|
I have to admit that I really don't like the idea of self-driving cars: I like driving and being in complete control behind the wheel. I know a lot of people are excited about them, though, and the argument in support of them is a good one, especially if it prevents more fatalities on the road (I'd rather we just actually teach people to drive, but there will always be room for human error). That said, I feel like a gun-nut, worried that some day the gubmint will take my steering wheels (hopefully not in my lifetime). So Google recently showed off this thing:  http://www.theguardian.com/technology/2014/may/28/google-self-driving-car-how-does-it-work and the experience sounds like a fate worse than death to me, but I'm guessing people have some good insight about it and what may follow, so here's a thread.
|
#
?
May 28, 2014 17:15
|
|
|
|
| # ? May 5, 2024 03:19 |
|
|
I hope I'm dead before self driving vehicles become something that is mandatory. That said, I think the technology behind them is really neat and they could have some real benefits in/around cities.
|
|
#
?
May 28, 2014 17:18
|
|
|
The idea of them as an automated form of taxi sounds fine to me. Having aggressive taxi drivers replaced with docile google cars can only be a good thing for everyone except the taxi drivers. I'd say they're the only ones who have much to fear in our lifetimes.
|
|
#
?
May 28, 2014 17:39
|
|
|
Driving for enjoyment and driving because you have to rarely overlap. Who here actually wants to drive in gridlocked rush hour traffic? Or drive 8 hours of straight flat highway? I absolutely love driving on the track or late at night on deserted roads but I loving hate being forced to deal with traffic and lovely drivers. Maybe automobile driving will join horseback riding and sailing and become something that's mostly done for pleasure. How possible would it be to have autonomous buses and OTR trucks? Bulk Vanderhuge fucked around with this message at 18:18 on May 28, 2014 |
|
#
?
May 28, 2014 18:15
|
|
|
The way I see this transition taking place is that certain roads will be transformed into "auto-highways" in which you'll only be allowed on them if you have a self-driving car in self-driving mode. I think there's a decent chance we'll see conversion kits for existing cars. As the technology deploys over time, you'll see a higher percentage of self-driving roads until it becomes practical to use something like Google's car that is self-driving only. I think we'll start seeing this transition within the next five years.
|
|
#
?
May 28, 2014 21:01
|
|
|
revmoo posted:The way I see this transition taking place is that certain roads will be transformed into "auto-highways" in which you'll only be allowed on them if you have a self-driving car in self-driving mode. I think there's a decent chance we'll see conversion kits for existing cars. As the technology deploys over time, you'll see a higher percentage of self-driving roads until it becomes practical to use something like Google's car that is self-driving only. I think five years is insanely optimistic to even see a single functional mass-market self-driving car on the road, and that they'll by and large be a gimmick that never pans out. The maintenance costs alone will be staggering, and given that big chunks of the populace just plain cannot buy anything resembling a factory new car these days, any such transition will take at least a decade and necessarily be coupled with an epic public transportation revamp. Or to put it in simpler words, that poo poo ain't happening. That said, this is one of the times I'm glad to live out in flyover country, where having something that -can- drive offroad and go places that aren't freeways is pretty much necessary.
|
|
#
?
May 28, 2014 21:40
|
|
|
I wonder how well the self driving cars do in the rain and snow where the droplets will be blocking some of the laser scanners.
|
|
#
?
May 28, 2014 21:46
|
|
|
Radar is largely unaffected by rain, as far as I know, and any good solution using current technologies would probably use both. Or maybe there are lasers on funky wavelengths that aren't affected by such things anyway?
|
|
#
?
May 28, 2014 21:56
|
|
|
Self driving cars are a fantasy for every day conditions still, no matter what Google says. And when / if they arent a nerd dream? You'll take control of a car out of my cold dead hand. I'm sure as gently caress not getting in one as a passenger either.
|
|
#
?
May 28, 2014 22:25
|
|
|
revmoo posted:The way I see this transition taking place is that certain roads will be transformed into "auto-highways" in which you'll only be allowed on them if you have a self-driving car in self-driving mode. I think there's a decent chance we'll see conversion kits for existing cars. As the technology deploys over time, you'll see a higher percentage of self-driving roads until it becomes practical to use something like Google's car that is self-driving only. Five years is insanely optimistic. It took Google five years of lobbying to get a couple states to do an evaluation study to see if autonomous cars might one day be allowed on the road for consumer use. I personally know folks who have left the field because of the political hurdles (both from government and manufacturers). I think five years also ignores the reality that Google's millions of hours driving without incident or whatever is all in ideal scenarios. They don't take the vehicles into "difficult" terrain (hills, tight bends, etc) nor into "bad" weather (rain, night). You want to see what happens when you throw a moderate curveball at an autonomus vehicle? Take a look at what happened in the first DARPA Grand Challenge around mile marker 8, which was a tight, but in no way blind, curve. Krakkles posted:Radar is largely unaffected by rain, as far as I know, and any good solution using current technologies would probably use both. The kind of radar you can strap onto a vehicle is limited to about 300 feet and provides very little information at that range. Right now, LIDAR does the heavy lifting, while RADAR is used for proximity things, much like commercial blind spot detection.
|
|
#
?
May 28, 2014 23:59
|
|
|
Brigdh posted:I think five years also ignores the reality that Google's millions of hours driving without incident or whatever is all in ideal scenarios. They don't take the vehicles into "difficult" terrain (hills, tight bends, etc) nor into "bad" weather (rain, night). I agree, this is why I think that augmented "self-driving" roads will be the way forward. I think they'll probably start with some sort of "HOV" lane that is separated from a primary freeway with jersey barriers, and that area will be fully monitored and have sensors and signaling equipment. I definitely think a five-year span is optimistic and hard to conceive. But I also see it happening. Technology moves fast, especially when there are inherent economic incentives both on a micro and macro scale. Also I completely understand the knee-jerk reaction to the concept. I get a very creepy feeling about the idea and there are tons of things I hate about it, but you simply cannot argue with it when looking objectively at the pros and cons. At this point it's honestly inevitable. What's more likely; governments worldwide invest heavily in driver education and licensing on a level comparable to UHC in an attempt to make roads safe, or we go with current technology that makes lives better as a whole?
|
|
#
?
May 29, 2014 00:07
|
|
|
revmoo posted:I agree, this is why I think that augmented "self-driving" roads will be the way forward. I think they'll probably start with some sort of "HOV" lane that is separated from a primary freeway with jersey barriers, and that area will be fully monitored and have sensors and signaling equipment. I'm not trying to mock you here, but have you put any thought into just how absurdly expensive that lane you're talking about is? A quick bit of googling says a mile of a single lane of freeway runs 1.6-3.5 million dollars, and that's in rural areas. Blocking off that kind of highway real estate to standard traffic use is absurd to even consider from a return on investment of tax dollars perspective. To give it perspective, the Eisenhower Freeway System alone is 47,714 miles of road. That doesn't include any of the state roads, surface streets, or such that are actually necessary to get anywhere a commuter would want to go. Adding 1 lane to that would start at around $75 billion, going sharply up when you consider how much of it is in urban areas where there is no room for additional lanes to handle current overcrowding.
|
|
#
?
May 29, 2014 00:15
|
|
|
revmoo posted:I agree, this is why I think that augmented "self-driving" roads will be the way forward. I think they'll probably start with some sort of "HOV" lane that is separated from a primary freeway with jersey barriers, and that area will be fully monitored and have sensors and signaling equipment. Liquid Communism is right. Augmented roads would be insanely expensive. CDOT was too broke to repair and expand a 10 mile stretch of a 4 lane highway to a 6 lane highway. I'm not even sure augumented road tech even exists, much of the R&D focus seems to be assuming it doesn't. I don't see it as a knee jerk reaction. My employer makes the tech and has active R&D for autonomous vehicles, yet the progress I've seen is actually pretty slow. Same news from my contacts in the field (old school friends, etc). My experience in the tech field indicates that current tech can evolve pretty fast, but inventing new poo poo is drat slow. People have been trying to do flying cars for 60 years now, many decades longer than autonomous ones, yet that hasn't really gone anywhere either. Eventually it may happen, but I'm not putting many bets down that it'll happen anytime soon, much less in my lifetime.
|
|
#
?
May 29, 2014 00:52
|
|
|
I can't help but see this in my head when I look at that Google car.
|
|
#
?
May 29, 2014 02:25
|
|
|
So this is kind of what I do for a living and I'm open to answering any questions you guys might have. First thing I'll say is that the biggest hurdle for these cars won't be technology, but insurance companies and lawyers.
|
|
#
?
May 29, 2014 03:40
|
|
|
Frinkahedron posted:So this is kind of what I do for a living and I'm open to answering any questions you guys might have. First thing I'll say is that the biggest hurdle for these cars won't be technology, but insurance companies and lawyers. That's pretty much the biggest hurdle for 99% of ____________.
|
|
#
?
May 29, 2014 09:01
|
|
|
I think it would be super cool to have for cargo. Imagine a bunch of autonomous trucks just driving along. Unless the cargo was high-priority or perishable, it could be made really efficient. Drive only when traffic is light, drive at slow speed to conserve fuel, divert to another destination ad hoc. This could be coupled with "autonomous-only" lanes as already mentioned, and maybe just between certain hubs, leaving the local distribution to normal box trucks. Of course, at this point, it might make more sense to just make all this a railroad network (or convert the one we have.) The metro in Copenhagen is driverless right now, and has been for ten years. I don't know how autonomous it is, though, as there's 24/7 human monitoring. It's just monitored, not operated, though. I believe it, in theory, could run 100% on its own, at least for some time.
|
|
#
?
May 29, 2014 10:25
|
|
|
I'm all for this as long as it doesn't interfere with my ability to drive for fun. For the vast majority of people's day-to-day car use, autonomous cars are going to be safer and more efficient.
|
|
#
?
May 29, 2014 14:17
|
|
|
That is the ugliest piece of crap.
|
|
#
?
May 29, 2014 14:52
|
|
|
CommieGIR posted:That is the ugliest piece of crap. You're not gonna own it; do you care what your taxicab looks like?
|
|
#
?
May 29, 2014 16:00
|
|
|
Was it BMW that proposed "packs" of high-speed, self-driving cars on highways? I'd be all for autonomous highway cars that are legally allowed to travel at ~100+mph. Plus, interstate highway driving is typically awful. PLEASE get this system working for interstate truck traffic.
|
|
#
?
May 29, 2014 16:06
|
|
|
Brigdh posted:People have been trying to do flying cars for 60 years now, many decades longer than autonomous ones, yet that hasn't really gone anywhere either. This is so far different than self driving cars, it's comical. Nobody is working on flying cars except some crazy buttcoin basement dwellers.
|
|
#
?
May 29, 2014 17:03
|
|
|
Yeah, the way I see this coming about is more like the process electric cars have gone through - a limited production (Tesla) available to the fairly wealthy (expensive), legalities sorted out by those interests, and the technologies trickling down. There's no particular reason first gen autonomous cars have to be combined with a "$75 billion+" road upgrade - we're perfectly capable of building (and selling) a $500k (just throwing a number out, it might be less, it may be well more) car that can do it on it's own, with existing infrastructure. I see the first iterations of it basically trying to make them as unintrusive as possible to a non-autonomous public - they'll behave themselves in traffic, follow the same rules, etc. As they become more ubiquitous, you might see the "special highways just for autonomous cars" or "roving packs of high speed BMWs"*. * - wait, we have these now Bumming Your Scene posted:This is so far different than self driving cars, it's comical. Nobody is working on flying cars except some crazy buttcoin basement dwellers. They're certainly not mainstream, but they're still a thing.
|
|
#
?
May 29, 2014 17:09
|
|
|
Bumming Your Scene posted:This is so far different than self driving cars, it's comical. Nobody is working on flying cars except some crazy buttcoin basement dwellers. Hey now, I don't think kastein's into buttcoins.
|
|
#
?
May 29, 2014 17:09
|
|
|
I would have no reservations betting that a small, affluent city with regularly maintained traffic control devices could have a small fleet of these as a local only taxi service, within a few years, easy.
|
|
#
?
May 29, 2014 17:26
|
|
|
KYOON GRIFFEY JR posted:You're not gonna own it; do you care what your taxicab looks like?  I have standards.
|
|
#
?
May 29, 2014 18:04
|
|
|
Un-l337-Pork posted:Was it BMW that proposed "packs" of high-speed, self-driving cars on highways? I'd be all for autonomous highway cars that are legally allowed to travel at ~100+mph. Plus, interstate highway driving is typically awful. It's a part of an EU research project known as SARTRE, and afaik Volvo have been doing the trying and testing so far. Like so.
|
|
#
?
May 29, 2014 18:24
|
|
|
Bumming Your Scene posted:I would have no reservations betting that a small, affluent city with regularly maintained traffic control devices could have a small fleet of these as a local only taxi service, within a few years, easy. And it'd travel to that city to check that poo poo out in a hearbeat. It's kind of like going to NYC, you don't really need to take the subway, but it sure is fun and an interesting way to get around a city. And while I feel sorry for my immigrant brothers, taxi cabs need to go the way of the elevator butler. bolind posted:I think it would be super cool to have for cargo. Imagine a bunch of autonomous trucks just driving along. Unless the cargo was high-priority or perishable, it could be made really efficient. Or even if it's perishable, just have the trucks cooled. I was thinking about this the other day, with the aging population, we don't want these old fucks driving anyway, so why not take grandmas car away and replace her weekly grocery run with an automated system. Morphix fucked around with this message at 19:14 on May 29, 2014 |
|
#
?
May 29, 2014 19:11
|
|
|
The way how I see it is it will be to expensive for the normal person to own a car in the automated future. So you will pay 200-400 dollars a month to have a share in a car service company like Uber or ZipCar. You tell the company your normal schedule like work, day care, and weekly chores/outing. And it assigns a car for you on that day to pick you up and drop you off. The. The car takes other people around when you don't need it, a another car picks you up. If you have a extra trip to take one will be by to pick you up according to the service level you pay for. The you will not need garage, insurance, or maintenance. The only issue being you will never own it. And as for human driven cars they will only be allowed on some roads and be more for sport then an everyday thing.
|
|
#
?
May 29, 2014 22:24
|
|
|
Un-l337-Pork posted:
I think there are some people in AI that will disagree with that statement.
|
|
#
?
May 29, 2014 22:29
|
|
|
Bumming Your Scene posted:This is so far different than self driving cars, it's comical. Nobody is working on flying cars except some crazy buttcoin basement dwellers. No, it's not. When you look at the real world, it's just as hysterically not going to happen without a massive technology step forward. Yes, you have "working" prototypes, but they are so heavily compromised, no sane person would consider them. Google should stick to building it's AI and leave the self driving car fantasy to the comics where it belongs for the next couple of decades.
|
|
#
?
May 29, 2014 22:41
|
|
|
Cat Terrist posted:No, it's not. When you look at the real world, it's just as hysterically not going to happen without a massive technology step forward. Yes, you have "working" prototypes, but they are so heavily compromised, no sane person would consider them. There are new Mercedes models that can steer themselves, right now, and it will keep it's distance to the car in front and brake automatically. Cars can park themselves now. The self driving cars can already detect pedestrians and bicyclists and bicyclist hand signals. What massive step forward is missing? https://www.youtube.com/watch?v=SxGY4iH5AAc Like, people say "oh how is this going to work at night when you can't see??". They put human deficiencies on it. Pop night vision or infrared on there, tech that has existed for decades. Longpig Bard fucked around with this message at 23:19 on May 29, 2014 |
|
#
?
May 29, 2014 23:12
|
|
|
So are self driving cars the new tech singularity thing here? Instead of "I don't have to work out because I'll be able to download myself into a robot body" it's "All road problems will be solved in 5 years because self-driving cars  " It's kind of obnoxious to see so many people buying into the we won't need stoplight and will drive 100 mph on all roads all the time, because computers will handle it all. I dunno, I like the idea of self driving cars, and would very much like to see auto-pilot as a feature on cars, but I don't think they're going to be replacing human driven cars wholesale just yet. I think it'll be a novelty feature on higher end cars that might trickle down given a decade or two, and maybe well see some taxi companies using them and the oddball tech nerds will buy a few, but that's it. I also really hate the idea of not owning the car and having to wait on it to be dispatched from some central depot to come pick me up. I'd really rather have own car, that I own, that's been made comfortable for me, and that I can jump in at 2 am to go to Taco Bell.
|
|
#
?
May 29, 2014 23:18
|
|
|
ExplodingSims posted:and that I can jump in stoned at 2 am to go to Taco Bell. There ya go.
|
|
#
?
May 29, 2014 23:21
|
|
|
keykey posted:There ya go. Again, a problem fixed by an auto-pilot accessory. Just run into the garage, sit my stoned rear end down and activate auto-pilot to go to the nearest 'Bell. No phone calls, not dealing with some call center shitlord telling me this is exceeds my car rental level and will be charged extra, no waiting 30 mins for the drat thing to show up, etc, etc...
|
|
#
?
May 29, 2014 23:26
|
|
|
I'm seeing more of a kiosk type thing. The fleet of 5 cars are around town doing their thing, you go to kiosk A12, see a movie theater at D23, swipe your card and it tells you the nearest car will pick you up in 13 minutes. poo poo, steal that Tesla automatic battery swap idea and they would barely have any down time  I don't think they'd start off with 2 am door to door Taco Bell rides. But, I guess that's more money coming in if you can go to their address. Longpig Bard fucked around with this message at 23:44 on May 29, 2014 |
|
#
?
May 29, 2014 23:33
|
|
|
Bumming Your Scene posted:There are new Mercedes models that can steer themselves, right now, and it will keep it's distance to the car in front and brake automatically. Cars can park themselves now. The self driving cars can already detect pedestrians and bicyclists and bicyclist hand signals. What massive step forward is missing? They detect these things in ideal conditions. There is plenty that isn't detected that is necessary to handle a random trip to the grocery store, much less doing it when its raining or dark or snowing. It seems like you are pretty set in your belief that we can do this tomorrow, despite having no evidence that you are in any way even in this field. However, I'll bite. One massive step forward is advances in what is known as computer vision. Computer vision is basically having a computer analyze the data from a camera and understand what it is "seeing". Such technology is necessary for reading speed limit signs. You can't dump a data base of every speed limit for every road, because the speeds change. Construction will take a 75 zone and drop it to 55. Hell, the signs even move so you can't plot the GPS coordinates of every one. The problem with computer vision, is its an extremely hard problem to solve. No one has figured out how to do it much better than brute forcing the problem. That takes a lot of computational power a lot of time to do. The state of the art computer vision systems I've seen tend to choke if the vehicle is moving faster than 30 mph. A surprising amount of driving relies on seeing things, and that's something that computers massively suck at. Its why LIDAR, RADAR, and SONAR are primarily used for autonomous vehicles, but each system has its own drawback. Oh, and that video you referenced? It was "staged". The vehicle was fed a bunch of advanced information that made that one particular trip possible, but wouldn't work in the general case. Sort of a proof of concept - if a bunch of hard problems were removed because we know in advance what we want the car to do, then yes, it can drive down the street. Impressive none the less, but still shows a lot of work to be done.
|
|
#
?
May 29, 2014 23:45
|
|
|
Brigdh posted:One massive step forward is advances in what is known as computer vision. Computer vision is basically having a computer analyze the data from a camera and understand what it is "seeing". Such technology is necessary for reading speed limit signs. You can't dump a data base of every speed limit for every road, because the speeds change. Construction will take a 75 zone and drop it to 55. Hell, the signs even move so you can't plot the GPS coordinates of every one. A truly autonomous car will be able to ascertain based on road conditions (cones, traffic ahead, open areas in what should be a smooth road, etc) what a construction zone is, as well as sensing what traffic around it is doing, and negate the need for it to even read the signs. The car would be programmed to not exceed a safe limit, would not have the human tendency to make errors in judgement regarding that, and thus wouldn't require the same limitations.
|
|
#
?
May 30, 2014 00:06
|
|
|
I won't even guess what google uses to determine where the road is, but I would bet reading signs and striping was a cake walk. Traffic control devices are retro reflective, that trumps any other thing out by the road. I have taken photos with a pocket camera with flash on a couple hundred feet away in broad daylight and could still see the sign brightening up. I didn't say they could do this across the country tomorrow. I said a municipality with traffic control devices under decent maintenance. I also don't believe these would be anywhere where it snows for a long time.
|
|
#
?
May 30, 2014 00:09
|
|
|
|
| # ? May 5, 2024 03:19 |
|
|
Brigdh posted:
I did my graduate research on computer vision for unmanned ground vehicles. With a modern laptop, I could process about 6 cameras worth of images at roughly 10 frames per second, and that was more a proof of concept for the cameras themselves and not so much focused on the speed of processing. The LIDARs used by Google and a lot of other companies spin at that rate, so we considered it a very good goal to hit for our dinky laptop. There's nothing wrong with throwing more computing power at a problem, it sometimes really is the best way, especially with statistical based methods such as machine learning. And luckily for us, computers double in speed every 18 months or so. Bumming Your Scene posted:I won't even guess what google uses to determine where the road is, but I would bet reading signs and striping was a cake walk. Traffic control devices are retro reflective, that trumps any other thing out by the road. I have taken photos with a pocket camera with flash on a couple hundred feet away in broad daylight and could still see the sign brightening up. You can actually read retroreflective signs with LIDAR, they're that good. (Both the signs and the LIDAR). Frinkahedron fucked around with this message at 01:03 on May 30, 2014 |
|
#
?
May 30, 2014 00:59
|
|