Aren't self-driving cars impractical?

[Ripley Twenty-Nine]

So what? That is just GPS systems that exist to day. That so trivial to include that it is pointless to mention. But being able to see signs is vital, unless you are suggesting some way that the car can have the deer or the car the broke down on its database.

And if you are postulating such a perfect database, why not have flying hydrogen powered cars as well?

You seem to be missing my point. I thought we were talking about visually recognizing road signs (Speed limits, One Way signs, Construction), compared to knowing information about the roads from a database.

Of course it would have to know about current conditions, such as what was currently in front of it, or what the condition of the road was.

I apologize if I wasn't clear about this.

 

[Dave1001]

Sure, I was just dealing with the speed limit aspect. I probably didn't word my post carefully enough, or got off on a tangent.

Here's another question/issue - a lot of times, I'm navigating by following somebody's directions. You know, go down the street 'til you see the Citgo, then take the second right, then the first left. Or, I'm driving down the road and looking for a restaurant. When I see one that appeals, I pop in. Or, I want to drop somebody off at the door of a business or whatever, then drive and find a parking spot. Databases don't work for that stuff.

That's not an argument that databases are useless. For long distance point to point stuff, they are good. For more spontaneous stuff (find a unmuddy place to park so I can unload my bike near the bike path in a section where it is pretty), they aren't useful at all. Any automatic system is going to have to support that kind of interactive driving and very complex problem solving.


That's not an argument against anything you said personally, Ripley, but more free form observations that the problem is quite difficult, and we aren't close to having the technology to deal with some of the problems of full automatic driving.

I disagree, I think all the technology is already in place for all the major problems. That sort of interactivity as you describe (spontaneously changing or delaying destinations to stop at something you see along your route) seems to be a relatively small deal to me.

 

[Ripley Twenty-Nine]

Sure, I was just dealing with the speed limit aspect. I probably didn't word my post carefully enough, or got off on a tangent.

Here's another question/issue - a lot of times, I'm navigating by following somebody's directions. You know, go down the street 'til you see the Citgo, then take the second right, then the first left. Or, I'm driving down the road and looking for a restaurant. When I see one that appeals, I pop in. Or, I want to drop somebody off at the door of a business or whatever, then drive and find a parking spot. Databases don't work for that stuff.

That's not an argument that databases are useless. For long distance point to point stuff, they are good. For more spontaneous stuff (find a unmuddy place to park so I can unload my bike near the bike path in a section where it is pretty), they aren't useful at all. Any automatic system is going to have to support that kind of interactive driving and very complex problem solving.

That's not an argument against anything you said personally, Ripley, but more free form observations that the problem is quite difficult, and we aren't close to having the technology to deal with some of the problems of full automatic driving.

I fully agree.. I don't know if we would ever want a fully automatic car that we couldn't switch to manual.

Thankfully for me, getting directions from people are a thing of the past. I have absolutely no sense of direction, and 'Drive North' means nothing to me. Just give me an address, or a general location, even if it's 'It's the building next to the McDonalds downtown on Elm street'.

When my wife and I went to visit friends in Boston, I planned a route door to door (12 hours worth of driving), and was also able to find all of the hotels right off of the highway about 4 hours into our trip, get the phone number for one of the hotels, and book a room. I could also avoid some toll roads along the way if I chose to; and all of this on my Pocket PC before I even left the house.

What can I say, I'm a major techology geek! I really enjoy thinking about how technology can help make our lives easier.

 

[Dave1001]

This type of system might work best with a database and being able to respond to visual cues. 'Hey, that sure doesn't look like a road! It looks more like a staircase!'.

Bingo.

 

[ponderingturtle]

You seem to be missing my point. I thought we were talking about visually recognizing road signs (Speed limits, One Way signs, Construction), compared to knowing information about the roads from a database.

Of course it would have to know about current conditions, such as what was currently in front of it, or what the condition of the road was.

I apologize if I wasn't clear about this.

We are, but you need a visual system and having it recognise signs as well is trivial.

 

[EternalSceptic]

I think road safety could be greatly increased without having to go to self-driving cars. Simply implant every car with a GPS receiver, and have detectors at every intersection of every road. When you pass by it calculates your average speed between that detector and the last one you passed. If it calculates that you were over the limit, you get a fine in the mail.

I would very much like something like that in place, everywhere.

Partly this is already working here in Europe. At some toll roads the time at one toll station is simply printed on your bill and if you arrive too early at the next station you are fined.

 

[timhau]

Partly this is already working here in Europe. At some toll roads the time at one toll station is simply printed on your bill and if you arrive too early at the next station you are fined.

...which is grossly unfair to those of us who drive cars with wormhole generators.

 

[Cuddles]

We are, but you need a visual system and having it recognise signs as well is trivial.

But flawed. Why design a system that is no better than what we can do now when it is easy to make it better? Just because cars could be made to recognise signs does not mean it is the best solution, especially since you have admitted that they would suffer from all the same problems humans do. Low power radio transmitters instead of signs would be far more reliable and much easier to update. Getting the information into the car is trivial and the technology could easily be rolled out tomorrow. It is making cars capable of acting on it that is the tricky part.

 

[Dave1001]

But flawed. Why design a system that is no better than what we can do now when it is easy to make it better? Just because cars could be made to recognise signs does not mean it is the best solution, especially since you have admitted that they would suffer from all the same problems humans do. Low power radio transmitters instead of signs would be far more reliable and much easier to update. Getting the information into the car is trivial and the technology could easily be rolled out tomorrow. It is making cars capable of acting on it that is the tricky part.

You make a good point that in the long term signs may not be the most efficient way to communicate road information to self-driving cars. But they make sense for the mixed use period where human drivers and self-driving cars will be sharing roads. During that period, it's a good idea to have road signs. And ponderturtle's recurring point is that given the visual tasks we'll have to give self-driving cars to avoid accidents on the road from things like foreign objects, it's extremely easy to add in the capability to read speed signs placed there for the human drivers, so we might as well add in that technology.

Once there are no more human drivers on public roadways, I agree with you that the incentive to maintain driving-related signage and the lines of software code for cars to read them may disappear.

 

[Ove]

Wow, when was the last time you shopped for cars? 1-200 dollars won't even upgrade you to powered seats.

I was talking costprice not what you should pay for it.

Not to mention, personally, I don't think governers solve anything. While speed definately contributes to the severity of a crash, it's not the cause, poor driving is. If everyone on the road is going 50mph, then the safest speed for you is 50mph. People driving 10mph faster/slower are driving dangerously, as well as those cutting through lanes, or stopping in the middle of a busy street.

Sorry, but that's just one of my pet peeves, because around here, cops only care about speeding. I've seen a person cut across 3 lanes on the highway, in sight of a cop, and get away clean. Sure you need to keep your speed within reason, but the real killers out there are the people driving like idiots, the people the cops ignore because they're not easy or profitable to prosecute. The only way to get rid of them is to automate driving completely, though I'd have to wonder how many of them would still opt for the "manual" drive option, just so they can cut around all the other cars and get to their destination those all important 5min earlier...

All the more reason for fully automated driving.

 

[ponderingturtle]

You make a good point that in the long term signs may not be the most efficient way to communicate road information to self-driving cars. But they make sense for the mixed use period where human drivers and self-driving cars will be sharing roads. During that period, it's a good idea to have road signs. And ponderturtle's recurring point is that given the visual tasks we'll have to give self-driving cars to avoid accidents on the road from things like foreign objects, it's extremely easy to add in the capability to read speed signs placed there for the human drivers, so we might as well add in that technology.

Once there are no more human drivers on public roadways, I agree with you that the incentive to maintain driving-related signage and the lines of software code for cars to read them may disappear.

Exactly. Sure speed limits and locations of stop signs might be helpful, but those are specificaly for interacting in a mixed enviroment. If you are in a pure enviroment, you can have the car decide the safe speed, taking into account things like sight distance, road conditions and such.

 

[roger]

Once there are no more human drivers on public roadways, I agree with you that the incentive to maintain driving-related signage and the lines of software code for cars to read them may disappear.

I'm trying to figure out what the interaction between automatic cars and humans would be. Our streets are currently multi-use, and I don't see any change to that, even in 50 years. Too much infrastructure already exists. Right now, as a pedestrian or bicyclist, I can mostly count on the fact that I can continue on into an intersection when the light is green, or if there is a stop sign and I arrive there before the vehicle. Right of way and other road rules make it somewhat safe to be a pedestrian.

If there were no rules for auto-cars, then it would be extremely difficult to be a pedestrian. How do you cross roads, how do you cross in front of the outlet to a parking garage, etc? The velocities are so different that these things require (so far as I can imagine, I admit my imagination has it's limits) road rules.

So long as we share the roadways in any capacity, some kind of human readable signage will be required. What sensors the vehicles rely on to respond to that signage (visual, radio transmitter, database) is kind of an implementation detail.

 

[roger]

I was talking costprice not what you should pay for it.

Current cost of Thum's package is 40K. He argues it will go down to 2K with mass production. Source.

I have to say I love his enthusiasm for the idea. I dislike driving, and love doing things like reading. I'd love an automatic driving car, so long as I don't have to sit there and monitor it, waiting to react for some anomaly.

I'm wondering about a 'joystick' mode that gives you limited manual control. You know, point the joystick where you want the car to go, and the car will try to go that way. You couldn't drift across a lane, or straddle two lanes, but the car would change lanes safely as necessary. Point to the side of the road and pull back, and it will pull over to the side of the road. Push the stick forward, and it will wait for a lull in the traffic, then pull back out. I think with our current infrastructure we need that level of manual control - there's no way I'm surfing a database to get the coordinates for the loading dock at Lowes, or messing with google earth to figure out where I should park to put my kayak in the river.


ETA: those kinds of limited control schemes can lead to massive instability, however, as we humans tend to overcorrect when a system does not respond predicatably to our inputs. I'm dealing with that with a system we are building at work. It's not an impossible problem, but the engineering challenges are not trivial, either.

 

[Dave1001]

I'm trying to figure out what the interaction between automatic cars and humans would be. Our streets are currently multi-use, and I don't see any change to that, even in 50 years. Too much infrastructure already exists. Right now, as a pedestrian or bicyclist, I can mostly count on the fact that I can continue on into an intersection when the light is green, or if there is a stop sign and I arrive there before the vehicle. Right of way and other road rules make it somewhat safe to be a pedestrian.

If there were no rules for auto-cars, then it would be extremely difficult to be a pedestrian. How do you cross roads, how do you cross in front of the outlet to a parking garage, etc? The velocities are so different that these things require (so far as I can imagine, I admit my imagination has it's limits) road rules.

So long as we share the roadways in any capacity, some kind of human readable signage will be required. What sensors the vehicles rely on to respond to that signage (visual, radio transmitter, database) is kind of an implementation detail.

Well, my aforementioned post made clear that it's at the end stage that road signage and the ability of cars to read them may no longer be necessary. The problems you mention don't apply much to highways, certainly not to the fast lanes of highways.

As for mixed use self-driving cars and pedestrians, I agree signage will persist until something better comes along. Such as direct communications to (the even more user friendly versions of) smartphones. Signs seem like too inefficient a form of long term communication of physical movement rules to persist for more than a few decades, in my opinion. A precedent would be messenger pigeons and the pony express. Sure they might help ensure communication redundancy in 2007, but why would society bother with the expense?

 

[Horatius]

I'm wondering about a 'joystick' mode that gives you limited manual control. You know, point the joystick where you want the car to go, and the car will try to go that way.

ETA: those kinds of limited control schemes can lead to massive instability, however, as we humans tend to overcorrect when a system does not respond predicatably to our inputs. I'm dealing with that with a system we are building at work. It's not an impossible problem, but the engineering challenges are not trivial, either.

They've done some work on joystick controls for cars, that could be integrated into this system:

http://www.snopes.com/photos/automobiles/newmercedes.asp

The neat thing here is, either front seat passanger can drive the car.

 

[Myriad]

Well, my aforementioned post made clear that it's at the end stage that road signage and the ability of cars to read them may no longer be necessary. The problems you mention don't apply much to highways, certainly not to the fast lanes of highways.

As for mixed use self-driving cars and pedestrians, I agree signage will persist until something better comes along. Such as direct communications to (the even more user friendly versions of) smartphones. Signs seem like too inefficient a form of long term communication of physical movement rules to persist for more than a few decades, in my opinion. A precedent would be messenger pigeons and the pony express. Sure they might help ensure communication redundancy in 2007, but why would society bother with the expense?

I have this mental image, now, of a road where all the signs are bar codes.

And another mental image of an automated car passing a bar-code sign, stopping, backing up, passing the sign again, and so on, five or six times, until it finally goes "beep" and continues on.

Respectfully,
Myriad

 

[Ove]

Current cost of Thum's package is 40K. He argues it will go down to 2K with mass production. Source.

I have to say I love his enthusiasm for the idea. I dislike driving, and love doing things like reading. I'd love an automatic driving car, so long as I don't have to sit there and monitor it, waiting to react for some anomaly.

I'm wondering about a 'joystick' mode that gives you limited manual control. You know, point the joystick where you want the car to go, and the car will try to go that way. You couldn't drift across a lane, or straddle two lanes, but the car would change lanes safely as necessary. Point to the side of the road and pull back, and it will pull over to the side of the road. Push the stick forward, and it will wait for a lull in the traffic, then pull back out. I think with our current infrastructure we need that level of manual control - there's no way I'm surfing a database to get the coordinates for the loading dock at Lowes, or messing with google earth to figure out where I should park to put my kayak in the river.


ETA: those kinds of limited control schemes can lead to massive instability, however, as we humans tend to overcorrect when a system does not respond predicatably to our inputs. I'm dealing with that with a system we are building at work. It's not an impossible problem, but the engineering challenges are not trivial, either.

I think that a combined system will be the first. Perhaps still with a steering wheel but with the option of "engaging autopilot" on select roads. It could perhaps start with simple systems like "autopark"(is in use) auto distance/anti collission (is being roadtested) and similar. I think we will se a gradual move to a automatic car during the next 10 years or so.

ohh and roger: A lot of GPS systems can find vacant parking spaces. Copenhagen City has a website that you can log into with your GPS which guides you to the nearest place where you can park.

 

[ponderingturtle]

But flawed. Why design a system that is no better than what we can do now when it is easy to make it better? Just because cars could be made to recognise signs does not mean it is the best solution, especially since you have admitted that they would suffer from all the same problems humans do. Low power radio transmitters instead of signs would be far more reliable and much easier to update. Getting the information into the car is trivial and the technology could easily be rolled out tomorrow. It is making cars capable of acting on it that is the tricky part.

So driving down a street marked as one way by sign if it is marked as not one way on the database is what you want?

 

[ponderingturtle]

You make a good point that in the long term signs may not be the most efficient way to communicate road information to self-driving cars. But they make sense for the mixed use period where human drivers and self-driving cars will be sharing roads. During that period, it's a good idea to have road signs. And ponderturtle's recurring point is that given the visual tasks we'll have to give self-driving cars to avoid accidents on the road from things like foreign objects, it's extremely easy to add in the capability to read speed signs placed there for the human drivers, so we might as well add in that technology.

Once there are no more human drivers on public roadways, I agree with you that the incentive to maintain driving-related signage and the lines of software code for cars to read them may disappear.

You will always have signs, but they might be the temperary sort, to note current differences in the actual roads from their default state. Having a physical sign as a back up to electronicly updateing information seems obvious at least to me.

 

[Cuddles]

So driving down a street marked as one way by sign if it is marked as not one way on the database is what you want?

I have never mentioned a database. A database is a stupid idea. If you read what I actually said, you will find I mentioned short range radio transmitters. These essentially act as signposts, but are much more reliable since they cannot be covered up or obscured, they use less power, are much less prone to vandalism and far easier to update, locally or remotely. There is no more chance of having an electronic sign wrong than there is of having normal signs wrong. The only drawback is that humans can't read them, but since they would be put in exactly the same place as regular signs there is no reason why both systems can't exist together during the changeover.