Why autonomous cars won't be able to drive properly...

Anything that doesn't fit elsewhere - doesn't have to be AD related.
User avatar
Mr Cholmondeley-Warner
Posts: 1118
Joined: Sun Sep 27, 2015 7:01 pm
Location: Swindon

Re: Why autonomous cars won't be able to drive properly...

Postby Mr Cholmondeley-Warner » Sun Jan 21, 2018 3:23 pm

Horse wrote:
jont- wrote:Well, quite. Or at least the average member of the public. My problem with a lot of them is that they are basing ML decisions on data acquired from lots of very average drivers. I want my autonomous car to drive like Andy Morrison.


If it's possible to write that style into rules, then it may well be possible - in time - to achieve that.

Not rules so much as experience. Imagine Andy's thought process on seeing an oncoming emergency vehicle using the "parting the waves" technique (or any other scenario you care to summon up). "I've seen this situation several thousand times. Mostly what happens is x. Other times y or z or a or b or c or d or ... and then there was that one time (oops no twice) when x[23] happened". By the time that sentence is finished, he's also assessed the drivers around him, how they're likely to behave, their age, apparent confidence levels, the vehicles they're driving, the road topology, the state of the verges, presence or not of kerbs, escape routes, has a main plan and two backups, and is putting plan A (ease off with half an eye on the mirror and position far enough left to make space) into operation, ready to adopt plan B or C (or X23) as the situation develops.

The autonomous vehicle will do the physical scan, and assess space, velocity etc, and implement its plan. There will be only one plan, and that won't change until a safety threshold is breached, when it will repeat the process. On implementing its plan, will it take into account its effect on surrounding vehicles? (the guy behind is too close, so I'll apply the brakes rather more gradually than I otherwise might...). And so on ...
Nick

Pyrolol
Posts: 73
Joined: Sun Oct 04, 2015 1:52 am
Location: San Francisco

Re: Why autonomous cars won't be able to drive properly...

Postby Pyrolol » Sun Jan 21, 2018 3:43 pm

I don't really see experience derived prediction being a weakness of an autonomous car. Few people will drive more than a million miles in their lifetime - autonomous vehicles could potentially do that in a day. The quantity of experience advantage is firmly on their side.

I don't have a great understanding of how modern machine learning works, but I don't think it's a single plan. I believe it tends to be a very large number of plans evaluated against heuristics for goodness. It may only execute one, but a car can only go one way.

User avatar
Mr Cholmondeley-Warner
Posts: 1118
Joined: Sun Sep 27, 2015 7:01 pm
Location: Swindon

Re: Why autonomous cars won't be able to drive properly...

Postby Mr Cholmondeley-Warner » Sun Jan 21, 2018 3:55 pm

So will they know that, for example, a Toyota Yaris is a vehicle beloved of ladies of advancing years (except when it isn't), look at the one in front of the oncoming ambulance, check if it's driven by such a driver, and factor that into the likely behaviour of that vehicle? Does it matter, anyway? Possibly not, but it's something that can't be written as a hard and fast rule.
Nick

User avatar
jont-
Posts: 1522
Joined: Mon Sep 28, 2015 7:12 am
Location: Herefordshire

Re: Why autonomous cars won't be able to drive properly...

Postby jont- » Sun Jan 21, 2018 4:47 pm

Pyrolol wrote:I believe it tends to be a very large number of plans evaluated against heuristics for goodness.

Indeed. Except no-one understands how those heuristics work, or that they can be easily (dangerously?) confused
https://www.theregister.co.uk/2017/10/3 ... gle_pixel/

User avatar
exportmanuk
Posts: 351
Joined: Thu Oct 15, 2015 6:56 pm
Location: Manchester

Re: Why autonomous cars won't be able to drive properly...

Postby exportmanuk » Sun Jan 21, 2018 6:16 pm

jont- wrote: Except no-one understands how those heuristics work, or that they can be easily (dangerously?) confused
https://www.theregister.co.uk/2017/10/3 ... gle_pixel/


So a sudden spec of dirt on a camera could have significant effects ? The more I read about these vehicles the scarier they seem. That added to the belief by many that these vehicle will bring about traffic utopia.
Andrew Melton
Manchester 500

User avatar
Horse
Posts: 3558
Joined: Mon Sep 28, 2015 9:20 am

Re: Why autonomous cars won't be able to drive properly...

Postby Horse » Sun Jan 21, 2018 8:00 pm

I think I've posted the YT link before of US fire service driver training using simulations, where scenarios are based on real incidents.

In a way it could be argued that 'higher/better's AD is really better prediction. Of course, it could be argued that a computer system might be better at consistent 'observation' and prediction . . .
Your 'standard' is how you drive alone, not how you drive during a test.

User avatar
Horse
Posts: 3558
Joined: Mon Sep 28, 2015 9:20 am

Re: Why autonomous cars won't be able to drive properly...

Postby Horse » Sun Jan 21, 2018 8:07 pm

exportmanuk wrote:
jont- wrote: Except no-one understands how those heuristics work, or that they can be easily (dangerously?) confused
https://www.theregister.co.uk/2017/10/3 ... gle_pixel/


So a sudden spec of dirt on a camera could have significant effects ? The more I read about these vehicles the scarier they seem. That added to the belief by many that these vehicle will bring about traffic utopia.


Yes quite worrying. Does your car have A B or C pillars, does the windscreen ever get dirty outside or in, etc?

In some of those cases the system confused dog and deer or cat. Perhaps a vision system will be cross-checked to IR and radar?

Did you read the article I posted a download link, of the police driver who saw a sofa and reported seeing an HGV? Perhaps these vision systems are actually realistic! ;)
Your 'standard' is how you drive alone, not how you drive during a test.

User avatar
Horse
Posts: 3558
Joined: Mon Sep 28, 2015 9:20 am

Re: Why autonomous cars won't be able to drive properly...

Postby Horse » Mon Jan 22, 2018 9:36 am

sussex2 wrote:I'm assuming, rather hoping, there will be an OFF button somewhere :)


Next you'll be insisting that you want to see a cashier on a checkout in the supermarket ;)

http://www.bbc.co.uk/news/business-42769096
Your 'standard' is how you drive alone, not how you drive during a test.

User avatar
Horse
Posts: 3558
Joined: Mon Sep 28, 2015 9:20 am

Re: Why autonomous cars won't be able to drive properly...

Postby Horse » Mon Jan 22, 2018 9:42 am

Horse wrote:
exportmanuk wrote:So a sudden spec of dirt on a camera could have significant effects ? The more I read about these vehicles the scarier they seem. That added to the belief by many that these vehicle will bring about traffic utopia.

Did you read the article I posted a download link, of the police driver who saw a sofa and reported seeing an HGV? Perhaps these vision systems are actually realistic! ;)


Here:
Horse wrote:
Mr Cholmondeley-Warner wrote: * we = advanced drivers / riders. I hope we look a little bit more carefully than average Joe

You'd like to think so . . .
https://www.dropbox.com/s/euud3txjhsvs1 ... v.tif?dl=0
Your 'standard' is how you drive alone, not how you drive during a test.

User avatar
dvenman
Posts: 261
Joined: Sun Oct 18, 2015 5:28 am

Re: Why autonomous cars won't be able to drive properly...

Postby dvenman » Tue Jan 23, 2018 8:31 am

I think this chap is hoping that fully autonomous cars make it to the market sooner rather than later...

https://arstechnica.com/cars/2018/01/tesla-owner-attempts-autopilot-defense-during-dui-stop/


Return to “General Chat”

Who is online

Users browsing this forum: No registered users and 31 guests