Urban75 Home About Offline BrixtonBuzz Contact

autonomous cars - the future of motoring is driverless

weltweit said:
Will the computer system be able to detect when a tyre is down on pressure like a driver can, will the computer be able to .. sorry I have to go, will be back later.
Click to expand...
This is the trouble with your argument though - yes it can, long before the driver notices unusual handling. TPMS has been around for a decade or so.

Edit: I pay attention when driving, and I drive a car that's fairly communicative. I noticed when cornering a while back that something was amiss. The tyre was pretty much flat (after changing to/from winters that day - didn't seal properly). TPMS would have told me well before that.
 
Last edited:
weltweit said:
Oh it is programmed, do you think some kind of empty computer just starts to take inputs and learn
Click to expand...
This is exactly how it works these days. Machine learning algorithms are now generic, and they're going to take over everything. It's exploding in popularity at the minute because now it's so easy.

See this video (I've linked to the relevant bit - it's only a few mins) to see a humvee learn to steer in 2 minutes:

This was a long time ago. Imagine what's possible now?

They don't even need to be programmed to communicate with each other to learn, btw, as some people are suggesting on this thread (though that would be useful!). Once they have been trained, they will work better than any human without additional training. You'll probably get software updates on your car the same way you do on your computer now. That will be based on the collective data of the fleet or by tests done by the manufacturer.
 
Last edited:
weltweit said:
"Whilst stuff about sensors and how cars react might be optimistic,"
This is what I think you are optimistic about, well that and machine learning being as easy to do as you and others seem to think, and the whole project being put into practice in the time frames mentioned by some in this thread.
Click to expand...

ah, well I think we've explored our differences on that completely and the post you responded to wasn't about that hence my confusion. I don't know how easy or difficult machine learning is to do but even if it's difficult I think Google, Apple, Uber and the might of the money of the car industry will manage it tbh. I have no concept of time frames because that's a fools game. It will happen at some point imo and I welcome it, having seen the footage of how the google cars drive around cyclists.
 
Fez909 said:
This is exactly how it works these days. Machine learning algorithms are now generic, and they're going to take over everything. It's exploding in popularity at the minute because now it's so easy.

See this video (I've linked to the relevant bit - it's only a few mins) to see a humvee learn to steer in 2 minutes:

This was a long time ago. Imagine what's possible now?
Click to expand...

In that segment the HumVee was taught much in the way a spraying robot is taught, by a human doing it and then copying with a little extrapolation. But the HumVee visual system is reduced in detail massively, the image it is working on does not see that there are leaves on the road, or the camber of the road, things an experienced human driver would take into account. And when the Humvee reaches the cross roads it just drives out onto the two lane road without giving way.

Yes it is a long time ago and yes things will have improved by now. But there are issues.

It reminds me of the convoy driving proposals. So cars can queue up one behind another and follow like a train down a motorway. Great, except for if the lead car makes a mistake like the Tesla did when driving into a truck, then, for the following cars, what then?
 
mauvais said:
This is the trouble with your argument though - yes it can, long before the driver notices unusual handling. TPMS has been around for a decade or so.

Edit: I pay attention when driving, and I drive a car that's fairly communicative. I noticed when cornering a while back that something was amiss. The tyre was pretty much flat (after changing to/from winters that day - didn't seal properly). TPMS would have told me well before that.
Click to expand...
TPMS is one of those nice to have options for more expensive cars, do you expect it to be available in all automated cars? if so what else? if these vehicles include every sensor system known to man in addition to their driving computers/systems, they will end up being prohibitively expensive.
 
weltweit said:
TPMS is one of those nice to have options for more expensive cars, do you expect it to be available in all automated cars? if so what else? if these vehicles include every sensor system known to man in addition to their driving computers/systems, they will end up being prohibitively expensive.
Click to expand...
It's available in middling Renaults and no doubt plenty besides, so yeah. And yes, they will contain all kinds of sensors, because they're self driving cars. Again you're looking at it wrong - mapping what we have today directly into this scenario without any flexibility. Think cars supplied as a service, think third party services for machine learning etc, think sensor suites and packages as a commodity.
 
weltweit said:
In that segment the HumVee was taught much in the way a spraying robot is taught, by a human doing it and then copying with a little extrapolation. But the HumVee visual system is reduced in detail massively, the image it is working on does not see that there are leaves on the road, or the camber of the road, things an experienced human driver would take into account. And when the Humvee reaches the cross roads it just drives out onto the two lane road without giving way.

Yes it is a long time ago and yes things will have improved by now. But there are issues.

It reminds me of the convoy driving proposals. So cars can queue up one behind another and follow like a train down a motorway. Great, except for if the lead car makes a mistake like the Tesla did when driving into a truck, then, for the following cars, what then?
Click to expand...
It was a steering algorithm only. The human was in charge of everything else. I hope you pay more attention when you're driving, than when you're discussing driving ;)

And yes, of course they've moved on a lot now. We've already got driverless cars on the roads in America that have proven themselves safer. We don't need to be hypothetical about it, we can look at the stats: https://static.googleusercontent.co.../selfdrivingcar/files/reports/report-0816.pdf

In August, Google cars did 170,000 miles. For 126,000 miles, the car was fully autonomous.

They had 5 crashes. 3 in manual mode, 2 in autonomous mode. Neither of the autonomous crashes were the fault of the car.

170,000 miles is the equivalent of 10 years of driving.

In July, one crash, also the other driver's fault: https://static.googleusercontent.co.../selfdrivingcar/files/reports/report-0716.pdf

June, two crashes, neither one the car's fault: https://static.googleusercontent.co.../selfdrivingcar/files/reports/report-0616.pdf

May, one crash. The crash was caused by the Google car. It was in human mode: https://static.googleusercontent.co.../selfdrivingcar/files/reports/report-0516.pdf

So there's about 40 years worth of driving miles. 9 crashes. None of them the fault of the computer.
 
BigTom said:
Different companies will have their own systems that probably won't talk to each other, but they will each have many thousands of vehicles on the roads, all learning together. Every company will very quickly have a data set way larger (and more accurate/reliable) than the one you've accumulated in 35years.
Click to expand...

I suspect that the major companies will play smart, and go for unitary standards, as - at least to start with - there would be little advantage to individual proprietary systems, but great advantage - pooled costs and risks, for a start - to a unitary system standard.
 
weltweit said:
I am technologically pessimistic, yes, but partly because I have seen how slow the automotive industry is to innovate with inventions that have a lot of testing behind them. The automotive industry is, or has been, risk averse, and the supplying components industry features massive testing regimes for even the simplest component. Then we have the concept that above all this we are going to add a computer system to drive the car collaborating with HAL9000 at base. Will the computer system be able to detect when a tyre is down on pressure like a driver can, will the computer be able to .. sorry I have to go, will be back later.
Click to expand...

Onboard systems can already accurately detect tyre pressure anomalies.
 
This project takes the self driving car to new heights, literally.


Urban Aeronautics have been testing their car sized Cormorant UAV which can carry a 500kg payload up to 50 km at speeds up to 185 km/h and altitudes up to 5500 m (18kft). Flying either manually via remote control or autonomously, it is driven by internal rotors and ducted fans housed in shields to prevent damage. The craft can take off and land vertically darting around in various directions once airborne. This model is intended to operate rescue or supply missions to vessels at sea, in remote or difficult to reach locations (obvious military applications too).

The firm are rumoured to be working on a smaller variant that could be used by the public as a flying car.
dronebigenou.jpg
 
More impressive city driving. Handling 4-way stops, pedestrians, cyclist undertaking, construction works, double-parked obstruction and other irregularities.

 
Ford talking about self driving cars

A report by Bloomberg says Ford is going to skip a step and go straight to fully autonomous driving. The article says that is because engineers who are testing the company's self-driving vehicles are falling asleep at the wheel because there is so little for them to do.

Ford tells me that only part of this story is true: "Reports that Ford engineers were falling asleep while testing autonomous vehicles are inaccurate."

But it goes on to say that "high levels of automation without full autonomy capability could provide a false sense of security".

That means it is difficult for the driver to suddenly take control if there is a situation where the technology is not up to it. And that's why it is going to head straight to what is known as SAE level four - "autonomous capability that will take the driver completely out of the driving process in defined areas".
Click to expand...

Driverless cars - no halfway house? - BBC News
 
Funnily enough, I've just finished watching Jay Leno (old and new) and he drove a Tesla. Sadly, did not go into much detail.
 
Crispy said:
More impressive city driving. Handling 4-way stops, pedestrians, cyclist undertaking, construction works, double-parked obstruction and other irregularities.

Click to expand...


How does it handle stop lines? Mapped or response to environment? In fact there a quite a few things that, intuitively, seem difficult to respond to - I mean a traffic light is just another light source and can be positioned in a number of different places within a (human) driver's field of view. A moving cyclist is at least a moving thing, and construction works are abnormal features in a road - to me it's perhaps more impressive that they respond to some dirty white lines and a hexagonal sign.
 
Cid said:
How does it handle stop lines? Mapped or response to environment? In fact there a quite a few things that, intuitively, seem difficult to respond to - I mean a traffic light is just another light source and can be positioned in a number of different places within a (human) driver's field of view. A moving cyclist is at least a moving thing, and construction works are abnormal features in a road - to me it's perhaps more impressive that they respond to some dirty white lines and a hexagonal sign.
Click to expand...

Traffic lights are very specific light sources though, I doubt there's many if any other light sources with Red, Amber, Green all the same size stacked on top of each other, especially once you take into account the phasing of the lights and that often the vehicle will have time to see the lights change and I really doubt anything else in nature or our urban landscape would mimic that (except by intent of course). I agree about the dirty white lines and hexagonal signs though and do wonder how they will deal with low contrast when paint is really dirty or under puddles.
 
You must log in or register to reply here.

Similar threads

A
First car / how do people afford cars when in low paid work?
2
Replies
59
Views
1K
Chz
Chz
danny la rouge
Traffic news apps for travelling in cars
2 3
Replies
73
Views
2K
eightball
E
H
Met police cars from 2022
Replies
6
Views
388
hash tag
hash tag
I
Motorcycles you have owned, loved and used down the years.
Replies
4
Views
145
Nivag
Nivag
ska invita
Car diagnostic computer scanners
Replies
8
Views
351
mauvais
mauvais
Back
Top Bottom