Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
That is not how it works. They use a mess of inputs...GPS, Inertial, topographic, features and lane markings. All that is run through a set of filters to determine an accurate position. If the car thinks it knows where it is but not precisely it may pick up a telephone pole or a fire hydrant and use that to get a precise location which it can then maintain by a combination of inertial and GPS.
And they will be pretty good at evaluating how well they know there position.
But they will have no trouble maintaining lane even where the snow has not been driven on and there are no obvious markings.
The fun cases is what does the car do when the lanes clearly being used are not correct. If you have driven a lot in snow you know that often one lane that is not aligned with the real road is often driven by everybody...now what does the autonomous car do? Does it follow the bad tracks...or does it try to hold a true lane. Now there is some fun programming.
They monitor the current traffic pattern. If cars have been driving in different lanes (of between lanes) it will use that information to make a decision.
First, it would just sit there. Because it would have no clue what to do. But if it could overcome that, by using utility poles, signs, trees, buildings to navigate, then it would drive with two wheels up in the snow bank, until it would get stuck. Which is why self-driving cars are not better then human drivers.
Ford’s, because that’s the only one that has even attempted to drive in the snow, and that was on a small well mapped out test track, with one inch of snow on the ground.
They monitor the current traffic pattern. If cars have been driving in different lanes (of between lanes) it will use that information to make a decision.
Yeah, right, it can do that. But if a city worker puts a couple of sand bags in the road for any reason it causes the car to freak out and crash into the side of a 60 foot articulated bus. But it can drive in snow with no problems. The technology that you are talking about is not part of what Google is testing.
You're missing the point. Bad weather makes things unpredictable. Computers and algorithms work on predictability and likely scenarios. There is no way to program for everything.
Didn't you read what they are using for algorithms? They DON"T program for everything. They program a few divergent scenarios then the algorithm LEARNS about the rest, just like a new driver learns. You don't teach a 16 year old driver everything there is to know about driving, do you? No, you give them the basics and then let the LEARN how to deal with unexpected scenarios as they arise, and then use what they've learned to deal with even more complex scenarios. That's exactly how the AI used in autonomous cars is working. The difference is that as the autonomous car learns, every other autonomous car gains the benefit of that leaning virtually instantly, making the entire population of them better at the same time. And each generation benefits and doesn't have to start from scratch, like a 16 year old human driver has to.
Add to that the idea that autonomous (and even new non-autonomous) cars will communicate with each other in real time and you see that very quickly, there will not be any problems interacting with each other and with current conditions at any point in time.
And, just like the answer to the EV argument about how the grid can't handle every car being replaced by EVs now, there is no physical way to manufacture enough autonomous cars to replace every non autonomous car in the next 50 years! There are 200 million cars in the US alone. Even if every manufacturer switched over to full autonomous car manufacturing tomorrow, it would take 30 years of continuous production to replace even HALF the total number of cars in the US. So worrying that your personal car will be taken away in your lifetime is stupid at best. What WILL happen is that the next generation of new cars will have transponders that communicate with each other, and it may be that your older car will be required to be retrofitted in order to conform to traffic laws and allow the few, but growing number of autonomous vehicles to coexist safely with non autonomous cars.
Yeah, right, it can do that. But if a city worker puts a couple of sand bags in the road for any reason it causes the car to freak out and crash into the side of a 60 foot articulated bus.
Except it won't do that again, and neither will any other future autonomous car, now that it's learned that scenario. THAT's the part you are willfully ignoring and can't seem to get through your head.
About a dozen miles from where I'm writing this, there's a physical obstacle which demonstrates why self-driving vehicles will be very difficult to implement once the "closed course" environment of the laboratory is left behind. It's a modest-sized hill, no more than 200 feet high. but the approaches on both sides consist of several curves, not all of them particularly tight, separated by short stretches of straight roadway, and the gradients on that highway are very dissimilar, actually incorporating several up- and downgrades within the main or principal slope on each side of the summit.
To top everything off (no pun intended), the road itself (a classic "two-lane blacktop") was originally developed sometime before the turn of the Twentieth Century, but is a low-numbered component of the state highway system. Minor accidents, usually involving a single vehicle leaving the roadway, are not all that uncommon; the drivers involved are often from out of the area and not familiar with the road. Injuries are rare, fortunately, due to the slower speeds mandated by permanent road conditions.
So I'm sure that any "self-driving" vehicle is going to be challenged by this little anomaly, and I wouldn't be surprised to discover that the people hoping to develop a broad-base autonomous vehicle system recognize that there are probably hundreds, even thousands of such pitfalls and challenges out there. Which leads me to conclude that the system will either have to involve a huge database, with local subsections offered for sale to motorists in a manner similar to maps and atlases, or that the next round of expansion beyond uniform city street grids would be on a relatively small number of major interstate highways-- preferably in flat country.
I am seeing the same problem with driverless cars arguments as other discussions we've had around here.
The naysayers want them to be 100% right and never have problems. If they are 99.99% right and manual drivers are currently 99.50% right they'll just keep pointing to the errors created by driverless and counting "all the mayhem" they cause.
I can tell you that with all the self-absorbed cell-phone using, pill-heads, drunks etc. out there I can see this product being a big improvement as they work out the kinks.
I am seeing the same problem with driverless cars arguments as other discussions we've had around here.
The naysayers want them to be 100% right and never have problems. If they are 99.99% right and manual drivers are currently 99.50% right they'll just keep pointing to the errors created by driverless and counting "all the mayhem" they cause.
I can tell you that with all the self-absorbed cell-phone using, pill-heads, drunks etc. out there I can see this product being a big improvement as they work out the kinks.
About a dozen miles from where I'm writing this, there's a physical obstacle which demonstrates why self-driving vehicles will be very difficult to implement once the "closed course" environment of the laboratory is left behind. It's a modest-sized hill, no more than 200 feet high. but the approaches on both sides consist of several curves, not all of them particularly tight, separated by short stretches of straight roadway, and the gradients on that highway are very dissimilar, actually incorporating several up- and downgrades within the main or principal slope on each side of the summit.
To top everything off (no pun intended), the road itself (a classic "two-lane blacktop") was originally developed sometime before the turn of the Twentieth Century, but is a low-numbered component of the state highway system. Minor accidents, usually involving a single vehicle leaving the roadway, are not all that uncommon; the drivers involved are often from out of the area and not familiar with the road. Injuries are rare, fortunately, due to the slower speeds mandated by permanent road conditions.
So I'm sure that any "self-driving" vehicle is going to be challenged by this little anomaly, and I wouldn't be surprised to discover that the people hoping to develop a broad-base autonomous vehicle system recognize that there are probably hundreds, even thousands of such pitfalls and challenges out there. Which leads me to conclude that the system will either have to involve a huge database, with local subsections offered for sale to motorists in a manner similar to maps and atlases, or that the next round of expansion beyond uniform city street grids would be on a relatively small number of major interstate highways-- preferably in flat country.
Any state or county numbered highway no matter how treachorous is not likely to be a problem. All such things are in the very earliest stages of mapping. It is very remote dirt roads not in any official system that will be the difficult ones. The skill level of the autonomous vehicle is very high when it comes to the mechanics of driving. And they have no ego to keep them from slowing when challenged.
The arguments about snow and limitied visibility really run the other way. These vehicles see much better than people and see multiple ways. Both video and lidar and radar and sonar may be in simultaneous use.
Navigation accuracy is improving rapidly. Cheap solutions to less than a foot are in hand. Inches are close. And it simply gets better and better as time goes on and new systems come on line. The real issue at this point in time is maintaining a high level of accuracy when no GPS signal is being gotten.
I would think we see the real thing by about 2020. I still think the first vehicle will be long haul trucks but who knows....
Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.
Detailed information about all U.S. cities, counties, and zip codes on our site: City-data.com.
Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
But if a city worker puts a couple of sand bags in the road for any reason it causes the car to freak out and crash into the side of a 60 foot articulated bus. But it can drive in snow with no problems.
The technology that you are talking about is not part of what Google is testing.
