving erratically, the trail of GPS positions from that car would alert the system to take caution. You could also do this with a vision system by analyzing the vehicle versus their projected path of travel, which is what Tesla does currently. To get the GPS positions, you would need cars to be equipped with gps receivers and for that receiver to talk to some sort of cloud infrastructure (there are privacy issues with this approach). Going 1 step further, the non-autonomous vehicle could also transmit wheel angle and speed to provide a complete projected path of travel.

As for GPS dead zones. GPS currently is not required for lane centering, as Tesla as currently demonstrated. They are not using GPS at all. But to your point, what if you wanted to go from point A to B, then yes, you will need GPS. However for dead zones, you have inertial measurement units that can very accurately navigate the vehicle. Think of GPS as a grid. Let's say you are at (0,0) meters and you want to get to (10,10) but you lost GPS at (0,0). If you didn't have absolute gps, you can still guide by me telling you to go straight 10 meters, turn left and go straight 10 more meters (depending on how you were originally oriented). The inertial measurement unit is just a more precise way of calculating these positions from accelerometers and/or gyroscopes.

You can also locate your position with vision systems, known as perception based localization. Let's say you don't know where you are because of a GPS dead spot, but you see a stop sign that is known to be at (5,5). You also know that you are 3 meters east and 2 meters south of the stop sign from LIDAR scans, so you can deduce your position that way.
Thanks for the detailed and well thought-out explanations. This is very interesting, especially the way you explained it all.
You are welcome! :)

I spend all day at work contemplating many of these things. It's just a shame because I like to talk about it with people but it's not exactly a conversation starter at parties :confused

Tamales
Posts: 1423
Joined: Sat Jul 05, 2014 10:47 am

Re: Will cars be autonomous in three years?

Post by Tamales » Mon Apr 24, 2017 2:03 pm

vitaflo wrote:People underestimate the speed of tech advancement. 10 years ago the iPhone didn't exist, and Facebook had just opened up the site to the public for the first time. 20 years ago less than 2% of people used the internet and only 4% had cell phones. I wouldn't underestimate the speed of tech change.
No comparison between iphone/facebook and autonomous driving.

First, iphone (and smartphones in general) wasn't a technology advancement per se. It was integration of existing pieces from the same platform or adjacent platforms. It was more of a packaging and marketing exercise.
And facebook isn't a technology or an advancement, but that a whole 'nuther discussion.

Autonomous driving on the other hand, requires new technology to be invented, adapted, and improved; lather, rinse, repeat. And since it involves safety of life and property, it will be many generations. The closest analogy would probably be air bags or ESC or ABS, but even those are a drop in the bucket compared to the scale and scope of full autonomous driving.

These types of things have to be mandated by law or they aren't likely to even reach 50% adoption. That will surely come, but it will be in many incremental steps over decades, before the "humanless carriage" affects the masses.

Right now, the nearest-term discussion (which we can't have here) involves the DoT's multi-year interest in safety-related communications between cars and the related DSRC mandate (which may or may not go into effect).

PS, actually "people" way OVERestimate the speed of technology advancement (speaking of true technology advancement, not integration and marketing posing as technology advancement). Analysts (who are in business of selling hype) apply a hockey-stick curve to everything, and people buy into it. There are entire books written about this.

bloom2708
Posts: 7089
Joined: Wed Apr 02, 2014 2:08 pm
Location: Fargo, ND

Re: Will cars be autonomous in three years?

Post by bloom2708 » Mon Apr 24, 2017 2:08 pm

Part of me thinks that cars do what cars do. It will be difficult to dislodge the car from the driver.

I'd rather see a technology like HyperLoop being worked on with R&D dollars.

It seems like that could be a much bigger "leap forward".
"People want confirmation, not advice" Unknown | "We are here to provoke thoughtfulness, not agree with you" Unknown | Four words: Whole food, plant based

themesrob
Posts: 261
Joined: Fri Jul 15, 2016 1:58 pm

Re: Will cars be autonomous in three years?

Post by themesrob » Mon Apr 24, 2017 2:09 pm

vitaflo wrote:
flamesabers wrote:
I think this is a very good point. I also am a person who would be afraid of riding in a self-driving car. There are already alternatives modes of transportation for people who are unable/unwilling to drive.
People also used to be deathly afraid of elevators. They automated them in 1900 but nobody wanted to dare use them because they didn't have operators. People got over it when they added big red Stop buttons to them. Now nobody thinks twice about elevators.

People don't like having control taken away, even if the alternative is much safer. Everyone thinks they're a good driver but everyone has made stupid driving mistakes. If there's one things humans are bad at it's not understanding their own limitations.
tadamsmar wrote:
themesrob wrote:A related interesting point which I did not know was that many "semi-autonomous" features (e.g. forward-collision warning, lane-departure tech) have actually been around for 10-20 years, but their market penetration is surprisingly low, and the people who own cars with those features usually deactivate them (because they feel they don't need them, or because those lane-departure beeps are annoying, I guess).

My opinion is that we don't see autonomous vehicles on the open road at any point. (So yes, buy the truck! :wink: )
Electronic stability control (ESC) has been around for more that 20 years. It's not annoying, most people probably don't even know they have it. Most people have probably never even heard of it. It has 100% market penetration because it was mandated by the NHTSA on all cars as of the 2012 model year. It was mandated because the vehicles that had it were evidencing 33% lower fatality rate, pretty good for something that cost around $250 in parts.

Foward collision avoidance (with autonomous braking) is reducing rear-end collisions according to the evidence.

My opinion is that the issue is not whether these measures are effective or whether, all other things being equal, having roadways of 100% autonomous cars would be safer than what we have today. I certainly believe that if all the various obstacles are overcome (note: that appears to be a big if), autonomous cars would make my life both much safer and more pleasant. But it seems that many people don't feel that way -- driving for them is a day-to-day activity, like using a computer or a telephone, and (1) they do not trust the technology, which as it currently stands is justifiable, and/or (2) they would prefer to do it themselves, regardless of how safe the technology is. Changing a behavioral preference is really difficult.

User avatar
flamesabers
Posts: 1807
Joined: Fri Mar 03, 2017 12:05 pm
Location: Rochester, MN

Re: Will cars be autonomous in three years?

Post by flamesabers » Mon Apr 24, 2017 2:11 pm

vitaflo wrote:People also used to be deathly afraid of elevators. They automated them in 1900 but nobody wanted to dare use them because they didn't have operators. People got over it when they added big red Stop buttons to them. Now nobody thinks twice about elevators.
It's not like stairs are obsolete though. And despite the advancement of elevators, accidents still happen. If there's a fire or a power outage, I'm sure people would much rather take the stairs then the elevator.
likegarden wrote:Look at all the variations in drivers, cars, roads, weather, day/night, construction going on, road signs and markings, accidents by others, etc. there will be a lot of law suits and happy lawyers in case some company introduces those too fast. There should be big signs on any of those cars, so human drivers can safely stay away from them. I am a retired engineer who worked in automatic controls.
If an automated car did get into an accident, who would be at fault? The owner of the vehicle or the manufacturer of the vehicle/software?

Jimmie
Posts: 431
Joined: Tue Mar 07, 2017 7:05 pm

Re: Will cars be autonomous in three years?

Post by Jimmie » Mon Apr 24, 2017 2:14 pm

Tamales wrote:PS, actually "people" way OVERestimate the speed of technology advancement (speaking of true technology advancement, not integration and marketing posing as technology advancement). Analysts (who are in business of selling hype) apply a hockey-stick curve to everything, and people buy into it. There are entire books written about this.
Watch the part toward the beginning of "2001: A Space Odyssey" on the spacecraft and think about how that wasn't around in 2001.

rkhusky
Posts: 7820
Joined: Thu Aug 18, 2011 8:09 pm

Re: Will cars be autonomous in three years?

Post by rkhusky » Mon Apr 24, 2017 2:29 pm

Cars already have autonomous functionality - antilock braking and stability/traction control.

Cars will have hands-off the steering wheel capability for highway driving in decent weather conditions in 3 years. Total autonomy in all road environments and weather conditions is a few decades away, unless the government intervenes and forbids human drivers and integrates sensors in all the roads and signs.

AlwaysBeClimbing
Posts: 143
Joined: Fri Mar 31, 2017 10:39 am

Re: Will cars be autonomous in three years?

Post by AlwaysBeClimbing » Mon Apr 24, 2017 2:59 pm

jharkin wrote:Another vote that there is absolutely ZERO chance the majority of cars will be autonomous in 3, 10 or even 20 years.

I work in high tech and I hear a lot of prognosticating about this but I dont buy it. For some reason Bogleheads seems to have a high concentration of anti-car people who see them as appliances and hate driving, but out in the real world Americans LOVE their cars, love driving, and wont give them up without a fight. Plus anyone who thinks its even logistacally possible must live in a Silicon Valley tech bubble and be blind to a lot of realities... The average car on American roads is 15 years old (i.e. turning over the entire fleet in less time than that is impossible)... Not to mention there is no way a farmer in Kansas is going to get an autonomous pickup truck to deliver feed...

I saw this article in my Facebook news feed recently and I think it makes some very good points. Many of these systems are making us LESS safe, not more because peoples driving skills are getting so bad:

http://newcartographer.com/combustion/augmented.html
+1 The linked article really hits the nail on the head IMO. It's just nuts that the airline industry's approach, augmentation rather than automation(taking the human out of the loop), isn't being adopted by the auto industry.

User avatar
greg24
Posts: 3748
Joined: Tue Feb 20, 2007 10:34 am

Re: Will cars be autonomous in three years?

Post by greg24 » Mon Apr 24, 2017 2:59 pm

jeep5ter wrote:Also: how many 15 year olds do you think are counting the hours until the day comes that they can sit in a car and not drive it?
You may not have noticed, but 15 year olds are not counting the hours until they can drive like they used to.

http://www.pbs.org/newshour/updates/man ... s-license/
The share of high school seniors across the country who have a driver’s license dropped from 85.3 percent in 1996 to a record low 71.5 percent in 2015

The drop has been sharpest in the South, where the share of high school seniors with a driver’s license fell from 88.6 percent in 1996 to 71.2 percent in 2015. High school seniors are most likely to have a license in the Midwest — 80.4 percent — and least likely to have one in the Northeast — 64.8 percent.

AlwaysBeClimbing
Posts: 143
Joined: Fri Mar 31, 2017 10:39 am

Re: Will cars be autonomous in three years?

Post by AlwaysBeClimbing » Mon Apr 24, 2017 3:02 pm

AlwaysBeClimbing wrote:
jharkin wrote:Another vote that there is absolutely ZERO chance the majority of cars will be autonomous in 3, 10 or even 20 years.

I work in high tech and I hear a lot of prognosticating about this but I dont buy it. For some reason Bogleheads seems to have a high concentration of anti-car people who see them as appliances and hate driving, but out in the real world Americans LOVE their cars, love driving, and wont give them up without a fight. Plus anyone who thinks its even logistacally possible must live in a Silicon Valley tech bubble and be blind to a lot of realities... The average car on American roads is 15 years old (i.e. turning over the entire fleet in less time than that is impossible)... Not to mention there is no way a farmer in Kansas is going to get an autonomous pickup truck to deliver feed...

I saw this article in my Facebook news feed recently and I think it makes some very good points. Many of these systems are making us LESS safe, not more because peoples driving skills are getting so bad:

http://newcartographer.com/combustion/augmented.html
+1 The linked article really hits the nail on the head IMO. It's just nuts that the airline industry's approach, augmentation rather than automation(taking the human out of the loop), isn't being adopted by the auto/tech industry.

User avatar
flamesabers
Posts: 1807
Joined: Fri Mar 03, 2017 12:05 pm
Location: Rochester, MN

Re: Will cars be autonomous in three years?

Post by flamesabers » Mon Apr 24, 2017 3:15 pm

greg24 wrote:
jeep5ter wrote:Also: how many 15 year olds do you think are counting the hours until the day comes that they can sit in a car and not drive it?
You may not have noticed, but 15 year olds are not counting the hours until they can drive like they used to.

http://www.pbs.org/newshour/updates/man ... s-license/
The share of high school seniors across the country who have a driver’s license dropped from 85.3 percent in 1996 to a record low 71.5 percent in 2015

The drop has been sharpest in the South, where the share of high school seniors with a driver’s license fell from 88.6 percent in 1996 to 71.2 percent in 2015. High school seniors are most likely to have a license in the Midwest — 80.4 percent — and least likely to have one in the Northeast — 64.8 percent.
Speaking as someone who lives and grew up in the Midwest, I'm not surprised in the slightest that the Midwest has the highest percentage of high school seniors who have a drivers license. While growing up it seemed like everyone had a drivers license except for those who didn't have access to a car and used public transportation to get around everywhere. It makes me wonder if it's just a Midwestern thing to be resistant to the idea of autonomous cars versus for people in the Northeast who may be more open to the idea because of not learning how to drive a car in their teen years.

AlwaysBeClimbing
Posts: 143
Joined: Fri Mar 31, 2017 10:39 am

Re: Will cars be autonomous in three years?

Post by AlwaysBeClimbing » Mon Apr 24, 2017 3:20 pm

Great interview here with airline captain "Sully" Sullenberger on safety and self-driving cars. This particular passage jumps out at me:

"When we assign technology as the doer and the human component as the monitor, we’re doing it backwards. Humans are inherently poor monitors."

We already have an epidemic of distracted drivers today and somehow these same nitwits are going to pay enough attention to react in time and take over from the computer to respond to some emergency situation. Sure they are.

Atgard
Posts: 420
Joined: Wed Apr 09, 2014 2:02 pm

Re: Will cars be autonomous in three years?

Post by Atgard » Mon Apr 24, 2017 3:24 pm

Jimmie wrote:Two things terrify me about autonomous vehicles and I am a control system engineer.

First, there is nothing that prevents a wild, out-of-control driver from hitting the autonomous vehicle. There are too many places when there is no "out" for it, regardless of how smart it is.
And what will you do when you're driving your car, and a wild out of control driver tries to hit you, but there is no "out" for it, regardless of how smart you are?

User avatar
Flymore
Posts: 297
Joined: Tue May 31, 2016 1:31 pm

Re: Will cars be autonomous in three years?

Post by Flymore » Mon Apr 24, 2017 3:52 pm

There will be many self-steering cars in three years. This will be the transitional phase to self-driving cars still many years off.
With self-steering cars, the driver behind the wheel will still be responsible for the automobile and there will be warnings about the driver must remain alert during the drive and of course many will ignore the warnings and let the car drive on its own.

Still for those who drive a lot, truckers and family vacations etc... self-steering will be fantastic and a relief.

User avatar
TomatoTomahto
Posts: 9737
Joined: Mon Apr 11, 2011 1:48 pm

Re: Will cars be autonomous in three years?

Post by TomatoTomahto » Mon Apr 24, 2017 3:54 pm

greg24 wrote:
jeep5ter wrote:Also: how many 15 year olds do you think are counting the hours until the day comes that they can sit in a car and not drive it?
You may not have noticed, but 15 year olds are not counting the hours until they can drive like they used to.
3 of 4 of my kids don't care about driving; it's the freedom to not need rides that appealed to them. An Uber is just as good as a car, and better if they're living in a city. The 4th likes racing cars, on a track, but has no particular fondness for driving among amateurs on city streets.

I asked the one soonest to graduate if he wanted a Tesla Model 3 as a graduation present (I would have had to put down a deposit well in advance of his graduation). Reply: Nah, no thanks, I'll probably live in a city.
Okay, I get it; I won't be political or controversial. The Earth is flat.

bloom2708
Posts: 7089
Joined: Wed Apr 02, 2014 2:08 pm
Location: Fargo, ND

Re: Will cars be autonomous in three years?

Post by bloom2708 » Mon Apr 24, 2017 4:06 pm

I envision the day where I self drive my car in podunk, midwest USA to the closest HyperLoop station.

I hyperloop to the BIG Citay, West or East Coast where no self drive cars are allowed. I exit the hyperloop and hop in a Google Tesla Uber to take me to my hotel.

Science fiction?
"People want confirmation, not advice" Unknown | "We are here to provoke thoughtfulness, not agree with you" Unknown | Four words: Whole food, plant based

Jimmie
Posts: 431
Joined: Tue Mar 07, 2017 7:05 pm

Re: Will cars be autonomous in three years?

Post by Jimmie » Mon Apr 24, 2017 5:16 pm

Atgard wrote:
Jimmie wrote:Two things terrify me about autonomous vehicles and I am a control system engineer.

First, there is nothing that prevents a wild, out-of-control driver from hitting the autonomous vehicle. There are too many places when there is no "out" for it, regardless of how smart it is.
And what will you do when you're driving your car, and a wild out of control driver tries to hit you, but there is no "out" for it, regardless of how smart you are?
Exactly my point. Autonomous vehicles offer little, if any, benefit over human control in these types of situations. Cars operated by either method of control are doomed to crash.

harikaried
Posts: 1294
Joined: Fri Mar 09, 2012 3:47 pm

Re: Will cars be autonomous in three years?

Post by harikaried » Mon Apr 24, 2017 6:00 pm

Jimmie wrote:Autonomous vehicles offer little, if any, benefit over human control in these types of situations.
There should be at least a few significant benefits over human control in these "wild out of control driver" situations: awareness, responsiveness, and control.

Ideally the autonomous vehicle can realize something bad is happening, e.g., a human driver might assume safety with other vehicles proceeding straight through a green light, but an autonomous vehicle can be constantly tracking all directions even if they're "usually safe" say an out of control vehicle going in the wrong direction of the crossing street. This awareness can help the autonomous vehicle decide not to enter the intersection and avoid an accident.

Some people have amazing reaction times when they expect to react, but autonomous vehicles could be ready to react at any moment. Being ready to handle a situation requires some amount of focus and energy on the driver, and maintaining that for any moment can be quite draining especially when the driver believes it to be a safe situation.

Not many people are prepared for accidents, so when one happens, it's quite possible that the driver also becomes an additional "wild out of control driver." Drivers can easily under-react or over-react, but computers at least can avoid the emotional aspect and try to do what it was programmed to do.

Here's a video of a failure of Tesla's AutoPilot (and driver attentiveness) in that it crashes into construction barriers, but interestingly, the vehicle immediately turns on hazard lights and maintains its lane-centering behavior, where a driver could have easily bounced multiple times into the barrier or swerved into the right lane causing another accident:

Video/gif: https://electrek.files.wordpress.com/20 ... -2-gif.gif
from https://electrek.co/2017/03/02/tesla-au ... te-to-use/

(On a side note relating to that video, it looks like the vehicle recording is following less than half a second behind the Tesla. An autonomous vehicle would probably follow further as there's no emotion of trying to get to your destination 2 seconds sooner.)

petiejoe
Posts: 46
Joined: Wed Feb 13, 2013 8:18 pm

Re: Will cars be autonomous in three years?

Post by petiejoe » Mon Apr 24, 2017 6:30 pm

Jimmie wrote:
Atgard wrote:
Jimmie wrote:Two things terrify me about autonomous vehicles and I am a control system engineer.

First, there is nothing that prevents a wild, out-of-control driver from hitting the autonomous vehicle. There are too many places when there is no "out" for it, regardless of how smart it is.
And what will you do when you're driving your car, and a wild out of control driver tries to hit you, but there is no "out" for it, regardless of how smart you are?
Exactly my point. Autonomous vehicles offer little, if any, benefit over human control in these types of situations. Cars operated by either method of control are doomed to crash.
If we show that human driving offers no benefit for catastrophic events (e.g. someone side-swiping you at a stoplight) and autonomous driving provides benefit for typical events (e.g. the person ahead of you brakes while you're looking at the radio), why wouldn't we select autonomous driving as the preferred mode?

It's hard to say how close we are to full autonomy being technically feasible without access to the research data within Tesla, Google, or Uber. I suspect we're close technologically but the current political climate makes it difficult to execute (1 person dying in a fully autonomous collision will set the programs back a lot more than 1000 lives saved will advance them). If I were to make a guess, I'd say the next 4-5 years for the technology being available for consumers. 10-15 years from now, governments will look at the apples-to-apples accident data and decide that they should require autonomous driving options for all future cars. Within the next 20-30 years, I expect we'll have areas (e.g. large cities) where human driving is outlawed. As someone who really does enjoy driving sometimes (on open country roads), I hope that they don't decide to outlaw human driving everywhere.

killjoy2012
Posts: 1094
Joined: Wed Sep 26, 2012 5:30 pm

Re: Will cars be autonomous in three years?

Post by killjoy2012 » Mon Apr 24, 2017 6:41 pm

In 3 years? Not a chance.

Go buy the F-150.

User avatar
tadamsmar
Posts: 8579
Joined: Mon May 07, 2007 12:33 pm

Re: Will cars be autonomous in three years?

Post by tadamsmar » Mon Apr 24, 2017 7:07 pm

themesrob wrote:the people who own cars with those features usually deactivate them (because they feel they don't need them, or because those lane-departure beeps are annoying, I guess).
The beeps are annoying. Some lane-departure warning vibrate the seat or steering wheel. I rented a Ford van that vibrated the seat near my right or left thigh. It did not annoy me because it was not alerting everyone in the car, and it was quite natural to react to a signal on the side you need to steer away from. The beeps are not directional.

Jimmie
Posts: 431
Joined: Tue Mar 07, 2017 7:05 pm

Re: Will cars be autonomous in three years?

Post by Jimmie » Mon Apr 24, 2017 7:22 pm

petiejoe wrote:If we show that human driving offers no benefit for catastrophic events (e.g. someone side-swiping you at a stoplight) and autonomous driving provides benefit for typical events (e.g. the person ahead of you brakes while you're looking at the radio), why wouldn't we select autonomous driving as the preferred mode?

It's hard to say how close we are to full autonomy being technically feasible without access to the research data within Tesla, Google, or Uber. I suspect we're close technologically but the current political climate makes it difficult to execute (1 person dying in a fully autonomous collision will set the programs back a lot more than 1000 lives saved will advance them). If I were to make a guess, I'd say the next 4-5 years for the technology being available for consumers. 10-15 years from now, governments will look at the apples-to-apples accident data and decide that they should require autonomous driving options for all future cars. Within the next 20-30 years, I expect we'll have areas (e.g. large cities) where human driving is outlawed. As someone who really does enjoy driving sometimes (on open country roads), I hope that they don't decide to outlaw human driving everywhere.
I could not agree more about what you have posted. However, the advantages you cite are available today as pre-emptive braking and other individual safety enhancements.

My point (and you agree apparently by the boldface above) is that full autonomous vehicle operation still has a way to go.

From an engineering standpoint, why the jump from every car being a beat-up Ford Fiesta to a full-blown autonomous car? This needs to be an evolution of technology with improvements made as both the technology improves and society learns to use it and embrace it.

User avatar
Watty
Posts: 18019
Joined: Wed Oct 10, 2007 3:55 pm

Re: Will cars be autonomous in three years?

Post by Watty » Mon Apr 24, 2017 9:03 pm

There are two components to making autonomous cars common.

1) The technology to do it.

2) Making it affordable enough that the majority of people can afford it.

It took about 25 years from the time airbags were introduced until they became mandatory and getting the cost down was one of the reasons.

FraggleRock
Posts: 362
Joined: Sun Feb 12, 2017 6:50 pm

cars, car service

Post by FraggleRock » Mon Apr 24, 2017 9:13 pm

No, in 3 years.
Yes, in 20+.
Lease?
Yes, because technology is moving too quickly in cars right now.
We own a 10 year old car with 43K miles. In 2020ish, we plan to lease a 100% electric car.
In 2023 we can make the keep or move on decision.
By 2040 I think we are looking ar at buying transportation services vs cars.

tryingmybest
Posts: 37
Joined: Tue Nov 04, 2008 12:23 pm

Re: Will cars be autonomous in three years?

Post by tryingmybest » Mon Apr 24, 2017 9:54 pm

May 2016 - Bill Gates - "It's certainly more than 15 years off before it'll be a meaningful percentage of cars driven,"
https://www.youtube.com/watch?v=paNduXK_gJs

Feb 27, 2017 - Warren Buffett – bets 10 years from now less than 10% of cars will be self-driving
https://www.youtube.com/watch?v=RZMotpUMxm4

randomguy
Posts: 8509
Joined: Wed Sep 17, 2014 9:00 am

Re: Will cars be autonomous in three years?

Post by randomguy » Mon Apr 24, 2017 9:58 pm

bubbly wrote:To expound on my previous points: Any other car manufacturer that I'm aware of other than Tesla, only offers adaptive cruise. Tesla is the only manufacturer that offers steering at high speeds, dubbed lane centering technology. EDIT: it seems like beginning this year, some non-Tesla vehicles will have relatively primitive lane keeping technology that steers the vehicle in certain situations.
Mercedes has had active lane keeping (I thought it was 150km/hr back then. The new one does 200km/hr) since at least 2012 and adaptive cruise control during the same period. They didn't give it a fancy name. They gave it some weird german one:) Pretty sure Volvo did also and probably a bunch of other ones. Here is a caranddriver from 2016 http://www.caranddriver.com/features/se ... ti-feature . Now the last is vastly outdated (volvo, audi, and tesla have all new systems. Pretty sure MB updated theirs for 2017 (2018 models) also). With tech like this it isn't a binary choice. It isn't like you have active lane keeping or your don't. Once you have it you have to figure out how well it works.

The one thing that tesla has done really well is update your system. Maybe the other manufactures were doing it also and not telling you but I sort of doubt it. How this plays out as tesla transitions from using mobileye technology to their own will be fun to watch. I find it hard to believe ANY oem is going to be doing a good job updating these systems when they get to 10+ years old.

randomguy
Posts: 8509
Joined: Wed Sep 17, 2014 9:00 am

Re: Will cars be autonomous in three years?

Post by randomguy » Mon Apr 24, 2017 10:11 pm

jeep5ter wrote:Maybe in 50 years there will be some impact in urban areas, but not in the majority of the country.

Also: how many 15 year olds do you think are counting the hours until the day comes that they can sit in a car and not drive it?
how many 80+ years are counting the hours til they can have a car that can drive them as they can no longer do it safely? The older population might be a lot bigger than the younger ones:) Most 15 years olds in my experience don't want to drive. They want the freedom to go places. It is easy to confuse those 2 desires as right now driving is the way most people get that freedom of motion.


50 years is an eternity in tech. Go look at 1967 computers:) You can look at how self driving cars have progressed since the first Darpa grand challenge in 2004 to now and the progress is totally insane. We are probably like 95%+ of the way to the self driving car. The hard part is figuring out is that last 5% solveable or not. Deep learning over the last 5 years has done an incredible job attacking problems that in the past were not considered solvable. How close we are to the limits of it is hard to say.

As you relax the requirements from 100% self driving (you give it a address and you take a nap), the problem gets a lot easier. I wouldn't be shocked to see highway self drivers (i.e. it gets on the ramp and beeps you when you get to yours) in 3-5 years. Handling urban areas is orders of magnitude harder. Look at how long google has been driving around mountain view and they are still working on corner cases.

Tamales
Posts: 1423
Joined: Sat Jul 05, 2014 10:47 am

Re: Will cars be autonomous in three years?

Post by Tamales » Mon Apr 24, 2017 11:03 pm

randomguy wrote:You can look at how self driving cars have progressed since the first Darpa grand challenge in 2004 to now and the progress is totally insane. We are probably like 95%+ of the way to the self driving car. The hard part is figuring out is that last 5% solveable or not. Deep learning over the last 5 years has done an incredible job attacking problems that in the past were not considered solvable. How close we are to the limits of it is hard to say.
I'm not sure that, collectively, all these individually developed proprietary systems and algorithms from many different players is a good measure of the progress since they are essentially all developing their technology in a silo, rather than a combined effort toward an end goal with shared experiences on fault conditions. They will all have different weaknesses and things that they don't know that that don't know. How much gets swept under the rug versus actually fixed is anybody's guess. But rest assured, they won't all be equivalent in how rigorously tested they are.

These systems will never be able to account for every permutation that nature (e.g. rock slide 10 seconds before you reach a mountain corner, or animal in the road) or location (e.g. road construction where you are diverted through cones, or fallen power lines) or random fools on the road (chair falls off the overloaded pickup in front of you) can come up with.

But it will be great for elderly people who lost their license, and for allowing people to drink heavily while out. Beyond that group, how about stricter standards and training for attaining a driver's license in the first place? I think the government should subsidize all of us going to the Bob Bondurant school of high performance driving as many times as we want ;o)

randomguy
Posts: 8509
Joined: Wed Sep 17, 2014 9:00 am

Re: Will cars be autonomous in three years?

Post by randomguy » Mon Apr 24, 2017 11:43 pm

Tamales wrote:
randomguy wrote:You can look at how self driving cars have progressed since the first Darpa grand challenge in 2004 to now and the progress is totally insane. We are probably like 95%+ of the way to the self driving car. The hard part is figuring out is that last 5% solveable or not. Deep learning over the last 5 years has done an incredible job attacking problems that in the past were not considered solvable. How close we are to the limits of it is hard to say.
I'm not sure that, collectively, all these individually developed proprietary systems and algorithms from many different players is a good measure of the progress since they are essentially all developing their technology in a silo, rather than a combined effort toward an end goal with shared experiences on fault conditions. They will all have different weaknesses and things that they don't know that that don't know. How much gets swept under the rug versus actually fixed is anybody's guess. But rest assured, they won't all be equivalent in how rigorously tested they are.

These systems will never be able to account for every permutation that nature (e.g. rock slide 10 seconds before you reach a mountain corner, or animal in the road) or location (e.g. road construction where you are diverted through cones, or fallen power lines) or random fools on the road (chair falls off the overloaded pickup in front of you) can come up with.

But it will be great for elderly people who lost their license, and for allowing people to drink heavily while out. Beyond that group, how about stricter standards and training for attaining a driver's license in the first place? I think the government should subsidize all of us going to the Bob Bondurant school of high performance driving as many times as we want ;o)
It is a sign that tech has reached a feasibility state when you have a dozen different people (i.e. including a guy in a garage who was hacking his acura) coming up with systems that do a reasonable job of driving. To some extent it is a sign of maturity in the market when you go from 1 company pushes the tech to a dozen all thinking it is feasible.

If autonomous cars miss 100% of all rock slides that happened 10s before, how many accidents would be caused?:) There will be accidents in self driving cars. The world isn't remotely deterministic to avoid all of them. Things like things in the road (doesn't matter if it is an electrical cable or a chair) can be handled. Road construction can be handled. Something to think about: you are driving a long and the random fool drops a chair in front of your car, what do you do? Do you think you have remotely enough time to figure out if the best course of action is to swerve left, right, brake, or hit the chair at speed? The computers ability to factor all that stuff in a fraction of a second is something you can not match. Granted we are not close to machines making these choices. The break through with deep learning is that the machines do a bunch of generalizations to avoid having to enumearate a million cases. You might only have to handle 10k:)

Improving driving isn't an either/or with self driving cars. It is an orthogonal issue. I doubt that making the driving standards tougher is going to catch on any time soon no mater how good of idea it would be.

It is hard to evaluate where we are because we get limited data. Musk likes to brag about how safe tesla autopilot is but we don't remotely know how the miles autopilot drives on compare to the same miles driven by model s's manually. Waymo talks about only needing user intervention 1 time per 5k miles driven but again we don't know what exactly those miles are. 1 million miles in mountain view isn't the same a 100k miles in chicago, boston, NYC, random midwest town or two, and so on where you see radically different road and weather conditions. I know this is tech I will let someone else beta test for a half dozen years:)

Iridium
Posts: 541
Joined: Thu May 19, 2016 10:49 am

Re: Will cars be autonomous in three years?

Post by Iridium » Mon Apr 24, 2017 11:48 pm

Jimmie wrote:From an engineering standpoint, why the jump from every car being a beat-up Ford Fiesta to a full-blown autonomous car? This needs to be an evolution of technology with improvements made as both the technology improves and society learns to use it and embrace it.
I think there are two ways to answer this:

1. They already ARE doing an evolution of improvements. The car I bought last year (Sonata) alerts me when my car leaves its lane, that a vehicle is coming from the side as I back up, that a vehicle is in my blind spot when I have my turn signal on, or if I am about to collide with a vehicle in front of me. The last one has already proven so effective at reducing the number and severity of accidents that the IIHS mandates it to get their 'Top Safety Pick' label. The car I owned before it didn't even have ABS. My car also has, as a convenience feature, 'adaptive cruise control' which means that it works the accelerator and brakes while I steer. This feature has undergone quite a bit of evolution over the last few years; it used to be limited to above 30mph, then limited to above 15mph. The latest technology allows it to adapt all the way down to stop and go traffic. The most recent innovations (which my car doesn't have) are the car automatically hitting the brakes after the driver has ignored the frontal collision warning, giving a gentle nudge to the steering wheel to help the driver stay in the middle of his/her lane, and the frontal collision system recognizing pedestrians and bikes. The technology is gradually but continually getting more sophisticated and ubiquitous (seems that every time a car undergoes a redesign, the new model supports these features). While it is true that the Ford Focus doesn't have any of these features, the step-up model, the Focus, does offer most of these features as an option. Volvo actually makes them standard.

2. As the technology improves, drivers will naturally become less engaged from the driving process. Humans are not well setup to remain on high-alert, but doing nothing, for hours at a time. Teslas, which can handle both the accelerator and the steering wheel, are probably already too intelligent for some drivers to operate safely. On the highway, drivers can get away with so little engagement with the driving process that there are video clips of people watching movies while 'driving'. That is, they can get away with it until it leads to an accident, because Tesla cars are not, in fact, self-driving vehicles (as it repeatedly reminds you). So, for as long as cars are too dumb to safely operate without an attentive driver, there is probably a limit to how smart they can get. I am confident that we'll cross that chasm at some point, but I think it is a pretty large chasm.
Tamales wrote:These systems will never be able to account for every permutation that nature (e.g. rock slide 10 seconds before you reach a mountain corner, or animal in the road) or location (e.g. road construction where you are diverted through cones, or fallen power lines) or random fools on the road (chair falls off the overloaded pickup in front of you) can come up with.
I actually think this is the very most exciting thing about autonomous vehicles. Part of what makes air travel so safe is that literally every single civil aviation accident is investigated by the NTSB. The result is a continually expanding base of knowledge of possible problems and safety measures to counteract them. It is looking like the autonomous era will get something at least somewhat similar. Weymo (Google) issues to CA a report on the cause of every single collision its vehicles are involved in. For the very rare incidents when the car could have done something to prevent the collision (so far, the vast majority of collisions have been the car getting rear-ended from behind while stopped at a red light), I have little doubt that the engineers have poured over the data to figure out how to improve their software and sensors. It would be impossible to train humans with every single possible situation on the road. However, it is quite possible for every new autonomous driving car to be virtually tested against every avoidable accident an autonomous vehicle has ever been involved in. So, yes, the first autonomous vehicle might not handle the mudslide well, but after the investigation and patch release, every autonomous vehicle would handle it expertly.
Last edited by Iridium on Tue Apr 25, 2017 12:09 am, edited 1 time in total.

app1
Posts: 10
Joined: Wed Nov 16, 2016 5:19 pm

Re: Will cars be autonomous in three years?

Post by app1 » Tue Apr 25, 2017 12:02 am

I predict by 2030 most cars will have the option for you to get on the highway and "put it into autopilot". I for one am looking forward to the day when I can get into my autonomous airstream interstate in Texas at 8pm, drive to the highway, get into bed, and end up at the ski resort the next morning instead of going to the airport.

wolf359
Posts: 1957
Joined: Sun Mar 15, 2015 8:47 am

Re: Will cars be autonomous in three years?

Post by wolf359 » Tue Apr 25, 2017 8:48 am

Tamales wrote:
vitaflo wrote:People underestimate the speed of tech advancement. 10 years ago the iPhone didn't exist, and Facebook had just opened up the site to the public for the first time. 20 years ago less than 2% of people used the internet and only 4% had cell phones. I wouldn't underestimate the speed of tech change.
No comparison between iphone/facebook and autonomous driving.

First, iphone (and smartphones in general) wasn't a technology advancement per se. It was integration of existing pieces from the same platform or adjacent platforms. It was more of a packaging and marketing exercise.
And facebook isn't a technology or an advancement, but that a whole 'nuther discussion.

Autonomous driving on the other hand, requires new technology to be invented, adapted, and improved; lather, rinse, repeat. And since it involves safety of life and property, it will be many generations. The closest analogy would probably be air bags or ESC or ABS, but even those are a drop in the bucket compared to the scale and scope of full autonomous driving.

These types of things have to be mandated by law or they aren't likely to even reach 50% adoption. That will surely come, but it will be in many incremental steps over decades, before the "humanless carriage" affects the masses.

Right now, the nearest-term discussion (which we can't have here) involves the DoT's multi-year interest in safety-related communications between cars and the related DSRC mandate (which may or may not go into effect).

PS, actually "people" way OVERestimate the speed of technology advancement (speaking of true technology advancement, not integration and marketing posing as technology advancement). Analysts (who are in business of selling hype) apply a hockey-stick curve to everything, and people buy into it. There are entire books written about this.
The DARPA Grand Challenge was held to spur autonomous vehicle technology. The first race of 100% self-driving vehicles was held on March 13, 2004, in the Mohave Desert. None of the robot vehicles finished the race. The best robot only got about 7 miles into a 150 mile course.

In 2005, only one year later, five vehicles completed the course. 22 of the 23 finalists bested the 7 mile record set only a year before.

In 2007, DARPA Grand Challenge was renamed the DARPA Urban Challenge, and the vehicles were required to navigate a simulated city environment. Six cars completed the course.

It is 2017, a little over ten years later. There are commercially shipping partially autonomous vehicles on the road today.

That is pretty fast technological change, especially considering it wasn't practical in 2003.

themesrob
Posts: 261
Joined: Fri Jul 15, 2016 1:58 pm

Re: Will cars be autonomous in three years?

Post by themesrob » Tue Apr 25, 2017 9:22 am

randomguy wrote: If autonomous cars miss 100% of all rock slides that happened 10s before, how many accidents would be caused?:) There will be accidents in self driving cars. The world isn't remotely deterministic to avoid all of them. Things like things in the road (doesn't matter if it is an electrical cable or a chair) can be handled. Road construction can be handled. Something to think about: you are driving a long and the random fool drops a chair in front of your car, what do you do? Do you think you have remotely enough time to figure out if the best course of action is to swerve left, right, brake, or hit the chair at speed? The computers ability to factor all that stuff in a fraction of a second is something you can not match. Granted we are not close to machines making these choices. The break through with deep learning is that the machines do a bunch of generalizations to avoid having to enumearate a million cases. You might only have to handle 10k:)
I agree 100% that the computers can handle the evaluations more quickly/accurately on what the best course of action is situation-to-situation. My hesitation with all this is that the rules for the computers' decisioning situation-to-situation will be set by humans, and I'm not sure how those rules will be agreed on -- for example, how the calculation of physical damage to the car versus risk to others will be set. In the case of the chair in the road (or a pothole, or a mudslide), driving and hitting the chair at speed has a 100% chance of damaging my car, but I assume arguendo that braking or swerving right/left will increase the chance of my car causing an accident or hitting a pedestrian by .1%. Will an increase of that risk be allowed to save me having to take my car in for an alignment after every pothole, or a paint job for every chair? How much of an increase will be allowed? If the answer is some quantity of risk, then if my car does cause an accident after taking that .1% chance, who's liable -- me or the company who set the parameters? If the answer is zero, would anyone buy a car that is unable/unwilling to dodge a pothole?

I just see a lot of issues down the line that seem difficult to resolve. (And that's after developing sensors that can detect potholes!)

wolf359
Posts: 1957
Joined: Sun Mar 15, 2015 8:47 am

Re: Will cars be autonomous in three years?

Post by wolf359 » Tue Apr 25, 2017 9:27 am

crazygrow wrote:Will most cars become autonomous drivers over the next three years? If so, should I LEASE my truck for three years and then buy an autonomous car then? I see that an explosion in car-related technology is in process and I expect major changes over the next three years.
To answer the OP, no, most cars will not become autonomous drivers in the next three years.

To do so the following things have to happen:

- All states and jurisdictions have to have laws that allow autonomous vehicles. It takes time to update and change laws, probably more than 3 years.
- Autonomous technology has to become accepted by consumers. It has not yet achieved this. Many consumers are leery.
- Autonomous technology has to be demanded by consumers. At that point, manufacturers will start putting it into cars in general, as opposed to only one or two models. Costs would have to also come down to a point to make this feasible.
- Consumers would have to replace their existing vehicles. The average age of a vehicle is over ten years old. Even if people want it and demand it, it would probably take 10-15 years for a new technology to be widely deployed in the fleet.

However, you will probably see lots of changes in vehicle technology in the next 3-5 years. You probably will be able to buy your self-driving car at that point. I plan to just keep holding onto my existing car as long as I can, then see what technology is available. I don't mind if other people work out the kinks of the new technology before I get it.

If you want a truck, buy a truck. Hold it for the normal 10-15 years. By the time you're switching again, driverless cars will be the norm. I wouldn't lease, because there's no guarantees that the technology will come through and be ready if you set a deadline like 3 years. You could always just sell the truck and switch earlier than 10 years if you decide you like something else better.

bubbly
Posts: 68
Joined: Tue Jun 16, 2015 4:21 pm

Re: Will cars be autonomous in three years?

Post by bubbly » Tue Apr 25, 2017 9:34 am

randomguy wrote:
bubbly wrote:To expound on my previous points: Any other car manufacturer that I'm aware of other than Tesla, only offers adaptive cruise. Tesla is the only manufacturer that offers steering at high speeds, dubbed lane centering technology. EDIT: it seems like beginning this year, some non-Tesla vehicles will have relatively primitive lane keeping technology that steers the vehicle in certain situations.
Mercedes has had active lane keeping (I thought it was 150km/hr back then. The new one does 200km/hr) since at least 2012 and adaptive cruise control during the same period. They didn't give it a fancy name. They gave it some weird german one:) Pretty sure Volvo did also and probably a bunch of other ones. Here is a caranddriver from 2016 http://www.caranddriver.com/features/se ... ti-feature . Now the last is vastly outdated (volvo, audi, and tesla have all new systems. Pretty sure MB updated theirs for 2017 (2018 models) also). With tech like this it isn't a binary choice. It isn't like you have active lane keeping or your don't. Once you have it you have to figure out how well it works.

The one thing that tesla has done really well is update your system. Maybe the other manufactures were doing it also and not telling you but I sort of doubt it. How this plays out as tesla transitions from using mobileye technology to their own will be fun to watch. I find it hard to believe ANY oem is going to be doing a good job updating these systems when they get to 10+ years old.
From what I've read, Mercede's lane keeping technology back then did not steer the car, only warned you when you were straying from the lane through braking or alarms/vibrations of the wheel. The difference in the technology now is that the car can actively steer the vehicle without user input, which is a huge step forward towards autonomous driving. As far as I've researched, only a few cars have started doing it this year and Tesla is obviously the most advanced commercial product out right now.

EDIT: It does look like active steering during cruising has been available for 2-3 years already.
Last edited by bubbly on Tue Apr 25, 2017 10:27 am, edited 1 time in total.

Tamales
Posts: 1423
Joined: Sat Jul 05, 2014 10:47 am

Re: Will cars be autonomous in three years?

Post by Tamales » Tue Apr 25, 2017 9:42 am

themesrob wrote:I agree 100% that the computers can handle the evaluations more quickly/accurately on what the best course of action is situation-to-situation. My hesitation with all this is that the rules for the computers' decisioning situation-to-situation will be set by humans, and I'm not sure how those rules will be agreed on -- for example, how the calculation of physical damage to the car versus risk to others will be set. In the case of the chair in the road (or a pothole, or a mudslide), driving and hitting the chair at speed has a 100% chance of damaging my car, but I assume arguendo that braking or swerving right/left will increase the chance of my car causing an accident or hitting a pedestrian by .1%. Will an increase of that risk be allowed to save me having to take my car in for an alignment after every pothole, or a paint job for every chair? How much of an increase will be allowed? If the answer is some quantity of risk, then if my car does cause an accident after taking that .1% chance, who's liable -- me or the company who set the parameters? If the answer is zero, would anyone buy a car that is unable/unwilling to dodge a pothole?

I just see a lot of issues down the line that seem difficult to resolve. (And that's after developing sensors that can detect potholes!)
This is an excellent point (that in any given potential collision scenario there are several possible avoidance action sequences that could be taken, each with different consequences (or damage results/costs)).

You are at the mercy of the algorithm developers, but also, so are the other "innocent bystander" vehicles/people/property around you (and they didn't choose to be). And as I mentioned earlier, since the systems are proprietary, they won't all choose the same path or have the same outcome.

So the question isn't just *whether* people trust autonomous driving, it's *which* particular system. That's just not something individuals will be able to judge.

wolf359
Posts: 1957
Joined: Sun Mar 15, 2015 8:47 am

Re: Will cars be autonomous in three years?

Post by wolf359 » Tue Apr 25, 2017 9:42 am

themesrob wrote:
randomguy wrote: If autonomous cars miss 100% of all rock slides that happened 10s before, how many accidents would be caused?:) There will be accidents in self driving cars. The world isn't remotely deterministic to avoid all of them. Things like things in the road (doesn't matter if it is an electrical cable or a chair) can be handled. Road construction can be handled. Something to think about: you are driving a long and the random fool drops a chair in front of your car, what do you do? Do you think you have remotely enough time to figure out if the best course of action is to swerve left, right, brake, or hit the chair at speed? The computers ability to factor all that stuff in a fraction of a second is something you can not match. Granted we are not close to machines making these choices. The break through with deep learning is that the machines do a bunch of generalizations to avoid having to enumearate a million cases. You might only have to handle 10k:)
I agree 100% that the computers can handle the evaluations more quickly/accurately on what the best course of action is situation-to-situation. My hesitation with all this is that the rules for the computers' decisioning situation-to-situation will be set by humans, and I'm not sure how those rules will be agreed on -- for example, how the calculation of physical damage to the car versus risk to others will be set. In the case of the chair in the road (or a pothole, or a mudslide), driving and hitting the chair at speed has a 100% chance of damaging my car, but I assume arguendo that braking or swerving right/left will increase the chance of my car causing an accident or hitting a pedestrian by .1%. Will an increase of that risk be allowed to save me having to take my car in for an alignment after every pothole, or a paint job for every chair? How much of an increase will be allowed? If the answer is some quantity of risk, then if my car does cause an accident after taking that .1% chance, who's liable -- me or the company who set the parameters? If the answer is zero, would anyone buy a car that is unable/unwilling to dodge a pothole?

I just see a lot of issues down the line that seem difficult to resolve. (And that's after developing sensors that can detect potholes!)
I think the safety issues will be worked out. They're an obvious requirement. I think there will be unexpected squishier issues.

One of the biggest challenges with a self-driving vehicle may be the fact that humans will be riding in them, and that the cars won't drive the way the humans drive. Humans run through yellow lights, tailgate, pass on the right, drive quickly and even accelerate when approaching stop signs and red lights, then brake at the last minute, drive 5-10 mph (or more) over the speed limit, and don't wait for pedestrians to clear the crosswalk completely before driving through. The self-driving vehicles will follow the laws because if there is an accident when the vehicle WASN'T obeying the law, the manufacturer will be held liable.

The first time the driverless car won't move while Granny is walking to the other side of the street but still in the crosswalk, and the cars behind you are honking, people are going to get frustrated. And when the car starts driving like Granny on a Sunday morning, they may be safe, but they're also likely to turn autopilot off.

michaeljc70
Posts: 6010
Joined: Thu Oct 15, 2015 3:53 pm

Re: Will cars be autonomous in three years?

Post by michaeljc70 » Tue Apr 25, 2017 9:44 am

No. It will probably be more like 10 years. It is very complicated navigating roads under construction, detours, other special situations (police roadblock) and interpreting all signs (no turn on red 7am to 7pm, not left turn 4pm-6pm, 20 mph while kids are in school). That is not even getting into things like parking in parking garage, parking on the street (knowing the regulations) and things like that.

In 3 years it might be able to handle 90-95% of what a driver comes across. If you are just looking for it to help you out, it might be there in 3 years. If you are looking for it to do everything (like you are not even in the car sending it to pick someone up or you don't have a drivers license), then it is going to be a while.

Another issue I've read about is should an autonomous car break the law (like speeding up) to prevent an accident? I think most people would say yes, but will regulators allow it? Regulation is WAY behind the curve on autonomous cars.
Last edited by michaeljc70 on Tue Apr 25, 2017 10:05 am, edited 1 time in total.

Tamales
Posts: 1423
Joined: Sat Jul 05, 2014 10:47 am

Re: Will cars be autonomous in three years?

Post by Tamales » Tue Apr 25, 2017 10:00 am

bubbly wrote:As far as I've researched, only a few cars have started doing [lane keeping assist] this year and Tesla is obviously the most advanced commercial product out right now.
Doing a quick search, there are apparently many cars with some form of lane keeping assist (or lane departure warning; probably goes by other names too). There are even top 10 lists on this.

I'm not sure why it's 'obvious' that Tesla has the most advanced product. How could that even be judged (not that that will stop media from judging, and people in turn repeating what they read, as fact). For all we know, the Cadillac or the Honda or the Toyota or the Lincoln or the Mercedes, etc., etc., may have the most advanced product for this particular feature. I guess it depends how you define 'advanced,' and who is doing the judging.

themesrob
Posts: 261
Joined: Fri Jul 15, 2016 1:58 pm

Re: Will cars be autonomous in three years?

Post by themesrob » Tue Apr 25, 2017 10:00 am

wolf359 wrote: I think the safety issues will be worked out. They're an obvious requirement. I think there will be unexpected squishier issues.

One of the biggest challenges with a self-driving vehicle may be the fact that humans will be riding in them, and that the cars won't drive the way the humans drive. Humans run through yellow lights, tailgate, pass on the right, drive quickly and even accelerate when approaching stop signs and red lights, then brake at the last minute, drive 5-10 mph (or more) over the speed limit, and don't wait for pedestrians to clear the crosswalk completely before driving through. The self-driving vehicles will follow the laws because if there is an accident when the vehicle WASN'T obeying the law, the manufacturer will be held liable.

The first time the driverless car won't move while Granny is walking to the other side of the street but still in the crosswalk, and the cars behind you are honking, people are going to get frustrated. And when the car starts driving like Granny on a Sunday morning, they may be safe, but they're also likely to turn autopilot off.
I think they can be worked out, but the solutions will not be easy, and possibly hurt marketability. But to your point, the Grant's article I mentioned above had some quotes from John Adams (author of "Risk," as well as this article: http://www.john-adams.co.uk/2016/08/17/ ... w-problem/), who hypothesized that the deference that will undoubtedly be programmed into driverless cars will change the behavior of bicyclists and pedestrians, who will walk right into the roads or ride in packs because they know traffic will stop for them.

bubbly
Posts: 68
Joined: Tue Jun 16, 2015 4:21 pm

Re: Will cars be autonomous in three years?

Post by bubbly » Tue Apr 25, 2017 10:12 am

Tamales wrote:
bubbly wrote:As far as I've researched, only a few cars have started doing [lane keeping assist] this year and Tesla is obviously the most advanced commercial product out right now.
Doing a quick search, there are apparently many cars with some form of lane keeping assist (or lane departure warning; probably goes by other names too). There are even top 10 lists on this.

I'm not sure why it's 'obvious' that Tesla has the most advanced product. How could that even be judged (not that that will stop media from judging, and people in turn repeating what they read, as fact). For all we know, the Cadillac or the Honda or the Toyota or the Lincoln or the Mercedes, etc., etc., may have the most advanced product for this particular feature. I guess it depends how you define 'advanced,' and who is doing the judging.

You judge the system based on the features offered and the complexity of the tasks that can be performed by the autonomous system. I don't understand how we turned "advanced" into a semantics debate but how do you quantify that a S class Mercedes is a "better" car than a C class. Obviously, for some people, it might not be but the options and packages offered by the S class can be listed out to show that it's more comprehensive than a C class. Same can be done with autonomous driving features. The 6 levels of autonomy defined by SAE is a start.

This car and driver article linked actually breaks down the systems and rates them against each other:

http://www.caranddriver.com/features/se ... lts-page-6
Last edited by bubbly on Tue Apr 25, 2017 10:15 am, edited 1 time in total.

wolf359
Posts: 1957
Joined: Sun Mar 15, 2015 8:47 am

Re: Will cars be autonomous in three years?

Post by wolf359 » Tue Apr 25, 2017 10:15 am

themesrob wrote: I think they can be worked out, but the solutions will not be easy, and possibly hurt marketability. But to your point, the Grant's article I mentioned above had some quotes from John Adams (author of "Risk," as well as this article: http://www.john-adams.co.uk/2016/08/17/ ... w-problem/), who hypothesized that the deference that will undoubtedly be programmed into driverless cars will change the behavior of bicyclists and pedestrians, who will walk right into the roads or ride in packs because they know traffic will stop for them.
Interesting article.

Not ALL traffic will stop for pedestrians and bikes. Hitting the point where 100% of the cars is driverless will probably not happen for decades, if ever, especially if owners have the option to switch autopilot off. And that wild car who is still subject to road rage will keep the pedestrians off the interstates.

Pedestrians have right-of-way over cars, even today. But, as my Daddy always said, "There's right, and there's dead right."

User avatar
bengal22
Posts: 1858
Joined: Sat Dec 03, 2011 6:20 pm
Location: Ohio

Re: Will cars be autonomous in three years?

Post by bengal22 » Tue Apr 25, 2017 10:18 am

I would be afraid to get in a car that was autonomous. I would at least want to direct the car's destination and duration of drive.
"Earn All You Can; Give All You Can; Save All You Can." .... John Wesley

User avatar
flamesabers
Posts: 1807
Joined: Fri Mar 03, 2017 12:05 pm
Location: Rochester, MN

Re: Will cars be autonomous in three years?

Post by flamesabers » Tue Apr 25, 2017 10:23 am

wolf359 wrote:One of the biggest challenges with a self-driving vehicle may be the fact that humans will be riding in them, and that the cars won't drive the way the humans drive. Humans run through yellow lights, tailgate, pass on the right, drive quickly and even accelerate when approaching stop signs and red lights, then brake at the last minute, drive 5-10 mph (or more) over the speed limit, and don't wait for pedestrians to clear the crosswalk completely before driving through. The self-driving vehicles will follow the laws because if there is an accident when the vehicle WASN'T obeying the law, the manufacturer will be held liable.
What if pedestrians are jaywalking though? Will the automated car stop and wait for the pedestrians to cross the road? Or will the automated car not pay any attention to the actions of pedestrians crossing the road unlawfully?

inbox788
Posts: 6694
Joined: Thu Mar 15, 2012 5:24 pm

Re: Will cars be autonomous in three years?

Post by inbox788 » Tue Apr 25, 2017 10:23 am

themesrob wrote:I agree 100% that the computers can handle the evaluations more quickly/accurately on what the best course of action is situation-to-situation. My hesitation with all this is that the rules for the computers' decisioning situation-to-situation will be set by humans, and I'm not sure how those rules will be agreed on -- for example, how the calculation of physical damage to the car versus risk to others will be set. In the case of the chair in the road (or a pothole, or a mudslide), driving and hitting the chair at speed has a 100% chance of damaging my car, but I assume arguendo that braking or swerving right/left will increase the chance of my car causing an accident or hitting a pedestrian by .1%. Will an increase of that risk be allowed to save me having to take my car in for an alignment after every pothole, or a paint job for every chair? How much of an increase will be allowed? If the answer is some quantity of risk, then if my car does cause an accident after taking that .1% chance, who's liable -- me or the company who set the parameters? If the answer is zero, would anyone buy a car that is unable/unwilling to dodge a pothole?

I just see a lot of issues down the line that seem difficult to resolve. (And that's after developing sensors that can detect potholes!)
It goes deeper than that. Turning over that much control to a preprogrammed machine questions our fate and free will.

There are conditions where there are ethical dilemmas we face that just can't be programmed. So even if we had the data, which we don't and won't, and even if we could act in time, which is unlikely, the minimum harm outcome is undefined. If the car become all seeing and all knowing, under what conditions if any would you allow the driver to take the grenade and sacrifice himself for the greater good or decide on which harm is lesser evil?

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

IMO, it will be a quite a while before we're comfortable with autonomous cars the way we are with elevators.

http://www.nytimes.com/2011/04/29/nyreg ... -york.html

I think it was another thread where I saw this funny cartoon: https://www.explainxkcd.com/wiki/images ... riving.png

And like the early GPS days, always be sure to check your destination, for example, if you're going to Albany: https://goo.gl/maps/uv5EtRxcxB42
Last edited by inbox788 on Tue Apr 25, 2017 10:37 am, edited 1 time in total.

tech_arch
Posts: 253
Joined: Wed May 27, 2015 11:47 am

Re: Will cars be autonomous in three years?

Post by tech_arch » Tue Apr 25, 2017 10:30 am

Waymo (Google) is about to start a new trial of their self-driving Pacificas in Arizona. Three years is still probably optimistic given regulatory concerns, but I wouldn't be surprised if the tech is ready within five years.

https://medium.com/waymo/apply-to-be-pa ... d996c7a86f

Locked