Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The latest chapter for the self-driving car: mastering city street driving

googleblog.blogspot.com

Post reply on HN
Print view View original

171–180 of 455 posts

Re: The latest chapter for the self-driving car: mastering city street driving

#171

Post by Qworg »

gfodor wrote: post #161

Question for self-driving car aficionados. When I'm driving and encounter a strange or dangerous situation these days I often try to think if a robocar will be able to handle it properly. One situation came up the other day that made me nervous. I was driving on a relatively empty highway at high speed (70mph) and there was a piece of debris in one of the lanes. I spotted the debris far ahead and safely changed lanes…

Debris is a really hard problem and I don't think they can get that resolution out of their LIDAR. Camera data may serve better, but that's hit or miss.

I'd call this one unsolved.

(Source: I'm a robotics/perception/sensors guy who has worked autonomous vehicles)

Re: The latest chapter for the self-driving car: mastering city street driving

#172

Post by mindslight »

mc32 wrote: post #163
mindslight wrote: post #124

Earlier quoted context omitted.

Don't get your hopes up, that brainwashing is still alive and well with the first word of Google's post being "Jaywalking".

Underscoring your sentiment is the origin of the term jaywalking http://paleofuture.gizmodo.com/the-invention-of-jaywalking-w... Although, to be sure, with the speeds of modern vehicles, it makes sense that people's paths be predictable since vehicles are much less able to be nimble and swerve out of someone's path.

Predictability and not jumping in front of cars is unrelated to not being in a crosswalk. Crosswalks actually worsen these aspects, as they make drivers more prone to needlessly slamming on their brakes to wave people out while also fostering a sense of pedestrian entitlement instead of respectful temporal negotiation.

It's really not hard to look at traffic (all types), spot the existing gaps, and navigate within them. On some roads this is impractical due to traffic levels, but in most areas these type of roads are the exception.

(I suppose it's not surprising to be downvoted for saying this, as California has some of the worst drivers I've seen. It's rare that I am able to start walking and successfully pass behind a moving car, without having the driver needlessly stop while looking annoyed and acting like they're doing me a favor - even in parking lots! But frankly, anyone thinking of themselves as a hacker should surely understand and respect the value of down-to-earth informal negotiation and the peril of overformalized systems).

Re: The latest chapter for the self-driving car: mastering city street driving

#173

Post by jlebar »

oceanplexian wrote: post #150
Zigurd wrote: post #142

Earlier quoted context omitted.

If the biking routes are that popular, one side of the street should be a bike lane separated from traffic. Until such time, cars need to be patient. Each one of those cyclists you are stuck behind is keeping a 3 ton multi-kilowatt biggest contributor to greenhouse gas emissions car off the road. Cars are huge polluters and energy consumers. Would you be equally pissed off if the residents along narrow streets had sp…

Cyclists are using roads that are subsidized and paid for by gasoline taxes and automobile registration fees. If you really want dedicated lanes and preeminence then pay your fair share. I don't think bicycling would be as attractive if cyclists paid for all the externalities of their hobby.

I'm not aware of any country where gas taxes cover the externalities of the emission of that carbon.

I'm also not aware of any country where gas taxes and car registration fees come close to paying for the total cost of roads.

Re: The latest chapter for the self-driving car: mastering city street driving

#174

Post by eigenvector »

rpedela wrote: post #170
TeMPOraL wrote: post #167

Earlier quoted context omitted.

My thought: LIDARs and other kind of sensors are much better than human eyes. I'd expect the current self-driving cars to have already better sensory resolution than human drivers.

Yeah and I think a larger problem is deer or other animals running into the road especially at night. I don't know if the Google car can handle that situation yet, but the sensors should be able to see animals way before humans can.

Many (most?) car/deer/moose collisions are unavoidable even for a human driver. In particular, because a human driver focuses on monitoring the road ahead, they cannot determine instantaneously whether it is optimal to continue straight and hit the animal or to swerve into oncoming traffic or the adjacent lane. The reflex reaction is to swerve which can often cause an even worse accident.

Humans also can't process the trade-off of hitting the animal vs. attempting to avoid it based on size fast enough whereas a self-driving car could conceivably be able to.

Re: The latest chapter for the self-driving car: mastering city street driving

#175

Post by bertil »

0x006A wrote: post #121
bertil wrote: post #49

Earlier quoted context omitted.

Indeed… I just hope that problem is the same as when someone asked them, 15 years ago, how their crawler could handle websites with “tens of thousands of pages” fearing libraries would put their ressources on-line, overburden and halt their effort.

In my experience driving is much more interactive and not so much rule based in those cities. You have to judge other drivers - there aggression and willingness to drive into your lane, there willingness to let you sneak in in front of them. Taking traffic from all sides into account might be something that works better for self-driving cars. Knowing when its ok to force someone else to slow down might be harder. I d…

Indeed. I actually believe — or rather, sense, because that tends to be a fairly instinctive appreciation for me, after modeling things for so 15 years — that aggressivity is a fairly straightforward parameter in driving style: basically, it’s the margin you allow yourself to touch, or the acceptable risk you‘d take. Choosing “the country”, on a vehicle with more GPS captors that anyone could remember is probably not something difficult, but there are major and new ethical issues around that point, precisely: what do we let the user control?

Can, should, will Google include a ‘Don’t abide to local driving laws’ mode? How explicit should it be? Modern ethics tend to surround several though exercices, most of which are centered around transportation accidents, namely one involving throwing, or letting fall, or pushing a button to release a lever… to have one obese man fall and prevent a car or a train from hurting three other people; those may or may not be relatives, have been careful in their behaviour, etc. What used to be theoretical questions, used to classify philosophers in their principles might become soon electronic options. Should your car, detecting a swerving vehicle is about to hit a child running after his ball, run that car into the nearby wall (after checking in 10 ms that it has working airbags) and stop the accident? Whose insurance pays the damages?

Re: The latest chapter for the self-driving car: mastering city street driving

#176

Post by TeMPOraL »

Shivetya wrote: post #148
anjc wrote: post #147

Are their cameras visible spectrum ones? How will it work at night? Night vision? IR?

Night, hell tell me how it works in snow, with fully covered roads. I would love to see how they make that leap that people do when looking at situations where the road is fully obscured, but you know exactly where its at based on objects your used too (like mailboxes/trees/etc) GPS ain't that accurate.

Computers can remember more objects better, recall them faster and calculate your position against them with much more detail, as well as apply inference rules to correct for any missing/moved objects much better than humans can.

Re: The latest chapter for the self-driving car: mastering city street driving

#177

Post by rictic »

oceanplexian wrote: post #143
TheCoelacanth wrote: post #81

Earlier quoted context omitted.

Human drivers can't predict that either, and the Google car should have a much faster reaction time than a human driver.

Reaction time isn't the problem. Cars can't make moral decisions. It's not uncommon that a pedestrian will jump in the road (sometimes intentionally to commit suicide). This happens to me personally on a daily basis in San Francisco. Sometimes the driver has a moral choice to ether run the car off the road killing people on the sidewalk or killing the pedestrian. What do you expect a Self-Driving car to do in that si…

"Cars can't make moral decisions."

There are moral problems that I'd be uncomfortable putting to a computer.

This isn't one of them. Am I missing something, is there something wrong with simply modeling likely impacts and trying to cause the least harm to the fewest number of people? That's what I would do, only with worse reaction time and less information.

Re: The latest chapter for the self-driving car: mastering city street driving

#178

Post by bertil »

gfodor wrote: post #161

Question for self-driving car aficionados. When I'm driving and encounter a strange or dangerous situation these days I often try to think if a robocar will be able to handle it properly. One situation came up the other day that made me nervous. I was driving on a relatively empty highway at high speed (70mph) and there was a piece of debris in one of the lanes. I spotted the debris far ahead and safely changed lanes…

Most human drivers tend to over-estimate the consequence of a collision — it’s common that swerving to avoid a wild animal poses a greater danger than hitting, for instance.

However, what a robocar can do that most humans can’t is organise the clean-up, and that’s hundred times more important, because presently, hundreds of cars will pass that debris.

Let’s say you have a radar under the chassis (not unlikely given Telsa’s latest PR issue): the car OS can warn the authorities instantly with an exact location; images from several cars can be compiled to form an exact estimate of what it is. All car with significant connectivity can be informed that this lane is closed a small portion before any physical equivalent (orange cones) has time to be set up, possibly allowing a safer clean-up if the (connected) highway detects that all cars in the next x minutes have been warned, and a brusher drone can be dispatched.

Re: The latest chapter for the self-driving car: mastering city street driving

#179

Post by theg2 »

josefresco wrote: post #106
Zigurd wrote: post #23

Earlier quoted context omitted.

Even before we get self-driving cars, autonomous braking should be mandated so that cars and trucks are literally incapable of coming too close to a cyclist.

I'll back that as soon as a system to force bicyclists to abide by traffic laws is required.

As someone who was nearly run down by a cyclist running a red while I walked a cross walk this weekend, I agree wholeheartedly. Boston is in a constant battle between cyclists and motorists and neither side is playing by the rules.

Re: The latest chapter for the self-driving car: mastering city street driving

#180

Post by mikeash »

deanclatworthy wrote: post #4

This is amazing, but if anything it's made me more wary of some of the challenges the project faces. - What if the cyclist fell off his bike in front of the car? How quickly can the computer process the real-time imagery and react compared to a human with their peripheral vision. - What if the cyclist swung from the pavement on to the road. A human driver will probably have spotted the hazard (we all train for that k…

I don't understand why people seem to always imagine that human drivers are all Michael Schumacher or Mario Andretti.

Most drivers are terrible at driving. They're unobservant, they have slow reaction times, they have inappropriate reactions, they don't adjust to conditions, they don't understand the laws, they can't anticipate other drivers, they can't plan ahead, you're lucky if they're actually looking where they're going half the time.

What if the cyclist fell off his bike in front of the car? I'd be surprised if a human was able to react to this in less than two seconds. The cyclist probably dies. It wouldn't take much for a computer-driven car to do better than this.

What if a fire hydrant bursts and makes the road really wet? I bet the computer would do something besides continue driving at full speed the way 90% of human drivers would.

These standards are just ridiculously high. Everybody is imagining scenarios that 90% of the time would not be handled correctly by human drivers, and then saying that anything less than a 100% success rate on the part of the computer is insufficient.

Post reply on HN