Google Self-Driving Car T-Boned By Interstate Delivery Truck

[Megalith]

A Google self-driving Lexus was in “one of the worst” accidents that any of their cars have ever been in. A van ran a red light and struck the car, even though the light was green for the Google vehicle at least six seconds before it entered the intersection.

Specifically, we’re told the crash happened in Mountain View, California, at the corner of W El Camino Real and Calderon Ave. We’re told that no one was hurt in the crash and all airbags deployed. “I only saw the tail-end of the crash, and the dazed Google employees sitting around afterwards waiting for their tow-truck. I had to be on my way,” a witness told us. The witness also mentioned that, based on their perspective, the self-driving car was not at fault. “From what I could see, it was the van’s fault entirely,” they said.

 

[SlayVus]

Probably one reason why self driving cars won't be completely safe. Until all vehicles on the road are self driving, there will always be a chance for someone else to hit you.

 

[THRESHIN]

Probably one reason why self driving cars won't be completely safe. Until all vehicles on the road are self driving, there will always be a chance for someone else to hit you.

that's the same with or without self driving cars in any number combination. kind of a moot point to make.

 

[cdabc123]

that's the same with or without self driving cars in any number combination. kind of a moot point to make.

does a self driving car monitor a situation the same way a human does. can it see a car not slowing down and change is path to lessen the accident? Can it notice "unsafe" driving and take actions to avoid that person? if every car has a predictable action its pretty easy to have a minimal amount of crashes.

 

[Jim Kim]

I blame the Tesla.

 

[Kor]

does a self driving car monitor a situation the same way a human does. can it see a car not slowing down and change is path to lessen the accident? Can it notice "unsafe" driving and take actions to avoid that person? if every car has a predictable action its pretty easy to have a minimal amount of crashes.

If anything the self driving car would be better at it.

 

[M76]

If anything the self driving car would be better at it.

But, but, but, that would be oppressive to the people who the system deems unsafe!

 

[raz-0]

If anything the self driving car would be better at it.

prove it.

Right now, if you calculate accidents per mile for the google self driving fleet, vs the average human, you wind up with ~ 1 accident for every 98k driven for the robot, and one accident for every ~100-110k for the human population.

Keep in mind that the gogle fleet doesn't have to deal with

snow
heavy rain
seriously pot holed streets
drunk driving
falling asleep
distracted driving
two wheeled vehicles that are extremely sensitive to fuck ups and traction problems.

And it is still not quite up to the human population.

It's not done yet, and given what should be benefits for the robot vs. human population, kind of sucks at the driving part. It'll likely get there, but it's not there.

 

[Wierdo]

Another score against human drivers.

 

[thesmokingman]

Biggest driving threat are humans, yet again.

 

[Biznatch]

prove it.

Right now, if you calculate accidents per mile for the google self driving fleet, vs the average human, you wind up with ~ 1 accident for every 98k driven for the robot, and one accident for every ~100-110k for the human population.

Keep in mind that the gogle fleet doesn't have to deal with

snow
heavy rain
seriously pot holed streets
drunk driving
falling asleep
distracted driving
two wheeled vehicles that are extremely sensitive to fuck ups and traction problems.

And it is still not quite up to the human population.

It's not done yet, and given what should be benefits for the robot vs. human population, kind of sucks at the driving part. It'll likely get there, but it's not there.

And how many of those accidents were CAUSED by the google self driving car, and not a human driver not paying attention? For those stats to have any meaning, it would need to be accidents where the google car was at fault vs human drivers. In your statistics, the accidents in both categories were caused by human drivers...

 

[Riouken]

prove it.

Right now, if you calculate accidents per mile for the google self driving fleet, vs the average human, you wind up with ~ 1 accident for every 98k driven for the robot, and one accident for every ~100-110k for the human population.

Keep in mind that the gogle fleet doesn't have to deal with

snow
heavy rain
seriously pot holed streets
drunk driving
falling asleep
distracted driving
two wheeled vehicles that are extremely sensitive to fuck ups and traction problems.

And it is still not quite up to the human population.

It's not done yet, and given what should be benefits for the robot vs. human population, kind of sucks at the driving part. It'll likely get there, but it's not there.

In statistics it is very hard to take a small data set like the google driving record and extrapolate or compare it to the overall human driving record. This is the same problem polls have in politics.

It would be a much more fair comparison if you had a data set of only the same geographical area that google drives in. Without that, and comparing a small data set to a much much larger data set you have to take a lot of statistical liberties and "guesses" to come up with a conclusion. This is the same reason so many political polls are wrong. Gallup gives up the horse race

 

[SBSuperfly]

And how many of those accidents were CAUSED by the google self driving car, and not a human driver not paying attention? For those stats to have any meaning, it would need to be accidents where the google car was at fault vs human drivers. In your statistics, the accidents in both categories were caused by human drivers...

Think about your argument, and keep thinking until you realize you're wrong. Both human drivers and self-driving cars have to avoid bad human drivers. Preventing accidents that wouldn't technically be your fault is all part of the game. If you have a green light, you still need to stop for assholes and emergency vehicles.

 

[Biznatch]

Think about your argument, and keep thinking until you realize you're wrong. Both human drivers and self-driving cars have to avoid bad human drivers.

Huh? Most of the google accidents were from being rear ended by a human driver. Guess the car should have jumped over them to avoid it right?.... So no, what I said is completely true. You will never be able to build something that can avoid all of the stupid distracted drivers on the road. So until they are completely gone, you will never have an autonomous vehicle with a perfect driving record.

 

[Archaea]

that's the same with or without self driving cars in any number combination. kind of a moot point to make.

I've avoided many accidents in my 21 years of driving by bending straight line rules. Id like to think I'd have had a chance to avoid or minimize a accident like this by paying attention to my surroundings at an intersection. This is the same reason I won't buy a motorcycle. You might be a safe motorcyclist but you aren't on the road alone.

I agree with the post that these automated cars won't be safe until all cars follow the same automated rules.

 

[amddragonpc]

Car batteries are heavy. Hope there was a wall between the batteries and the van driver. If not, probably has a severe backache and headache.

 

[NickJames]

I've avoided many accidents in my 21 years of driving by bending straight line rules. Id like to think I'd have had a chance to avoid or minimize a accident like this by paying attention to my surroundings at an intersection. This is the same reason I won't buy a motorcycle. You might be a safe motorcyclist but you aren't on the road alone.

I agree with the post that these automated cars won't be safe until all cars follow the same automated rules.

Same, the problem with automated cars if they follow the rules too well. I drive with the idea that everyone else on the road is a complete idiot so I tend to avoid any situation that can put me in harms way.

Making a left in the middle of the intersection, lights turning yellow, hmm that guy across from me is flying too fast and will probably blow through the red light, I'll wait.
That guy behind me is about the ram the shit out of me because he's not paying attention, let me flash the brakes a bit or move forward.
This woman just blew through the stop sign and cut me off, I'll have to swerve, incoming car in left lane, I'll swerve right onto this homes driveway to avoid the collision.
Semi is changing lanes right on top of me, better start honking and slowing down or speeding up depending on my position so I don't die.

Way too many possibilities for accidents that automated cars are just not prepared for.

 

[SBSuperfly]

Huh? Most of the google accidents were from being rear ended by a human driver. Guess the car should have jumped over them to avoid it right?.... So no, what I said is completely true. You will never be able to build something that can avoid all of the stupid distracted drivers on the road. So until they are completely gone, you will never have an autonomous vehicle with a perfect driving record.

What you said specifically-- "For those stats to have any meaning, it would need to be accidents where the google car was at fault vs human drivers." All accident involve at least 1 bad human driver, so the stats are totally legit. Automated cars and human drivers both have to deal with it. The stats show humans are better at avoiding bad humans.

 

[britjh22]

prove it.

Right now, if you calculate accidents per mile for the google self driving fleet, vs the average human, you wind up with ~ 1 accident for every 98k driven for the robot, and one accident for every ~100-110k for the human population.

Keep in mind that the gogle fleet doesn't have to deal with

snow
heavy rain
seriously pot holed streets
drunk driving
falling asleep
distracted driving
two wheeled vehicles that are extremely sensitive to fuck ups and traction problems.

And it is still not quite up to the human population.

It's not done yet, and given what should be benefits for the robot vs. human population, kind of sucks at the driving part. It'll likely get there, but it's not there.

Do you have the source for that 100-110k number? While you are correct that Google/Uber/Anyone else's self driving fleet isn't yet able to deal with all eventualities, I think that 100-110k number probably isn't fully accurate. The more interesting comparison would just to look at California drivers, or drivers in the same counties as the Google cars operate. Also, you can bet Google is reporting every single accident for obvious reasons, how many minor fender benders and such don't get reported? I would assume the human statistic is derived from insurance information, and plenty of people don't report minor incidents due to it not making financial sense (increased insurance costs).

 

[Kor]

prove it.

Right now, if you calculate accidents per mile for the google self driving fleet, vs the average human, you wind up with ~ 1 accident for every 98k driven for the robot, and one accident for every ~100-110k for the human population.

Keep in mind that the gogle fleet doesn't have to deal with

snow
heavy rain
seriously pot holed streets
drunk driving
falling asleep
distracted driving
two wheeled vehicles that are extremely sensitive to fuck ups and traction problems.

And it is still not quite up to the human population.

It's not done yet, and given what should be benefits for the robot vs. human population, kind of sucks at the driving part. It'll likely get there, but it's not there.

I figured the reasoning would be self evident but anyway.

A self driving vehicle has complete situational awareness at all times, it views and monitors any and all objects within range in a full 360 degrees around the vehicle with no blind spots. Humans are completely incapable of the same level of situational analysis, we simply can't observe and process the same amount of information in real time. The computer has real measurable data about the speed and distance of the incoming vehicle, any existing environmental hazards, and a full knowledge of the safest ways to avoid injury to the passengers on board. The only variable the computer is going to have a problem accounting for is how the sentient pile of meat behind the wheel of the other car is going to react, but to be fair neither would a human driver.

 

[HeavensCloud]

Could a human driving the car have seen the van barreling towards them from farther away that the Google car could have seen and reacted? How far down the road do they scan?

 

[Jagger100]

I figured the reasoning would be self evident but anyway.

A self driving vehicle has complete situational awareness at all times, it views and monitors any and all objects within range in a full 360 degrees around the vehicle with no blind spots. Humans are completely incapable of the same level of situational analysis, we simply can't observe and process the same amount of information in real time. The computer has real measurable data about the speed and distance of the incoming vehicle, any existing environmental hazards, and a full knowledge of the safest ways to e avoid injury to the passengers on board. The only variable the computer is going to have a problem accounting for is how the sentient pile of meat behind the wheel of the other car is going to react, but to be fair neither would a human driver.

Or self-driving cars are used on incredibly curated situations of low speeds and low traffic densities with no expectation of performance.

And a large part of driving is preventing accidents even when you've been dutifully following the rules because sometimes shit happens. Like a kid chasing their dog accross the road.

 

[SLee]

Do you have the source for that 100-110k number? While you are correct that Google/Uber/Anyone else's self driving fleet isn't yet able to deal with all eventualities, I think that 100-110k number probably isn't fully accurate. The more interesting comparison would just to look at California drivers, or drivers in the same counties as the Google cars operate. Also, you can bet Google is reporting every single accident for obvious reasons, how many minor fender benders and such don't get reported? I would assume the human statistic is derived from insurance information, and plenty of people don't report minor incidents due to it not making financial sense (increased insurance costs).

Google actually commissioned a study which found that the American driving population has, including unreported accidents, 4.2 accidents/million miles.

Automated Vehicle Crash Rate Comparison Using Naturalistic Data | Virginia Tech Transportation Institute

But one thing that makes comparison hard is that the Google Cars (and other so called self-driving cars) have human drivers who take over whenever a potentially dangerous situation arises, or whenever the hardware or software has a failure. But this information is only reported once a year and only after California forced them to. From 2014/09 to 2015/11, there were hundreds of such incidents and where at least 10 of them would have resulted in an at-fault collision with something.

http://static.googleusercontent.com...drivingcar/files/reports/report-annual-15.pdf

 

[M76]

Or self-driving cars are used on incredibly curated situations of low speeds and low traffic densities with no expectation of performance.

Self driving cars are in their infancy. The current devices are nor perfect, nor finished. It's like you're judging all human drivers by the performance of a student driver of his/her first day in traffic.
I'm sure by the time self driving cars will be a thing the actual performance will be the equivalent or better than the average human driver. Performance thrown out like this isn't even a metric of anything. In some aspects self driving cars perform magnitudes more efficiently than human drivers already. Like reaction times, and keeping up with the traffic flow. Human drivers always slack off in traffic constantly rubber banding to the car in front of them, which is one of the main causes for traffic congestions. (The Tesla is not a self driving car, it's a car with a glorified adaptive cruise control and lane assist)

And a large part of driving is preventing accidents even when you've been dutifully following the rules because sometimes shit happens. Like a kid chasing their dog accross the road.

How is that a large part of driving? It's about 0.000001% of the actual time driven. I think you meant the most important part of driving. But that's not an argument. Yes it's an important part of driving, and? Are you suggesting that sensors can't detect a kid chasing their dog across the road?
Let us see. A sensor works constantly without interruption without dividing attention. A human will be more slow, and might not even notice a dangerous situation until it is too late. I see that happen all the time on dashcam footage. That people are just oblivious to the signs or even the developing accident in front of them.

 

[StoleMyOwnCar]

Self driving cars are in their infancy. The current devices are nor perfect, nor finished. It's like you're judging all human drivers by the performance of a student driver of his/her first day in traffic.
I'm sure by the time self driving cars will be a thing the actual performance will be the equivalent or better than the average human driver. Performance thrown out like this isn't even a metric of anything. In some aspects self driving cars perform magnitudes more efficiently than human drivers already. Like reaction times, and keeping up with the traffic flow. Human drivers always slack off in traffic constantly rubber banding to the car in front of them, which is one of the main causes for traffic congestions. (The Tesla is not a self driving car, it's a car with a glorified adaptive cruise control and lane assist)



How is that a large part of driving? It's about 0.000001% of the actual time driven. I think you meant the most important part of driving. But that's not an argument. Yes it's an important part of driving, and? Are you suggesting that sensors can't detect a kid chasing their dog across the road?
Let us see. A sensor works constantly without interruption without dividing attention. A human will be more slow, and might not even notice a dangerous situation until it is too late. I see that happen all the time on dashcam footage. That people are just oblivious to the signs or even the developing accident in front of them.

This. Humans are limited by genetics and essentially other random factors as to how good they could possibly be at driving. Meanwhile, if anything, computers are still in their infancy (in such a complex task) yet already performing this well. If anything, I'd say this is a good sign. Give it a few years (and maybe a few technological advances), and I feel that our new automated drivers will be much better at the task than any average human (well that might already be true). In fact, it's inevitable. Especially when there are actually more automated cars on the road, and they actually start communicating WITH EACH OTHER. This is a big issue with human drivers, and that's that they essentially have no way of communicating with each other on the road aside from some lights and blinkers (which honestly there are a lot of retards out there that can't use the blinkers properly, even). Computers can actually have perfect transmission between each other as to where they're at and what their vector is. The more automated cars are out on the road, the safer each person in an automated car will be. Compare this to human drivers. The more human drivers there are out on the road, the more potential that one of them is some crazy idiot that will cause that one huge pileup on that interstate.

Essentially, I see it like this: automated cars will only get better and better as each little nook and cranny in the code is filled in. Unlike humans, which are each one individual entity dependent on genetics and proper learning (the latter of which can clearly be eschewed), and can evolve only to a limit. Eventually we'll get it to a state where it will be ready for prime time in 90% of use cases, and will clearly outperform humans in those use cases. On the side, we'll have people developing special-scenario software... like driving in snowy conditions. Then that will follow the same cycle and be ready for prime time. Again unlike one human which is a bundle of randomness and genetics, these computers will have multiple people each working to fill in the little nooks and crannies and gradually make the program better and better. There'll be bugs, but there are bugs in humans (some unfixable), too. Unlike the human bugs, you can actually work out programming bugs without fail.

The way I imagine the final product (ie after mass adoption), at first it seems like they'd follow speed limits and overall go slower... but think about how much safer and smoother the journey will be. If the car itself knew how to get there, it will select the optimal lane to be in at each area. There won't be any crazy three lane merges to get into that one exit lane. And even if there are, the cars will communicate with themselves and give the car that needs to merge ample space to make the transition. Transition done, they go back to normal following distance. IMO it'll be kind of like a conducted orchestra rather than the jumble of nonsense we have out on the roads right now. Your work commute will generally be just one set amount of time. And you can do something else while it's going on. Ideally, I think this is how our future transportation system should be. Yeah, there will be some people that love driving left and don't trust the technology (mainly old timers), but I think with the problems our transportation system currently has... this just needs to happen. The next evolution is third dimension traveling. Currently a massive amount of our transportation takes place in two dimensions. It's pretty much just a map, with X/Y coordinates. We need our future vehicles to have a z axis... and obviously be automated, because at that point everyone needs to be a pilot, and who the hell trusts the average man to pilot?

 

[M76]

I see one danger in the transitional periods. I think most people won'T be able to resist the temptation to use the large following distances kept by self driving cars, to flick flack between the lanes, causing havoc because they only care about themselves. Even I'm forced to keep a very small following distance in traffic jams which is both a waste of fuel and unneccessary, but otherwise idiots would decide to overtake me to be one car further ahead. This already happened to me more than once than some idiot did that. I was slowly crawling with constant speed to not have to stop and go all the time, because that's the biggest wear on the car, but the idiot decided to overtake me because I left 5 car lengths of empty road in front of me.

All I see regarding self driving technology is FUD. As soon as we get past that mass adoption is inevitable.

 

[Biznatch]

I see one danger in the transitional periods. I think most people won'T be able to resist the temptation to use the large following distances kept by self driving cars, to flick flack between the lanes, causing havoc because they only care about themselves. Even I'm forced to keep a very small following distance in traffic jams which is both a waste of fuel and unneccessary, but otherwise idiots would decide to overtake me to be one car further ahead. This already happened to me more than once than some idiot did that. I was slowly crawling with constant speed to not have to stop and go all the time, because that's the biggest wear on the car, but the idiot decided to overtake me because I left 5 car lengths of empty road in front of me.

All I see regarding self driving technology is FUD. As soon as we get past that mass adoption is inevitable.

LOL 5 car lengths... If I leave a gap bigger than 1 car here in LA, some douche nozzle will pinch in every god damn time to get 1/2 car length ahead. Which usually requires me and the 5+ people behind me to slam on their brakes.... People in cities are the biggest assholes and worst drivers. Nothing like almost ramming into some walnut who backed out of their driveway into the street without looking, who then proceeds to start yelling at me to watch where I'm going.... Dwight Schrute was right, we need another plague....

 

[IcePickFreak]

Surely they'll finally get security right and these would never get hacked and cause a 1000 car pile-up. You can hack regular modern cars, and just about everything else made in the last 10-15 years.. but they'll finally get it right this time!

And what about the other massive part of the country that has an actual winter? Snow covered roads, ice covered roads, ice & snow covered sensors? Or is this just going to be for the weirdos in California & Florida that can't manage driving in 70° & sunny weather?

 

[M76]

LOL 5 car lengths... If I leave a gap bigger than 1 car here in LA, some douche nozzle will pinch in every god damn time to get 1/2 car length ahead. Which usually requires me and the 5+ people behind me to slam on their brakes.... People in cities are the biggest assholes and worst drivers. Nothing like almost ramming into some walnut who backed out of their driveway into the street without looking, who then proceeds to start yelling at me to watch where I'm going.... Dwight Schrute was right, we need another plague....

Sure in moving traffic they go for 1 inch of space too. But in stop and go traffic, 5 car lengths is about when I see them start get agitated behind me.

 

[lcpiper]

If anything the self driving car would be better at it.

That's a big claim.

My old man was driving the family home from dinner out. The streets were pretty empty, it was dark, there was one car up ahead as we pulled up behind waiting for the light to turn green.

My dad had noticed something funny about how the car in front was driving earlier. He actually told us that the guy ahead of us was about to try and back into us and sure as shit, the white lights came on and the guy backed up sharply. So did my dad. When the guy realized he hadn't hit us as he expected, he tried a second time, and failed a second time. Finally the light turned green and the guy drove off.

Would your self driving car do it better then that?