REGISTER | LOGIN
Breaking News
News & Analysis

Biggest Robo-Car Roadblock Is Human

CES 2017 Automotive Recap
1/13/2017 09:30 AM EST
11 comments
NO RATINGS
Page 1 / 7 Next >
More Related Links
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
IJD
User Rank
Author
Re: Take it from my cold dead hands.
IJD   1/17/2017 1:08:59 PM
NO RATINGS
rohs_compliant, never is a very long time. Let me turn it round -- how much per year will you be willing to pay for insurance to not relinquish that control?

94% of accidents are caused by driver error, so any decent AI should be able to do better than that and -- unlike human drivers -- improve year-by-year. So once AI cars hit the road in increasing numbers, more and more accidents will be caused by a smaller and smaller number of human drivers, and the insurance companies will ensure that the culprits pay by escalating premiums. Once 90% of accidents are caused by the remaining 10% (and falling...) of human drivers, the cost of driving yourself will rise so rapidly that fewer people will do it every year, which will put the insurance costs up for those left, which will cut the number of human drivers...

So, would you pay 5k/year insurance? How about 10k? 20k?

Before saying "don't be ridiculous, the cost won't go up that much" -- by then, a multi-car pileup on a motorway which kills or cripples several people with total insurance costs of millions will all be dumped on the one unfortunate human driver involved.

Maybe human driving will be seen as so dangerous that society will view it like drink-driving today -- actually it ought to be seen as much worse, most accidents are caused by sober people...

apaDAV
User Rank
Author
Autonomous Ethics
apaDAV   1/17/2017 12:57:21 PM
NO RATINGS
So basically it is now up to engineers to solve the ethics problem that no philosopher dared to touch for 3000 years.
Now engineers must answer the question of the value of a specific human life. Is a tall person more valuable than a small person?
Better to risk the health of a train full of people or the life of one?
Is the owner of the car "worth" more than a pedestrian?

Do engineers have to be responsible for ALL of human progress or will the soft sciences finally contribute?

rohs compliant
User Rank
Author
Take it from my cold dead hands.
rohs compliant   1/17/2017 9:10:56 AM
NO RATINGS
As long as I have the mental and physical capacity to drive my own vehicle, I will never relinquish that control to an autonomous vehicle.

Victor.Lew
User Rank
Rookie
Validating an inductive learning system
Victor.Lew   1/15/2017 8:11:28 PM
NO RATINGS
https://users.ece.cmu.edu/~koopman/pubs/koopman16_sae_autonomous_validation.pdf

https://users.ece.cmu.edu/~koopman/pubs.html

https://betterembsw.blogspot.com/

Art_
User Rank
Author
Missing Element in Most Discussions on this Topic
Art_   1/13/2017 7:47:35 PM
NO RATINGS
Drivers base their actions based on certain assumptions about the other drivers around them. As Telsa has seen, a significant number of drivers interacting with automated systems (either in their own car or another vehicle) push the limits since the systems are not humans. I have not seen a clear plan for dealing with this natural tendency in discussions of mixed traffic of automated and human drivers. For example, how would an automated car ever make it into a crowded roundabout? Or merge during a traffic jam? A measured amount of human aggression keeps things flowing now. If a driver knows that another car is automated, would they tend to not let them merge? Or cut them off intentionally knowing that the automated car will avoid a collision? In a few cases, I would expect extreme behavior specifically testing the automation.

Taking this to the limit, should we build in some apparent risk for other drivers to ensure they behave appropriately around automated cars? i.e. zero-tolerance is counter-productive?

Rather than adding physical risk to add negative feedback to the system, we could consider a debit system based on driving behavior. Instead of relying on police monitoring and issuing citations, the automated car and ADAS sensor systems could flag hazardous behavior and issue complaints (with documentation) that would lead to automated fines :) The same decision methods driving the automated car's decisions would be applied to surrounding cars. This could lead to better behavior for all drivers!

junko.yoshida
User Rank
Author
Re: ASIL-D and AI should have been the heart of the
junko.yoshida   1/13/2017 4:09:38 PM
NO RATINGS
Victor, I don't have the answers to your important question.

That should be a separate story that needs a whole lot more digging than what I put tother here.

Don't get me wrong, I know what you are getting at.

In my defense, though, this was meant to be a story about CES automotive recap.

Thanks for your interest and we will work on it.

EELoser
User Rank
Freelancer
Re: ASIL-D and AI should have been the heart of the
EELoser   1/13/2017 3:33:43 PM
NO RATINGS
Do key ignition switches or airbag inflators,  pass ASIL-D standards?  How about Firestone tires that shred apart and kill people?

 

 

EELoser
User Rank
Freelancer
Can't even design a Ignition Switch or Airbag right.
EELoser   1/13/2017 3:31:05 PM
NO RATINGS
Will the first mistake put these companies out of business, like GM almost was before they went bankrupt and Takata probably will be soon.

Seems like ignition switches and airbag inflators are kind of easier to develop than automated cars and they can't even do those right.

I also don't want to have a fender bender and have 5000 dollars worth of damage because the sensors are knocked out, or the replacing the main computer board for thousands of dollars when my car is out of warranty. 

 

 

Victor.Lew
User Rank
Rookie
ASIL-D and AI should have been the heart of the
Victor.Lew   1/13/2017 2:13:55 PM
NO RATINGS
Can AI ever achieve ASIL-D status in next 10 years should have been the heart of the focus of the article.  In your reflection, which outlines a building block approach, the response contradicts your original premise.  This should not seem alien as automotive grade safety in an incredibly regulated product in which return on equity necessitates global platforms to be pursued by auto manufacturers alwayes yields to verifiable chain of command .

junko.yoshida
User Rank
Author
Re: Has the most important fact been omitted
junko.yoshida   1/13/2017 11:40:28 AM
NO RATINGS
That's a very good question, Victor.

I think companies are pursuing a dual track strategy -- working on the traditional ASI-D certification based on the current Computer Vision model (which can be certified), while working on AI as an additional "intelligence." But again, I am not sure how those two will meld, or if one takes over another, then, what happens?

Page 1 / 2   >   >>

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed