More driverless cars

[AdamSmith]

Interesting piece on not Google's driverless-car project but a like effort by Volvo.

http://www.theguardian.com/technology/2014/jun/07/driverless-volvo-s60-car-review

[RA1]

Driverless cars in a way sound scary, somewhat like a drone. Both can operate independently but both need a "minder". OTOH, going 75 MPH 2 feet away from who knows who, meaning mental state, physical state, high or low, is not exactly a wonderful feeling.

Pilots have to face re-current training, re-current medical exams and various amounts of initial training in order to fly. Drivers have something almost a rite of passage with a minimum of training, no medical exam and no re-current training or evaluations. In other words a fee paid to the state that allows one to drive. Scary.

I have to say I am not in favor of drones in their current state of development (meaning regulations) but I am in favor of developing driverless cars. One caveat, how shall we survive when the whole USA is covered in concrete or asphalt to accommodate vehicles and there is no land left for crops, parks, living space, etc.?

Best regards,

RA1

[MsGuy]

One caveat, how shall we survive when the whole USA is covered in concrete or asphalt to accommodate vehicles and there is no land left for crops, parks, living space, etc.?

Actually, the general idea is to eliminate our vast fleets of individually owned cars.

You just click your i-phone for transport when you need a car, the nearest vehicle drives itself to wherever you are and is waiting for you at the curb by them time you get to the street. You get in, tell it where you want to be and off you go. Way more efficient both in terms of cost to you and in terms of the total number of vehicles on the road.

These guys aren't really talking about driverless cars as an end in themselves. The real payoff is in a new paradigm for personal transportation. All the efficiency of public transport w/o the hassle and inconvenience; all the convenience of personal vehicles w/o the cost and aggravation.

[RA1]

Aren't these called taxis?

Best regards,

RA1

[AdamSmith]

Taxis only seem driverless.

[Guest PasadenaCA]

Taxis only seem driverless.

Nail on the head!

[AdamSmith]

One more point: all the key people in the driverless-car development initiatives, at Google and Volvo and some others (I happen, ahem, to know just about all of them from one line of the consulting I do for work, or what passes for work ) make the point that once the sensors and even more the algorithms are sufficiently developed and matured, autonomous vehicles will be on average vastly safer than human drivers. They all rhapsodize, quite rightly I think, about this technology slashing the appalling rate of traffic fatalities and injuries today.

We can of course make endless cracks about how if Microsoft can't even get laptops to operate right, etc. But the better comparison is possibly aircraft autopilot software, or the stability and braking control and all-wheel-drive software in cars which, one or two unintended-acceleration episodes excepted , works extremely reliably. Today's luxury car has like 10 million lines of code in it, almost 10 times the amount of software in a modern commercial aircraft.

So in sum, I trust the autonomous-vehicle guys to get it right. And to make sound judgments about when they have in fact done so.

[Guest PasadenaCA]

It's one of these goals, that if you connect the dots, you can see it coming. One key thing that has to happen is that cars will have to interact with each other--something like a very sophisticated version of TCAS. This interaction will lessen the programming burden, trying to anticipate what could happen.

[AdamSmith]

What an insightful but funny conclusion. In fact Chris Urmson and Sebastian Thrun, now with Google, and also the 2 Volvo guys quoted above, and other geniuses (whom, again, one knows personally ) stay awake at night dreading specifically the unforeseen reactions of their robots with one another, as their potentially worst engineering oversights. All the other problems are much more tractable, they think.

[Guest PasadenaCA]

What an insightful but funny conclusion. In fact Chris Urmson and Sebastian Thrun, now with Google, and also the 2 Volvo guys quoted above, and other geniuses (whom, again, one knows personally ) stay awake at night dreading specifically the unforeseen reactions of their robots with one another, as their potentially worst engineering oversights. All the other problems are much more tractable, they think.

Driverless cars need the awareness--of some sort--that we have. When in doubt, stop.

[RA1]

RE: The autopilot comparison.

One night a friend of mine who is now a retired FEDEX DC-10 captain, was flying a coupled autoland approach. This involves 3 certified autopilots + special monitoring by the crew. The FO looks for the runway and the captain watches the AP and the FO. My friend reported to me the next day that one by one the autopilots dropped off line. The aircraft landed itself and used autobrakes to slow. It was only when the captain disengaged the autobrakes and started his taxi that anyone realized all autopilots had failed.

Eventually, driverless cars likely will be safer than one driven by human drivers but not because driverless cars are so good but because human drivers are so bad.

Best regards,

RA1

[AdamSmith]

This Stone Age kid needs education: FO?

[AdamSmith]

PPS: The job of engineers:

(1) To make for one cent that which anybody could make for two cents.

(2) To do that while also discovering and mitigating all the ways in which their work could kill many people.

A somewhat daunting job description, come to think.

[RA1]

You are far from a Stone Age kid but we could have had fun back then. I always wanted to BBQ a Mastodon.

FO = First Officer, now commonly called SIC or Second in Command, previously known as co-pilot.

Years ago, it was, "don't touch anything unless I tell you to" which slowly became CRM or Cockpit Resource Management, which is usually defined as using all the resources available and don't leave anyone out.

Best regards,

RA1

[RA1]

Daunting is my watchword.

Best regards,

RA1

[AdamSmith]

Driverless cars need the awareness--of some sort--that we have. When in doubt, stop.

If you can give an algorithmic, quantifiable definition of 'doubt,' these guys will pay you millions.

[AdamSmith]

Seriously: that, as they describe it to me, is their main challenge. You can remove all risk if you simply prohibit the vehicle from proceeding at all.

But how to balance the introduction of some risk in order to let the thing proceed down the road, reason at intersections, etc.? Calculating the risk in diving at 10mph vs a reasonable road rate is the central challenge of autonomous vehicles.

[MsGuy]

5,642 vehicle pile up on quiet residential street:

Engineers blame faulty algorithm.

[Guest PasadenaCA]

How would collision insurance work for a driverless car?

[JKane]

How would collision insurance work for a driverless car?

The Google car is a weird exception and more of a golf cart, but the general idea is that it is your responsibility to supervise the car and be prepared to take over if necessary. Wired just had a great article about how the car could tell you when it was uncertain about something, maybe an odd shadow or bump that could be a pet or something...

The simple fact is there will be almost no collisions (front end anyway, the cars will still get hit...) and what there are will be people who took over or were self-driving and then tried to blame the automation. But (for liability reasons for the manufacturers) everything will have been recorded in detail.

But the fact that within the next couple years automation will clearly be better (from a safety and efficiency standpoint) will be ignored by many and the clothes rending panic at every incident involving any car with any automation will grow to a roar despite the numbers of actual incidents and severity of injury being *far* in favor of the automation. "Tesla Fires" X10.

Because people fear change and existing businesses exploit that.

[RA1]

Similarly, the way to prevent aircraft injuries and deaths is to ground all aircraft. No flying, no problems. Except no commerce or fun ensues.

The aircraft industry (general aviation) limited their liability by excluding products produced many years ago. Prohibiting ambulance chasers would be a big help. In other words, having driverless cars perform for an extended period of time successfully would go a long way towards limiting liability and legislation for same could not hurt.

I do not think that having occupants ready and willing to take command of a driverless car would be a long term or satisfactory solution. With cruise control, automatic braking and other tech features of modern cars, is that not approximately what we have now?

Best regards,

RA1

[Guest PasadenaCA]

Similarly, the way to prevent aircraft injuries and deaths is to ground all aircraft. No flying, no problems. Except no commerce or fun ensues.

The aircraft industry (general aviation) limited their liability by excluding products produced many years ago. Prohibiting ambulance chasers would be a big help. In other words, having driverless cars perform for an extended period of time successfully would go a long way towards limiting liability and legislation for same could not hurt.

I do not think that having occupants ready and willing to take command of a driverless car would be a long term or satisfactory solution. With cruise control, automatic braking and other tech features of modern cars, is that not approximately what we have now?

Best regards,

RA1

Basic skills in driving are bound to get rusty, so you may end up with people both driving for the first time and trying to avoid an accident. I'd agree that there are likely to be far fewer accidents, but if and when one were to occur, chances are that an inexperienced driver would add to the problem.

Drawing on an airline analogy again, I think that the investigation into the Asiana crash at SFO last summer will show that the pilot in charge could supervise the equipment but lacked the required experience to actually fly it. His skills were rusty from lack of use.

[RA1]

There was far more to the Asiana crash than that. No one noticed that the auto throttles were not engaged. One pilot did notice the slowing airspeed but too late. There were 3 qualified pilots in the cockpit watching or participating in the approach. There was a "culture" problem in that many other than US (and some US) pilots have a strict adherence to authority. The captain is the captain, etc.

I think what you have in mind is that this airline trained their pilots to rely too much on electronic approach aids. The runway used by this Asiana flight had the ILS inop for maintenance. The weather was good, the crew had the airport in sight and, as is very normal, they were cleared for the visual approach. Even so, they were used to relying upon electronic guidance to the runway but it was not available that day. Still, it was a simple procedure except they were concentrating on doing the visual without aid and did not notice the auto throttles not engaged. The plane got slower and slower until it was unable to fly any longer and hit the sea wall short of the runway and then continued onto the runway. Unfortunately there was loss of life. The pilots weren't rusty, they were insufficiently trained.

Best regards,

RA1

[AdamSmith]

Interesting, though hard-to-read, diagram of the different kinds of sensors used on autonomously guided vehicles:

From this article on a new prototype Mercedes self-driving freight truck: http://www.cnn.com/2014/07/04/tech/mercedes-future-truck/index.html?hpt=hp_bn5

[Suckrates]

Well, if I dont need to drive it, then I probably shouldnt PAY for it either ? they should just GIVE them to us GRATIS...... and PAY for the insurance as well....?

That certainly would be progress at its Best !