|
Post by rubbertoe on May 12, 2011 7:19:41 GMT -8
Wow, Google is trying to get Nevada to pass legislation allowing for autonomous vehicles: www.nytimes.com/2011/05/11/science/11drive.htmlI didn't think this would see the real world for many years. But the idea of working out the legal and technical kinks in a smaller geographic area makes sense. The last line where the person picks up their phone and says "Zipcar, come pick me up" would be a truly revolutionary moment. The car that you drive to work and home could serve multiple other users when you are at work. It could go take other people different places, then come back to work when you need it. Cars would become part of the transit infrastructure instead of personally owned. The cost of use would be much less, they would (theoretically) be much safer, more fuel efficient, and fewer would be required since they are doing much more than they do now when owned by a single individual. Less parking spaces would be required, since the need to park the vehicle for long periods goes away. Scheduling this would require a data connection and smart computers, which are readily available almost everywhere now. Smart cars driving from LA to Vegas could speed up the traffic and thats just for starters .If Google can pull this off, it could be a game changer, in more ways than one. RT
|
|
|
Post by jeisenbe on May 12, 2011 20:22:09 GMT -8
So, this would be Personal Rapid Transit, without the grade-separated guideways
I don't think it will work. Humans have large parts of their brain devoted to pattern recognition and spatial imaging, and we still spend months practicing driving to get comfortable at it, and to learn the explicit and cultural rules of the road.
How will a driverless car react to a childs ball that rolls into the street? Will it know to stop, in case a kid darts out after it? How much room will it use to pass someone on a bike, who is riding next to an irregular line of parked cars? Will it yield properly at roundabouts?
Consider that there are hundreds of grade separated metro lines in the world which still use human drivers, despite having not conflicts with cross-traffic and totally predictable patterns. There are a handful of automated, grade-separated lines, but modern practice requires the installation of automatic gates at stations to make them foolproof.
And there are hundreds of partially at-grade train and bus rapid transit and commuter rail systems that all use human drivers, even though automating them would be a million times simpler than automating a car that can drive on any number of streets and highways.
Before we see driverless personal automobiles, we need to see automated light rail, then automated BRT, then automated mixed-traffic streetcars, then mixed-traffic buses. Only then should someone attempt a driverless private car.
|
|
|
Post by rubbertoe on May 13, 2011 7:55:15 GMT -8
Yes, PRT without the guideways is a good way to think of it.
Concerning the "technical" ability of a car to drive itself, I would think that the scenarios that you mentioned: 1) ball in street, 2) bicycle passing, 3) irregular parked cars, 4) roundabouts, and even 5) all combinations of above. These all fall into the category of "known" events that need to be accounted for when driving. There is no doubt in my mind that all these and more were planned for and tested well before Google turned their cars loose on the streets. Remember, these cars drove a combined 140,000 miles so it isn't like they did a one day test. There are even more challenging ones: plastic bags floating around busy freeway traffic, flattened cardboard boxes being uplifted by cars on a freeway, things sitting between lanes on a freeway that cars need to avoid while still at speed, the list is endless...
I also read that they recorded every drive of every car, and analyzed each event that required the driver to hit the "red button" that returned control to the human driver. Which then allowed for incorporating the additional "lessons learned" into the software running the cars. I haven't read any more detailed documents of the entire process, like a scientific paper or anything. It could be that Google is keeping this closely guarded as you might imagine.
There are current systems available that assist in keeping cars centered in a given lane of traffic. This is available on higher end models, and is an amazing safety feature. The NHTSA is considering making it mandatory on all vehicles. If you think about it, a large percentage of accidents have to be started by distracted drivers (texting, phones, etc.) accidentally swerving into other lanes and triggering collisions. So, some level of automation is already creeping in, and there is also the adaptive cruise control too to automatically follow other vehicles.
The beauty of this system is that it is entirely self-contained. No monstrously expensive external infrastructure (guideways, embedded road magnets, dedicated proprietary data feeds, etc.) is required, the car just needs to be smart enough to do what a human driver would do. No doubt these are expensive now, but electronics cost has a way of rapidly declining, and these kind of systems would be on the same kind of curve, especially given the possibly huge safety and cost gains.
The other interesting question is about how safe they are. I'm sure the Nevada legislation (I need to find this BTW) first sets up a framework for deciding where, when, how, and the timetable for allowing this to happen. No doubt also addressing the liability issues. It poses an interesting dilemma: Say there are 50,000 automobile related deaths a year. And then assume that if all the cars were automated, this would be reduced to 500 total. If you could cut the rate by a factor of 100, would it be justified to then require all cars to eventually be automated? How the insurance works in the 500 cases is also something that needs to be addressed, obviously. Is it better to have 5 people killed by a software problem, or 40 people killed by reckless drivers?
I think there are several things that are going to cause this to come into existence sooner rather than later: 1. Safety (less accidents) 2. Economy (fewer cars required, less road building, easier carpooling) 3. Gas consumption: Imagine if 3 people in your neighborhood all work at factory A. You might not know them, and you all drive to work separately. With this system, a single car could pick all 3 up and not increase your commute by much time, and get 3 cars off the road. 4. Less parking required: Fewer cars sitting around all day long means less space required for parking. This translates into more compact downtown areas which benefits everyone, including current transit riders.
The other thing is, this isn't an expensive game to get into, relatively speaking. Google put together 7 cars that cost maybe $300,000 each. The FLIR rotating at 10 RPM on top of the car is the only thing that really stands out as unusual. And even for that, you could have your engineering team try and get it to be more reasonably sized and blend with the car. Given that the hardware is available,the job then becomes one of writing the software and testing it to death (no pun intended). I wouldn't be surprised if there aren't 10 other companies working on this as we speak. I look forward to seeing how this all develops.
Scary video on top of parking garage:
RT
|
|
|
Post by rubbertoe on May 13, 2011 9:10:51 GMT -8
Here it is, AB 511, requires a 2/3 majority vote, Section 8 of the bill is the autonomous vehicle section... Sec. 8.1 The Department shall adopt regulations 22 authorizing the operation of autonomous vehicles on highways 23 within the State of Nevada. 24 2. The regulations required to be adopted by subsection 1 25 must: 26 (a) Set forth requirements that an autonomous vehicle must 27 meet before it may be operated on a highway within this State; 28 (b) Set forth requirements for the insurance that is required to 29 test or operate an autonomous vehicle on a highway within this 30 State; 31 (c) Establish minimum safety standards for autonomous 32 vehicles and their operation; 33 (d) Provide for the testing of autonomous vehicles; 34 (e) Restrict the testing of autonomous vehicles to specified 35 geographic areas; and 36 (f) Set forth such other requirements as the Department 37 determines to be necessary. 38 3. As used in this section: 39 (a) “Artificial intelligence” means the use of computers and 40 related equipment to enable a machine to duplicate or mimic the 41 behavior of human beings. 42 (b) “Autonomous vehicle” means a motor vehicle that uses 43 artificial intelligence, sensors and global positioning system 44 coordinates to drive itself without the active intervention of a 45 human operator. – 5 – - *AB511_R1* 1 (c) “Sensors” includes, without limitation, cameras, lasers and 2 radar. Looks like they are focusing on the operation on highways. And the bill just tells the DMV to start the process of developing the guidelines. Kind of makes sense with all the long stretches of open road. And the testing will be in limited areas. The other interesting thing is the wording is such that they define "autonomous vehicle" to allow for the possibility of no human operator even being present. And a link to follow the bill: nelis.leg.state.nv.us/76th2011/App#/76th2011/Bill/Overview/AB511RT
|
|
|
Post by jeisenbe on May 13, 2011 18:23:49 GMT -8
In most states, "highways" includes all streets and roads.
If this were limited to just grade-separated freeways I would be fine with it. But I'm not sure I trust even Google to design something that can be used on urban streets and residential areas.
Let them perfect an automatic bus for BRT first, and then we can talk about personal cars.
|
|
|
Post by rubbertoe on Jun 4, 2011 16:11:20 GMT -8
AB511 passed the Nevada Assembly 36-6 on 5-30-11. The bill was moved to the Senate and is now being considered by the finance committee. All autonomous vehicle provisions remain intact.
RT
|
|
|
Post by rubbertoe on Jun 25, 2011 15:31:49 GMT -8
The bill has now been signed into law by the Governor. The government website was a little fuzzy about the process, but now I have seen a couple links to the story: cyberlaw.stanford.edu/node/6688technolog.msnbc.msn.com/_news/2011/06/24/6937087-nevada-gives-robo-cars-a-green-light-I shot off a letter to the Nevada DOT a couple weeks ago to see if there was anything more they could tell me about the process or the timeline. Here is what they had to say: The USDOT’s National Hwy Traffic Safety Administration (NHTSA) sets regulations governing vehicle equipment standards. NHTSA is in the process of testing and setting standards for such vehicles. Ultimately, there will be a collaborative effort involving NDOT, NHP and DMV to set NV State requirements based on NHTSA’s findings and guidance.
Obviously, this will take some time.The second story says the law takes effect 3-1-12, but no word whether they would start formulating the plans before then. RT
|
|
|
Post by rubbertoe on Jun 25, 2011 15:59:22 GMT -8
And it looks VW is also looking at the potential of semi-autonomous cars. They seem to be focusing on highway driving: blogs.forbes.com/jimgorzelany/2011/06/24/a-car-that-can-drive-itself-at-80mph/TAP = Temporary Auto Pilot. Interesting alternative. By restricting the autonomous functions to highway driving, you make the problem MUCH simpler. Highways after all being "exclusive use" for vehicles, no people to worry about for the most part. Cars already have adaptive cruise control and lane assist. Those two capabilities alone are basically highway driving. This makes me think that Nevada might initially restrict the autonomous cars to highways, maybe a certain area for starters. They could roll it out in a less dense corridor than the Vegas/SoCal 15 area, then allow it there. Not sure off the top of my head whether having 10% self driving highway vehicles would speed things up overall, versus say 30%. There has got to be a point at which having 20-30 vehicles in an "autonomous train" will save some time, unless of course there are 2 cars driving side by side going 55 MPH. While you could restrict the autonomous cars to only use a single lane, that would kind of defeat the purpose of the autonomous vehicles being able to work with other vehicles, especially if you need to build more lanes, which you should be able to avoid. Maybe when the 15 gets to 3 lanes as it gets closer to Vegas, they could have 1 lane be for AV's only. I haven't been to Vegas in awhile, but there are only about 40 miles from the border to the city itself, and a lot of that is already 3 lanes wide. Another very good candidate for AV technology would be toll lanes, like the Express Lanes being started here on the 10 and 110. The 110 has 2 lanes in each direction in the corridor, you could make the leftmost lane AV only. If those cars could travel closer and faster, it could increase capacity, though you still have the problem of them merging into the regular traffic when the HOT lanes end. In a perfect world, the HOT lanes would end in a manner that would allow them to empty/merge in a manner that wouldn't create an even bigger backup than they do now. RT
|
|
|
Post by jeisenbe on Jun 27, 2011 22:22:26 GMT -8
I don't think automation would make a big increase in capacity for limited-access highways. Cars will still need to merge on and off the highway, so even if ALL cars, buses and trucks had an auto-pilot, there would need to be gaps for merging. And not every vehicle can accelerate as quickly, or brake as quickly, so you need some space in between.
Traffic engineers say that a limited-access, grade-separated motorway (freeway) can carry about 2000 cars per hour per lane, max, compared to about 1000 per hour for a lane on a street with traffic lights. 2000/60min is a car every 1.8 seconds, on average, meaning that sometimes the cars are even closer together. Could automation really improve on that?
The benefit would be basically allowing "PRT" (personal transit), or more appropriately "driverless taxis", where people could text/call/talk/write on this forum while in a car. Personally, I think the bus is a great solution to that need.
|
|
|
Post by rubbertoe on Aug 5, 2011 10:14:32 GMT -8
|
|
|
Post by rubbertoe on Nov 14, 2011 7:40:12 GMT -8
Todays New York Times article says that Google is considering manufacturing the automated car in the US: www.nytimes.com/2011/11/14/technology/at-google-x-a-top-secret-lab-dreaming-up-the-future.html?_r=3&pagewanted=all"Google may turn one of the ideas — the driverless cars that it unleashed on California’s roads last year — into a new business. Unimpressed by the innovative spirit of Detroit automakers, Google now is considering manufacturing them in the United States, said a person briefed on the effort." Google apparently has a secret lab or labs where they are working on some of the more "further out" ideas. Kind of makes sense, given all the money they have. A lot of the stuff that we take for granted now was the stuff of science fiction as little as 5-10 years ago. I'm still waiting for the day that one of the Google co-founders steps up to the plate and writes a check to make the "private investment" that gets the HSR system built. I recently drove past and tried to get into the Ivanpah SEGS on the California/Nevada border. That will be quite a facility when it gets finished and comes online. RT
|
|
|
Post by rubbertoe on Feb 17, 2012 10:16:57 GMT -8
|
|
|
Post by rubbertoe on May 7, 2012 17:18:43 GMT -8
|
|
|
Post by rubbertoe on May 22, 2012 7:15:37 GMT -8
|
|
|
Post by Elson on May 22, 2012 12:35:16 GMT -8
Would they be called....AUTONOMOBILES?
|
|
|
Post by rubbertoe on Sept 25, 2012 17:38:13 GMT -8
|
|
|
Post by rubbertoe on Dec 5, 2013 6:16:09 GMT -8
|
|
|
Post by Gokhan on Dec 5, 2013 7:31:59 GMT -8
While I like Google a lot and use their products and Android phones, I think they grossly overestimate their abilities.
Take for example the Google GPS Navigation. It's still inferior to most commercial navigation systems in reliability. You can't always rely on it to take you to your destination but you will rely on Google automatic pilot not to send your car into a ravine?
Also, their constant highly unnecessary yet drastic updates on their products have got to the point of being very annoying. Take for example Google Maps. A few years ago, it was an excellent product. Now, it's alternating between not nearly useful and difficult and awkward to use with every update.
They've also stopped spending enough effort on their Android system. The latest updates they've had are very incremental and the 4.1.1 version I have has severe power-management problems, which I doubt have been fixed in 4.4.
So, is Google doing well these days? I don't think so. I think the company is being poorly managed and the quality of their products is on a decline, and their poor management is getting worse and worse.
|
|
|
Post by rubbertoe on Dec 10, 2013 7:44:54 GMT -8
While I like Google a lot and use their products and Android phones, I think they grossly overestimate their abilities. Take for example the Google GPS Navigation. It's still inferior to most commercial navigation systems in reliability. You can't always rely on it to take you to your destination but you will rely on Google automatic pilot not to send your car into a ravine? Also, their constant highly unnecessary yet drastic updates on their products have got to the point of being very annoying. Take for example Google Maps. A few years ago, it was an excellent product. Now, it's alternating between not nearly useful and difficult and awkward to use with every update. They've also stopped spending enough effort on their Android system. The latest updates they've had are very incremental and the 4.1.1 version I have has severe power-management problems, which I doubt have been fixed in 4.4. So, is Google doing well these days? I don't think so. I think the company is being poorly managed and the quality of their products is on a decline, and their poor management is getting worse and worse. Gokhan, While the post title specifically mentions Google, there are a bunch of other companies also working on this technology. Also, while GPS is used by the Google vehicles, their primary knowledge of their surroundings is from the LIDAR on the roof. I attended a 2 hour presentation and they showed video of the cars perspective of where it is. The LIDAR showed the exact location down to the centimeter level, including all surrounding vehicles, moving pedestrians, and inanimate objects. And they also showed the GPS location of the car, which was often 10-20 feet away from where the actual vehicle was. Which is why the cars can't rely on GPS alone. Heres a few link to more recent non-Google news for anyone interested. Volvo will be testing 100 cars in Sweden: www.theverge.com/2013/12/3/5169848/volvo-drive-me-autonomous-car-pilot-projectAnd Nissan is on record that they will be selling these as a product by 2020: www.theverge.com/2013/8/27/4664272/nissan-will-bring-self-driving-cars-to-market-by-2020I guess the main points that I would make here are the following, and the last will show why these developments are incredibly important to transit advocates: 1. This is something that there is no doubt is going to happen. Guaranteed. Will every car on the road in 2025 be autonomous? No, probably less than 1%. But the important idea is that the amount of money that will be saved is so enormous, and also the safely implications, which is why every car maker and technology company is going full speed ahead. I can point you to a series of Forbes articles where the author estimates the impact of driverless autos will be in the multiple trillions of dollars, including who will be the winners and losers. You don't want to send your kids to school to be cab drivers or auto insurance salesmen for example. 2. Does the autonomous car need to be completely autonomous 100% of the time, in 100% of the possible conditions? No. Certain road conditions are very tricky. The presentation mentioned icy roads as one of the biggest challenges for example. But here in SoCal you aren't going to be seeing a lot of icy road issues, so the car isn't facing that 99.999% of the time. Also, and this is probably how they will initially be incorporated onto the roads, designated freeway lanes will be turned into "autonomous enabled lanes". On these lanes autonomous vehicles will be allowed to operate, along with other vehicles. This will have an immediate and dramatic effect on freeway congestion. 3. Even a limited number of autonomous vehicles mixed in with regular traffic on freeway lanes will greatly reduce congestion. I read a great article on this recently, but I can't seem to find the link. Will post it later if I can dig it up. 4. The biggest impact for transit will be that since road congestion at some point will no longer be such an issue, at that point there will no longer be a need for freeway expansion. When you look at how much Federal money flows to highways versus other transit, there may now be much more money available if freeway expansion becomes unnecessary. Essentially what is happen is that this technology will allow the current roads to carry many more people much more quickly and safely, getting closer to their optimum carrying capacity. This is a win, win, win situation for all involved: safety, hours lost in traffic, gas consumption, etc. Concerning building more freeways, I like the following analogy. Autonomous vehicles are at about the point right now that cars were when they were first invented. If you were a company that owned a large horse drawn buggy business, would you go "all in" and expand your business to be able to build more horse drawn buggies, or would you start paying close attention to the new technology in case it might adversely affect your business? I think all government entities should start looking ahead a few years to make sure that they aren't making any spending decisions that might in the long run prove to be pointless. I'll give one example here, though I don't claim that this is the best one or one that it will actually prove to be disastrous. Metro is looking to find a partner to build a rail tunnel under the Sepulveda Pass. This may include a separate tunnel for trains, maybe not. In any case, the basic idea is that the private contractor will pony up a bunch of money so that they can charge large tolls for the next 20-30 years for use of the "faster" tunnel. Well, remember that that there are 5 (?) lanes of traffic on the 405 going in both directions. And yes, it is crazy congested, so much so that you almost can't imagine freely flowing traffic going South at 7:30am on a weekday. If the introduction of autonomous vehicles allows for average speeds to increase (possible substantially), at some point in the future, even if it does take 10-15 years to reach that point, what effect do you think that will have on the company charging $10 to drive through the tunnel? Any company ponying up several billion $ to help build this tunnel is going to be thinking about that possibility long and hard before signing a check that big. If I were Metro, I would get this PPP project kicked off as soon as possible, otherwise the possibility of doing it will have been overcome by events. RT RT
|
|
|
Post by rubbertoe on May 22, 2014 11:54:15 GMT -8
|
|