Google Self-Driving Car Hits Municipal Bus

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
Hmmm, a self-driving car swerves to miss some sand bags, ends up hitting a bus instead. The explanation claiming "the vehicle and the test driver believed the bus would slow or allow the Google (autonomous vehicle) to continue" is pretty sketchy. Just wait until it swerves to miss some sand bags and you go off a cliff. :eek:


Alphabet Inc's Google self-driving car struck a municipal bus in a minor crash earlier this month, a recent filing showed, in what may be the first case of one of its autonomous cars hitting another vehicle. In a Feb. 23 report filed with California regulators, Google said the crash took place in Mountain View, California on Feb. 14 when a self-driving Lexus RX450h sought to get around some sandbags in a wide lane.
 
This accident is actually a good thing. The engineers need more data to come up with optimal responses. Most likely, if the google car went more than "2 mph" to go around the sandbags, they would have completely missed the bus.
 
My problem, is that the car moved into the side of the bus. That is 100% the car's fault, you did not properly navigate a merge. The information is clearly there, the car was traveling at 2 mph and the bus was going about 15. While the speeds don't seem that big of a deal, but when you are doing 60 on a highway and someone is doing over 70 next to you, they go by pretty fast.

Their software did a poor job of realizing they either need to speed up or slow down/stop to make the merge work.

I honestly can't wait till all driving has to be automated like iRobot or other various shows that have shown things like that. Then we can bitch at the road engineers that have created terrible issues with our highways, roads, and light setups. Because there will be traffic jams even with traffic computers controlling traffic in an orderly fashion. ;)
 
This accident is actually a good thing. The engineers need more data to come up with optimal responses. Most likely, if the google car went more than "2 mph" to go around the sandbags, they would have completely missed the bus.
Exactly. The bus probably didn't slow down because people expect other drivers to speed up when merging. It's the definition of merging, for crying out loud.

"Not once was the self-driving car the cause of the accident," Google said at the time.

That's another way of saying our autonomous vehicles are not yet ready to drive in everyday traffic.
 
So what did the test driver do? Fall out of the car and start rolling around on the ground exclaiming, "My neck. My neck."? :D
 
Exactly. The bus probably didn't slow down because people expect other drivers to speed up when merging.
Not to mention, humans have a sense of self preservation and a bus is the last thing someone will try to muscle in front of. The bus driver probably saw the slow as hell google car and figured they would yield to the much bigger vehicle that is actually moving at more than a glacial pace.
 
From now on, our cars will more deeply understand that buses (and other large vehicles) are less likely to yield to us than other types of vehicles
Gee big shocker on this, bus drivers tend to do this. Add to this cab drivers, humans with cell phones attach to the side of their face (regardless of laws), humans who breath in nitrogen and oxygen regularly, and pretty much every car.
 
Google Engineers said:
"From now on, our cars will more deeply understand that buses (and other large vehicles) are less likely to yield to us than other types of vehicles, and we hope to handle situations like this more gracefully in the future."
Huge ego boost for myself, I am blown away this was not an early, and integral portion of the autonomous design.

These engineers may be more the stuff of WORK SUPER HARD, and not the WORK SUPER SMART variety (which I normally associate with Google).

The absolute basis of any autonomous structure (for driving) should be a safety core that can calculate the worst case decisions of other drivers, and basis it's "thinking" from that platform.

It sounds really hard, but after some thought play I've designed some rudimentary approaches that would work beyond a 2d road environment.
 
It was probably stuck behind one of those aluminum cages on wheels while savage children were taunting/screaming/throwing things at the car.
 
Highlighting why I as a SW Engineer don't think self driving vehicles are going anywhere: Too many cases like this that will eventually cost everyone involved a great deal of money.
 
Highlighting why I as a SW Engineer don't think self driving vehicles are going anywhere: Too many cases like this that will eventually cost everyone involved a great deal of money.

I think it's an all or nothing situation. There also needs to municipal and regional infrastructure to support self-driving vehicles.

On top of that: In my experience transit bus drivers do not slow down, or make room for anyone. Ever.
 
I think it's an all or nothing situation. There also needs to municipal and regional infrastructure to support self-driving vehicles.

On top of that: In my experience transit bus drivers do not slow down, or make room for anyone. Ever.
In Florida, I know the law is you have to yield to a bus merging from a stop. It's probably the same or similar in most other states. Unfortunately around here this has turned bus drivers apathetic to the traffic conditions around them. Let's put it this way: if I'm on a three lane road I will be sure to be in the far left lane when approaching a bus whether or not they're making a stop.
 
Even relatively stupid people are going to be better drivers than self-driving cars. Those who think otherwise are grossly under-estimating the complexity of driving.
 
It was going 2mph and STILL couldn't figure it out? Good thing it wasn't going faster. I figured something as basic as a non-moving sandbag wouldn't be such an issue to endanger a bus full of people.

So, has Google released the video of the wreck? Anyone have a link? How about the other wrecks? I want to see these for myself. Google seems to be pretty forthcoming with all of their street view pictures and other data... Not these autonomous vehicles though, eh? I have yet to see any.
 
"Nearly 1.3 million people die in road crashes each year, on average 3,287 deaths a day. An additional 20-50 million are injured or disabled." - asirt.org

If driver-less cars are even just 1% safer than humans, we all win. My guess is that the are 99% safer than humans. I bet you can't find even 1 study that shows that driver-less cars are less safe than humans. Sure driver-less cars may not be perfect (yet), but they are still much safer than humans. Also you can't prevent all accidents. For example, someone high on drugs deciding to bolt across the freeway on foot isn't likely to end well as cars, even if they can react instantly, still take time to stop when they are going 60+ on the highway.
 
I wonder if the speed of the bus, the proximity of the bus, and the size of the bus, along with a potentially stationary status when first identified; led the googlemobile to identify it as a wall or building, or at least a parked vehicle.
 
Takes more than believing you have the right of way to avoid an accident, whether you actually have it or not.
 
Only idiots assume buses or trailers will yield that's how to enter a blind spot and get side swiped.
 
Their software did a poor job of realizing they either need to speed up or slow down/stop to make the merge work.

Apparently the driver in the google car did a poor job as well, estimated likewise and lost the gamble to the aggressive bus driver. So much for humans.
 
Enough with the freaking self-driving car!!! This is not going to work anytime in the near future.

"There has been no official determination of fault in the crash. Google has previously said that its autonomous vehicles have never been at fault in any crashes."

So there has been more than one crash caused by this thing?

Exactly. The bus probably didn't slow down because people expect other drivers to speed up when merging. It's the definition of merging, for crying out loud.
My driver's instructor told me you are supposed to maintain the same speed as traffic when merging/changing lanes. Most people speed up. Some freak out and slow down. This is the kind of thing you can't program...

"Nearly 1.3 million people die in road crashes each year, on average 3,287 deaths a day. An additional 20-50 million are injured or disabled." - asirt.org

If driver-less cars are even just 1% safer than humans, we all win. My guess is that the are 99% safer than humans. I bet you can't find even 1 study that shows that driver-less cars are less safe than humans. Sure driver-less cars may not be perfect (yet), but they are still much safer than humans. Also you can't prevent all accidents. For example, someone high on drugs deciding to bolt across the freeway on foot isn't likely to end well as cars, even if they can react instantly, still take time to stop when they are going 60+ on the highway.
They probably are safer than humans... most of the time. The problem is when a sensor fails on the freeway and they kill a bunch of people. The redundant backups you would need to add would make them uneconomical. Never going to work!
 
Last edited:
It's really impossible to tell what happened here without a video. End of story.

Agreed.

That being said, if we compare the number of at-fault accidents per million miles driven, my money is on that the Google autonomous car is well below the average human driver rate :p
 
Some of the responses in this thread are mind blowing. Our ancestors never would have made it to fire with the way some of you dildos think.
 
Some of the responses in this thread are mind blowing. Our ancestors never would have made it to fire with the way some of you dildos think.
There are the ancestors who invented or adopted fire. And then there were those ones who thought fire would make a nice fashion accessory on their clothes. Would like to think the latter all died out, but maybe not.
 
Last edited:
Agreed.

That being said, if we compare the number of at-fault accidents per million miles driven, my money is on that the Google autonomous car is well below the average human driver rate :p
The Google Car would be much higher, if it didn't have human drivers saving it.

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

The car's human drivers prevented at least 10 at fault accidents in 420,000 miles. The estimated accident rate in the US is 4.2/million miles (including unreported ones).

National crash rate for conventional vehicles higher than crash rate of self-driving cars, report shows | Virginia Tech News | Virginia Tech

And that's with the Google Car doing virtually all of its driving in California and in good weather.
 
This is stupid. I bet you the software is being written by ppl who actually don't do a whole lot of driving. How about have their software programmers go drive normal cars in traffic for a year. At least 40 hours per week, so they can see how traffic actually is. Then start programming.

Not only that, but they'd have to do it multiple different countries, cause drivers are different in other places. Germans don't drive like Americans. Brits don't drive like Americans. Japanese don't drive like Americans. Koreans don't drive like Americans. French don't drive like Americans.

Do I see self driving cars happening? Yes, I really do. About 20+ years from now.
 
Not to mention, humans have a sense of self preservation and a bus is the last thing someone will try to muscle in front of. The bus driver probably saw the slow as hell google car and figured they would yield to the much bigger vehicle that is actually moving at more than a glacial pace.

My own driving experience shows this to be incorrect. Also, my step-dad drove 85,000 lb trucks for 40 years, and 4-wheelers were the bane of his existence because of their persistent desire to be run over by him.
 
Not to mention, humans have a sense of self preservation and a bus is the last thing someone will try to muscle in front of. The bus driver probably saw the slow as hell google car and figured they would yield to the much bigger vehicle that is actually moving at more than a glacial pace.

LOL. I've seen someone in a sedan cut-off a loaded 53' tractor-trailer off ramping from the interstate going 30MPH over the speed limit, about to have their car split in half by an Interstate median/wall.

If you think a bus is the "last thing someone will try to muscle in front of"....drive on I-35 in Oklahoma or Tx some time.

This is stupid. I bet you the software is being written by ppl who actually don't do a whole lot of driving. How about have their software programmers go drive normal cars in traffic for a year. At least 40 hours per week, so they can see how traffic actually is. Then start programming.

Not only that, but they'd have to do it multiple different countries, cause drivers are different in other places. Germans don't drive like Americans. Brits don't drive like Americans. Japanese don't drive like Americans. Koreans don't drive like Americans. French don't drive like Americans.

Do I see self driving cars happening? Yes, I really do. About 20+ years from now.

Because the Department of Transportation cares rightfully about how the vehicles can follow the law as it is written. Not getting robots to mimic how f'ing maniacs actually are out to try and kill people by being reckless.
 
Because the Department of Transportation cares rightfully about how the vehicles can follow the law as it is written. Not getting robots to mimic how f'ing maniacs actually are out to try and kill people by being reckless.

Ya, cause driving into the middle of the bus is following the law as it is written
 
Ya, cause driving into the middle of the bus is following the law as it is written

In the first case, self-driving cars don't need to be perfect...they just need to do a better job than humans...which for a computer-system is a stupid-simple not-even-a-challenge. This is one not-even-fatal incident this year. There are over 30,000 driving fatalities every year, in the USA alone.

In the 2nd case, spend 5 seconds.

Seriously. Spend 5 seconds. Think about WHO the customer for the self-driving car will be? It will NOT be a consumer vehicle for a (very) long while. Remember how badly the Volt sold, because of how expensive it was? How many Teslas do you see on the road? Tech like this is always priced beyond 99% of consumers reach when debuted. Hell for as popular as people think the Prius is, it has a staggeringly low market penetration (something like 1-2% of consumer vehicles on road in the USA last I knew)

Did you spend 5 seconds thinking about who the customer will be for the self-driving car? If consumers inevitably cannot afford it how will car companies make money off of it?



EASY.

ANYONE who employs delivery drivers. FedEx/UPS/USPS/PizzaHut, you name it. Who else? Semi-truck drivers. Bus drivers. Of which between the three there are over 3,500,000 people trying to earn a living....and all will be unemployed. Tell me. Do you REALLY want a 53' semi truck or a school/city-bus or a UPS truck driven in the same manner a soccer-mom drives? I mean really? Think this through just a tiny bit. You and everyone should want these vehicles FOLLOWING THE LAW like the Bible, because you will not be able to buy one for a very long time...instead all the delivery/courier vehicles on the road will not have a human driving FAR sooner.
 
Even relatively stupid people are going to be better drivers than self-driving cars. Those who think otherwise are grossly under-estimating the complexity of driving.
i disagree, most drivers are idiots. google has millions of miles racked up already with zero crashes except one.
 
The Google Car would be much higher, if it didn't have human drivers saving it.

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

The car's human drivers prevented at least 10 at fault accidents in 420,000 miles. The estimated accident rate in the US is 4.2/million miles (including unreported ones).

National crash rate for conventional vehicles higher than crash rate of self-driving cars, report shows | Virginia Tech News | Virginia Tech

And that's with the Google Car doing virtually all of its driving in California and in good weather.

Those numbers just don't seem right. I wonder how they came up with them.

I've never been in an at fault accident, but I keep being hit by other people who don't seem to know what they are doing.

I'd estimate my "hit by idiots" rate as once per ~50,000 miles, nowhere near the once per 4.2 million...

I know my own experiences are just anecdotal in the grand scheme of things, but being ~84 times higher than those figures suggests to me that something is off...
 
In the first case, self-driving cars don't need to be perfect...they just need to do a better job than humans...which for a computer-system is a stupid-simple not-even-a-challenge. This is one non-fatal incident this year. There are over 40,000 driving fatalities every year.

Uh, do you even know how super complex it would be to create a computer system to actually drive in every single weather condition, every road condition, every anything that happens on the road. It's hardly a "stupid-simple not-even-a-challenge" type of thing. If I hang a bed sheet in front of the car, can it tell it's a bed sheet or will it think it's a concrete wall? How does it detect there's ice on the road? The Google car can't even drive in rain yet. The google car can't even detect temporary speed limit changes, but hey. Must be stupid simple. Also of course they have non-fatal incidents. The google car can only travel 25 mph.

Google has reported tons of cases where the car failed and the human had to take over. Where the car couldn't see traffic lights, yield to pedestrians, or commited traffic violations. Not to mention dealing with other drivers or the car just making unwanted maneuvers. That's on perfect days. We haven't even seen how it'll be with their new testing in rain.

In the 2nd case, spend 5 seconds.

Seriously. Spend 5 seconds. Think about WHO the customer for the self-driving car will be? It will NOT be a consumer vehicle for a (very) long while. Remember how badly the Volt sold, because of how expensive it was? How many Teslas do you see on the road? Tech like this is always priced beyond 99% of consumers reach when debuted.

Did you spend 5 seconds thinking about who the customer will be for the self-driving car? If consumers inevitably cannot afford it how will car companies make money off of it?



EASY.

ANYONE who employs delivery drivers. FedEx/UPS/USPS/PizzaHut, you name it. Who else? Semi-truck drivers. Bus drivers. Of which between the three there are over 3,500,000 people trying to earn a living....and all will be unemployed. Tell me. Do you REALLY want a 53' semi truck or a school/city-bus or a UPS truck driven in the same manner a soccer-mom drives? I mean really? Think this through just a tiny bit. You and everyone should want these vehicles FOLLOWING THE LAW like the Bible, because you will not be able to buy one for a very long time...instead all the delivery/courier vehcles on the road will not have a human driving FAR sooner.

How about you spend 5 seconds. I mean, seriously. Spend 5 seconds. Now tell me, who takes that package, pizza, truck load, etc and give it to the customer? Who verifies that the package gets signed for? Who verifies they receive payment for the pizza delivery? Who verifies the load is dropped off and the correct load is picked up? Who deals with getting the semi-truck weighed? Who verifies the bus gets cleaned after someone throws up in it? Also following the law like the bible? What law? Every state has their own laws. Every country has their own laws. Some countries have no laws.


Either way, I never said autonomous vehicles aren't coming. They definitely are, but I don't see it happening quickly. Probably within the next 10 years, we'll have some kind of auto-pilot for highway driving. 20 years, everywhere else.
 
If you're gonna have your first at-fault accident, go big and hit a bus with 15 passengers! This won't be a problem as soon as buses are self-driving also ...
 
i disagree, most drivers are idiots. google has millions of miles racked up already with zero crashes except one.

How about reporting how many crashes were avoided due to the human driver intervention?

Every 1200 miles, a human had to take control of the google car for late 2014 to late 2015 testing. How many crashes would have happened if a human didn't? It's gotten better, where a human only has to take over every 5000 miles, but that's also in good weather. I wonder how much it'll drop again with these new batch of tests they'll be doing in rain.
 
If you're gonna have your first at-fault accident, go big and hit a bus with 15 passengers! This won't be a problem as soon as buses are self-driving also ...

Only if we go to one system. Everyone seems to be producing their own systems. I really hope they decide to standardize a lot of these systems, so the different systems actually talk with each other. If not, we'll still have all these issues even if the bus was self-driving too.
 
How about reporting how many crashes were avoided due to the human driver intervention?

Every 1200 miles, a human had to take control of the google car for late 2014 to late 2015 testing. How many crashes would have happened if a human didn't? It's gotten better, where a human only has to take over every 5000 miles, but that's also in good weather. I wonder how much it'll drop again with these new batch of tests they'll be doing in rain.

this more has to do with humans not following the rules of the road, not the google system fucking up. if anything the google system is too vanilla which is getting better all the time.
 
Back
Top