Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow • Hacker News Board

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#211

Post by Piskvorrr » Tue, Sep 18, 2018, 10:41 AM UTC

imtringued wrote: post #201

x3n0ph3n3 wrote: post #26

Earlier quoted context omitted.

It has nothing to do with the trolley problem. The trolley problem is a headline-grabbing non-problem in the actual design of driverless cars.

The real trolley problem: [ ] Let self driving cars kill people during development but try to make it up by saving thousands of lives after they are perfected. (Uber) [ ] Make self driving cars extremely safe from the start but more people end up dying from manual driving because development takes longer. (Waymo)

The underlying assumption "general autonomous driving can be made much safer in a short timeframe [years]" is not granted. It may be that the first option could have a death toll higher than what it would prevent - but the thing is, we don't know, it's an unknown, an assumption "it will be much better once everybody drinks the koolaid."

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#212

Post by emodendroket » Tue, Sep 18, 2018, 10:41 AM UTC

LandR wrote: post #182

emodendroket wrote: post #63

Earlier quoted context omitted.

Yeah I don't think most bosses in Boston are going to let you stay home every time it snows.

The last 2 jobs I've had we've always been told pretty explicitly if the roads are bad with snow DO NOT try to get to work. Hell, I've been told I've not even to walk to work when it's snowing. ALthough saying that the last time it snowed here badly, I tried to walk to shop and fell over twice. Maybe my employer just knows me too well...

Yes, but the experience of a software developer is not representative of most people's experiences. A lot of jobs they're going to want you there if the roads are not literally closed, come hell or high water.

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#213

Post by adrianN » Tue, Sep 18, 2018, 10:53 AM UTC

jakobegger wrote: post #181

umichguy wrote: post #174

Earlier quoted context omitted.

"I’m curious how self driving focused groups are solving lane detection and managing surrounding traffic on snow covered roads. It’s not uncommon for a multi lane highway to be completely covered with snow and multiple tracks for different ‘lanes’ that overlap. It’s also not uncommon to see cars just making their own lanes. This occurs even after the snow has stopped flying." Yep. As a Michigander, I totally agree. Y…

Humans are good at improvising. Not using all the lanes when the lane markings are covered by snow, and the road is slippery, sounds like a reasonable strategy. And nobody ever told us that's what we should do. With autonomous vehicles, you need to plan every situation in advance. Our current AI technologies can't improvise. It seems to me that current AI is basically just a decision tree with some neural networks sp…

We don't want machines to improvise because we want to be able to test their whole gamut of behavior as thoroughly as possible. We can build AI that can improvise, but that's not the kind of software you want to be liable for as a manufacturer.

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#214

Post by trashtester » Tue, Sep 18, 2018, 10:56 AM UTC

tiagoafpereira wrote: post #204

I still cannot comprehend the current push for automated driving. We, collectively, not just the US, have a problem with too many cars and instead of reducing said number of cars we want to make them self crashing? Instead of investing heavily on public transit systems we keep feeding this fable that somehow, in the near future, cars driven by computers will do at least as good a job of driving as sentient, sober peo…

Self driving cars, when perfected, will dratically reduce the number of cars.

Once there Uber self driving cars will then be both cheaper, safer and more reliable than your own car. And you don't need to park them. Only those with special requirements (ie professionals that need to store equipement in their car) need to have their own, the rest of us can use a taxi service.

Furthermore, when human driven cars are removed from the roads, road capacity could increase by 50-100% or more, as the distance between cars (in both dimensions) can be safely reduced.

Also, you will not need to pick up your children/parents/others that cannot drive. For instance, if you drive your kids to school, you likely have to drive both ways, doubling traffic (and wasting your time).

For commuters, self driving cars also lend themselves were well for ride sharing, having cars that function like full office spaces or entertainment sources, meditation rooms or whatever you can think of. If all cars are self driving, the speeds will could be regular, and uncomfortable acceleration/deceleration avoided.

Finally, the tech going into this is likely to be generalizable to other kinds of machinery, such as mining machinery, construction machinery, automated chefs, as well as the functions in manufacturing that have not already been automated.

I think the potential of the end state should be quite clear. The total economic potential is huge. Anyone able to attain near monopoly for one of the technologies going into this (software, sensors, traffic control centers, etc) stand to be the next Google.

That said, I think it is unlikely that it will take off in full until between 2025 and 2035. Now, we are in a situation similar to the .com era around 1998. If history repeats itself, 2023 may be a great year to invest in robotics startups, provided we see a bubble bursting.

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#215

Post by Tor3 » Tue, Sep 18, 2018, 10:56 AM UTC

maxerickson wrote: post #207

Moru wrote: post #198

Earlier quoted context omitted.

If you read the manual for the traction control it is full of big warnings on every page,'Does not cancel out physical laws'. It is not a solved problem, it can only do as best it can uphill until car does not have enough traction to continue upwards. At which point the car will slowly glide down or into the ditch.

So what's your point? Are you saying human drivers can safely navigate those same hills or something?

What he's saying is that a human will, in those conditions, build up a speed in advance that is exactly enough to get up there before the speed has dropped too much (due to having to back off on the speed to keep traction. The computer can do the latter, but can it do the former? Probably when we get AI. But we don't have AI. Just machine learning.)

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#216

Post by krapp » Tue, Sep 18, 2018, 11:03 AM UTC

tiagoafpereira wrote: post #209

krapp wrote: post #206

Earlier quoted context omitted.

> I still cannot comprehend the current push for automated driving. Google and Uber believe they can make an insane amount of money with the transition to automated driving. This is being sold to the public as the solution to a public safety crisis with the narrative that human controlled cars are death machines which need to be taken off the roads as soon as possible.

I wonder how long will it take them to realize that it won't happen before someone cracks the artificial general intelligence problem... I don't know what ticks me off the most, people believing that these companies can easily solve the automated driving problem or these companies stubbornly pushing for it.

I'm sure plenty of engineers actually working on the problem are aware of how difficult it is... but they're not the ones doing the pushing or making the grandiose claims.

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#217

Post by mikejb » Tue, Sep 18, 2018, 11:06 AM UTC

jakeogh wrote: post #173

mikejb wrote: post #94

Earlier quoted context omitted.

I'd argue that Self-driving cars are better capable of recognizing ice patches. They don't just rely on visible light to see the ice, but have a more diverse sensor array. The challenge there of course is having the car learn to interpret the input of these sensors. But that doesn't seem like a unpassable hurdle. On the contrary, it would be pretty high on the list of things to tackle when aiming for driving in snowy…

Ignoring the real problem you mentioned, camera dynamic range (for example) is pathetic compared to the human eye.

Depends on the camera, but mostly yes. OTOH, it's (relative) trivial to improve that for cars; it's way harder to improve it for humans.

(Also, as a data point, I think there was a guy who drove his Tesla with lights off, and it was able to stay within the lines)

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#218

Post by trashtester » Tue, Sep 18, 2018, 11:06 AM UTC

purplezooey wrote: post #152

Well boys, looks like it'll be another couple of decades

Or it could start taking off 2-3 years after the bubble bursts. Amazon lost 90% of its value between 2000 and 2002, and many of the big players today entered the market between 2002 and 2005, when confidence in the "new economy" was at a minimum.

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#219

Post by mannykannot » Tue, Sep 18, 2018, 11:14 AM UTC

adrianN wrote: post #213

jakobegger wrote: post #181

Earlier quoted context omitted.

Humans are good at improvising. Not using all the lanes when the lane markings are covered by snow, and the road is slippery, sounds like a reasonable strategy. And nobody ever told us that's what we should do. With autonomous vehicles, you need to plan every situation in advance. Our current AI technologies can't improvise. It seems to me that current AI is basically just a decision tree with some neural networks sp…

We don't want machines to improvise because we want to be able to test their whole gamut of behavior as thoroughly as possible. We can build AI that can improvise, but that's not the kind of software you want to be liable for as a manufacturer.

I am not at all convinced that improvisation is an easy add-on. While AI has demonstrated some creative ways to satisfy goals that have left options open (sometimes inadvertently), real-world improvisation has to be appropriate for the situation and the unusual difficulties it poses, and, in the case of driving, it must not create additional problems or (of particular relevance here) present significant dangers.

In general, I think, improvisation requires causal reasoning. It also often needs a broader knowledge than that needed for the nominal task (for example, when driving on a potholed road, one needs to have some understanding of how it affects the car's response, and sometimes of issues such as ground clearance.) Constraining the vehicle to avoid such issues would constrain the option to improvise.

Re: Self-Driving Cars Can Handle Neither Rain nor Sleet nor Snow

#220

Post by apcragg » Tue, Sep 18, 2018, 11:15 AM UTC

Animats wrote: post #156

The article is kind of vague. What they seem to be doing is operating a ground penetrating radar under the vehicle so that underground details can be used as positional references. This is useful only if you have a map which includes that data. At best, this tells you where you are. It tells you nothing about what's ahead. LIDAR units that return "first and last" returns are helpful in dealing with rain and fog. The…

Small nitpick. The range resolution improves with increasing pulse/chirp/waveform bandwidth, not increasing carrier frequency. It's true that at higher carrier frequencies there is more bandwidth available and thus better range resolution is possible but the relationship is slightly different than "delta_r ~ frequency"