Saturday, 23 August 2014

I, for one, welcome our new Google overlords


more ...

Google's driverless cars have clocked up 700,000 miles around California by late April 2014 - with only two accidents : one occurring in 'manual' mode , one was rear-ended while stopped at lights.
I guess that puts the algorithm in the top 5% of good drivers, but doubt if accurate statistics for humans exist!



It seems ministers will review the Highway Code to allow Google self-driving cars in the UK next year.
But that may not mean a revised Highway Code : I don't see any reason why the algorithm can't obey the old code better than humans.

A Google software engineer said Google's self-driving cars are programmed to exceed speed limits by up to 10mph, if surrounding traffic makes that safer. Hmmm ...

Google wants to prove its algorithms further in a Virtual Reality simulator.
That VR platform could be used to improve the 'human algorithms' of the Highway Code, and experiment with different road infrastructure scenarios, such as the Bow Roundabout, which seems to be causing problems.

Initial concept - 28 Mar 2012

Launch animation - 22 Aug 2013


Note these are mainly graphics, not behaviour simulators.

In practice, it is not so much an 'early start' as a 'double stop' - not at all what cyclists were led to expect!
Is this just 'poor management of expectations', or deliberate deception by TfL ?
NB The Times even refers to them as 'cyclist priority' lights !





What many commentators miss is that the cars will almost certainly be networked to each other, and to static servers. Thus they can produce live traffic reports, know what's around corners, report temporary obstacles like roadworks etc. Even cyclists.

How driverless cars will change the way the British moan about traffic - [ flexed.co.uk ]
"Driverless cars are built on logic and the passionless scanning of the road ahead by a computer," says Flexed.co.uk 's Mark Hall. "It's going to be all function and no discretion, with no nipping out of road junctions before that chap towing a caravan doing 25 mph. Speaking as a British driver, can you imagine the boredom?"
1. Sitting and watching other drivers watching films, playing video games or having fun, when you’ve forgotten to bring something interesting to pass the time.
2. Mini roundabouts. We all know what happens when three cars get to a mini roundabout all wanting to turn right – British politeness sets in and you could be there until tea-time waiting for someone to go first. Robot cars just won’t be able to make that decision, it’ll be chaos, and Isaac Asimov will be turning in his grave.
3. Car crime. Who wants to steal a car if it drives you straight to the police station? Better still, it could driver criminals to the local water treatment works and slowly fill the car up with sewage until the criminals apologise.
4. The art of hitch-hiking is likely to die out, unless robots take up hitch-hiking.
5. Cars occasionally turning up at destinations with heart-attack victims inside, which won’t be a pleasant experience for anybody
6. Road rage perpetrated on your own car because it slavishly follows its programming and refuses to overtake the car in front that’s tootling along at 2mph below the speed limit.
7. Banging your head on the dashboard in frustration as your car reacts too slowly to get the last space in a car park
8. Not a problem: Other drivers won’t be able to nip up the ‘wrong lane’ in a traffic jam, and jump into a gap near the front of the queue. What is a problem: Neither will you.
9. Higher insurance rates for your make of robot car, because its driving logic is worse than a more expensive model
10. Pranksters placing cardboard cut-outs of people on crossings.
11. People ‘hacking’ their cars to let it break the law.
12. The first conviction for a couple having sex in a driverless car.

For the origin of the title, "I, for one, welcome our new insect overlords", see knowyourmeme.com.

But they are no substitute for infrastructure

Mark Treasure on Twitter: "This only makes sense if ‘bad driving’ is the sole reason for building cycling infrastructure. It isn’t."
For one thing, I don’t want to cycle on roads clogged with motor vehicles, even if they are perfectly driven. I want space and priority.  Equally, cycling in front of driverless HGVs and buses is far from attractive, either for the person cycling, or the vehicle occupants
Also kids do not always cycle straight and can be unpredictable.
Driverless cars won't make a traffic jam any more pleasant to cycle past but separated space will.
Heavy vehicles tear up roads. Damage is mainly proportional to sum of cubes of wheel loads. bus damage 300000x bike.
Driverless cars and the sacred cow problem - [ John Adams ]  - full PDF
The promoters of driverless cars have demonstrated remarkable progress in their ability to program their vehicles to respond with extreme deference to pedestrians, cyclists, and cars with human drivers. Such programming confers sacred cow status on all road users not in self-driving vehicles. The developers of autonomous vehicles acknowledge the need for new road safety rules to accommodate these revolutionary vehicles on public highways. But would-be regulators have yet to propose a set of rules that would allow these sacred cows to move about freely in dense urban areas without creating a state of deferential paralysis for those in autonomous vehicles.

See also