REGISTER | LOGIN
Breaking News
Blog

Mobileye's Safety Formula Doesn't Quite Add Up

Page 1 / 2 Next >
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 3   >   >>
junko.yoshida
User Rank
Author
Re: Follow-up to this article
junko.yoshida   10/29/2017 8:14:04 AM
NO RATINGS
@Victor.Lew Hope this partly answers your quesitons.

junko.yoshida
User Rank
Author
Follow-up to this article
junko.yoshida   10/29/2017 8:09:36 AM
NO RATINGS
OK. EE Times posted a follow-up to this piece by reaching out to experts in academia whose research interests range from robotics and embedded computer systems to autonomous vehicle safety and human-robotic interactions. They include: Phil Koopman at Carnegie Mellon and Missy Cummings at Duke. Here's a link:

Experts Weigh in on Mobileye's AV Safety Model

https://www.eetimes.com/document.asp?doc_id=1332505

 

 

jpimente
User Rank
Author
Interesting Paper but highly incomplete
jpimente   10/28/2017 11:42:52 PM
NO RATINGS
I read the 30 page paper referenced by the article and found it particularly interesting however incomplete and a little self-serving. I particularly liked the principles used in the study, a) "Complete avoidance of every accident scenario is impossible;" b) without a "formalized model for fault," the auto industry can't develop driving policy software that avoids accidents caused by autonomous vehicles; and 3) the industry needs a solution that avoids "the data-intensive validation process that most AV developers seem to be planning." . I also fully agree with "some industry analysts suggesting that Mobileye's proposal has "good potential for improving autonomous driving."   I particularly liked their formulation of AV safety taking into accounts multiple agents (a termed they call multi-agent safety. I found this noteworthy since multi-agent safety is new for AV and it is not found in older industries that deal with safety, e.g., nuclear, aviation, and process control.   Issues with the paper: The authors give the impression that they solved the AV safety problem, far from it. Although they have made an important contribution, they have only uncovered the surface and much more needs to be done along the same philosophy.   Use of the term blame as being a bit self-serving. As noted in various comments, blame involved the law. In safety engineering one refers to "hazard sources" rather than "blame sources".   As noted by multiple researchers, safety is complex and the authors have only addressed one particular type of safety hazards, those arising in "behavioral planning".  As noted above, behavioral planning does not arise in older industries that have addressed safety. But AV involve safety hazards in other sub-systems as well such as "motion (path or trajectory) planning, control, actuation, and in the perception system. Thus we need similar efforts to address safety in a more holistic fashion before we can say that the AV safety problem is solved.

MikeD
User Rank
Author
Re: Impossible dilemma
MikeD   10/23/2017 8:54:29 PM
NO RATINGS
Well Mr. Lew, I don't think I know you, but happy to discuss my credentials sometime. I invite you to come to one of Linley Group Technology Conferences, and I'll even buy you a coffee! And by the way, it's not my "own bio", but in fact a brief description of my background posted by my employer on our company website. But thanks for reading. Regardless, this article is first and foremost a processor story. It's specifically about a methodology for establishing that a propriety processor running proprietary and undisclosed algorithms (i..e. a black box) to control an autonomous vehicle could not be at "fault" for causing an accident. As much as I would also enjoy reading an article on the broader topics of autonomous-vehicle safety, regarding the subject at hand you won't find a more qualified analyst.  

junko.yoshida
User Rank
Author
Re: Impossible dilemma
junko.yoshida   10/23/2017 11:33:30 AM
NO RATINGS
@Victor.Lew, thanks for your reponse. Don't worry, I am already in touch with some people you listed in your comment. This piece was written last week, right after Intel issued a newsbyte. (Reporters have to work fast, as you can imagine.) But more analysis is coming soon.

pcambou
User Rank
Author
Re: Impossible dilemma
pcambou   10/23/2017 3:17:19 AM
NO RATINGS
page 54 "Toyota should have logged errors" what a coincidence, robots too do not want any evidence to be recorded

Victor.Lew
User Rank
Rookie
Re: Impossible dilemma
Victor.Lew   10/22/2017 5:23:47 PM
NO RATINGS
http://pegasus-projekt.info/files/tmpl/pdf/AVT%20Symposium%202017%20Test%20Specifications%20for%20HAD_Folien%20.pdf

Victor.Lew
User Rank
Rookie
Re: Impossible dilemma
Victor.Lew   10/22/2017 10:13:36 AM
NO RATINGS
http://homepage.cs.uiowa.edu/~tinelli/classes/181/Fall14/Papers/Barr13.pdf

Victor.Lew
User Rank
Rookie
Re: Impossible dilemma
Victor.Lew   10/22/2017 10:02:59 AM
NO RATINGS
http://www.pegasus-projekt.info/en/home

Victor.Lew
User Rank
Rookie
Re: Impossible dilemma
Victor.Lew   10/22/2017 9:40:25 AM
NO RATINGS
so, before spitting coffee all over myself this morning, one reads Mike Demler was referenced in first paragraphs in this article (whose own bio states "He is frequently quoted in EE Times"), rather than an individual such as Phil Koopman, or Ted Keller, or Michael Barr, or John Munson, or Gene Spafford, or Yoav Hollander, or George Stark, or one of several primary airline safety industry experts, or ...?. Status quo as usual for these articles.   

Page 1 / 3   >   >>
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed