REGISTER | LOGIN
Breaking News
Blog

$6,000 of Electronics in Car by 2022

NO RATINGS
Page 1 / 2 Next >
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
perl_geek
User Rank
Author
Re: what is needed for robo cars
perl_geek   6/28/2017 2:29:56 PM
NO RATINGS
Sw guy is quite right about "white-out" conditions, which still reap a crop of accidents, especially helicopters.  (They're hard enough to keep upright at the best of times. On a surface covered with loose snow, they can generate their own zero-visibility cloud within seconds.)

Greg504
User Rank
Author
Re: what is needed for robo cars
Greg504   6/28/2017 1:39:44 PM
NO RATINGS
I don't like using poles but I am generally going under 20 miles per hour when using that method but I have gotten home that way.  And yes drifts is the correct word for piles of snow resulting from blowing winds.  Also since I am from Indiana we have lots of flat fields so the worst that happens if you get off is you get stuck.

sw guy
User Rank
Author
Re: what is needed for robo cars
sw guy   6/28/2017 8:06:36 AM
NO RATINGS
Beware of poles: When I was young, a rally driver got a serious accident (may be was killed) because he assumed road would always follows poles, and this missed a turn at the top of a hill.

BTW, snowy conditions are really difficult for us mere humans. My brother (he lives in a mountain area) once saw a car hiiting heap of snow (google translate tell me its called drift, right ?) beside of the road because driver was no longer able to see a 90° dihedral marking limit of the road: both plans were white as snow, after all. In the former days of aviation, there were crashes due to inability for pilots to see difference between white ground and white sky (as it may happen sometime), making them unable to discover they were dangerously nearing earth.

realjjj
User Rank
CEO
Re: Autonomous Vechicles by 2020??
realjjj   6/27/2017 8:41:00 PM
NO RATINGS
All that makes sense.

With full autonomy, parents need to let their kids ride in such vehicles and if in some situations the car doesn't make the best decisions, the circus in the press would be very harmful - look how hard EET is going after Tesla for one incident where they are at little fault. In theory any reduction in road incidents is a positive ofc but i don't see full autonomy as commercially viable before they can offer much more.

Not sure, maybe traction control can determine the state of the road without the need for extra sensors. A lot can be done and will be done with high res maps, high accuracy positioning, real time data, fleet learning, sensors.

Have you seen Levant's presentation at Hot Chips? https://youtu.be/cGS11oJUnuA?t=3674  They got an active suspension system. It's reactive for now but, if am not mistaken, inthe Q&A section someone asks about going proactive and they are looking at that. Teslas today (with air suspension) adjust height based on your previous actions in a particular location and that's cool but so much more can be done.

I do think that too many people, not necessarily you, are focusing on very predictable issues that are part of the basic navigation, the 99% of the time that is easy. You can't even begin to claim full autonomy if you can't find the road or handle snow and the car will have to be able to handle far tougher situations. The corner cases are the hard part. Something that can't ace basic navigation is a Roomba not an autonomous car LOL.

 

elizabethsimon
User Rank
Author
Re: Autonomous Vechicles by 2020??
elizabethsimon   6/27/2017 12:30:03 PM
NO RATINGS
@realjjj

You are right that it would be amistake to commercialize something that's just better than the average driver. Whether it would have to be much better than the best driver in every situation, I'm not sure. It might be good enough to be at least as good as the best driver. What I was talking about was the potenial to reduce accidents which would happen at a lower competence level.

I agree that the cars already have better sensors than ours in certain conditions. What is lacking at this point is the algorithms that would allow them to properly interpret what they "see". This has the potential to be useful even before we get to L4/L5. For example, my pickup has a temperature sensor. When the temperature gets below a certain point, I get a warning that ice is possible. What would be really useful is if it could use other sensors to detect that there really is frozen water on the road. Even better if it could tell if the frozen water was ice or snow...

Greg504
User Rank
Author
what is needed for robo cars
Greg504   6/27/2017 7:14:04 AM
NO RATINGS
Can electronics look at a snow covered road, use the utility poles on each side to make a guess at where the road is?  Can it do it if there are only polels on one side?  What about if instead of utility poles there is a fence?  On some roads the poles or fence are closer to the road edge and on some they are farther from the road edge.  The human driver either knows from prior experience or notices the distance during better conditions a couple of miles earlier or when turning onto the road.  They tend to do this without having to even realize they need to think about it.

Also robo cars will have to be much better than drivers to get acceptance.  You see this in airline traffic vs car traffic.  Airlines on thier worst days are safer than travelling by car.  If you have to many airline accidents you will have people driving places instead of flying even if flying is still statistically safer.

realjjj
User Rank
CEO
Re: Autonomous Vechicles by 2020??
realjjj   6/26/2017 10:39:25 PM
NO RATINGS
LOL your comments are very Dada but Dadaism is fun

Tesla offers Level 2 while this convesation was about L4/5. If your toaster is not a better driver than you are , it doesn't mean that you are the best driver in the universe.

However, a current Tesla can see through fog and heavy snow better than we can.

EELoser
User Rank
Freelancer
Re: Autonomous Vechicles by 2020??
EELoser   6/26/2017 10:05:57 PM
NO RATINGS
WRONG.

Teslas or any other car cannot drive in autopilot on snowcovered roads.  It can't "find" the road edges if there is no curb or landmarks, so on a snowy day on a rural road, you could end up in a cornfield.  

Humans do this much better.

 

realjjj
User Rank
CEO
Re: Autonomous Vechicles by 2020??
realjjj   6/26/2017 9:50:54 PM
NO RATINGS
Radar enables cars to go beyond optical so yes cars have better sensors than we do, especially in those kind of conditions.

EELoser
User Rank
Freelancer
Re: Autonomous Vechicles by 2020??
EELoser   6/26/2017 8:04:34 PM
NO RATINGS
The cars sensors are better than ours?

Not really, they use the lane striping a lot.  What if the lane striping is covered by snow?

I doubt any current day Tesla can do autopilot on a snow covered road.

They can't even avoid potholes like a human can!

And what happens when those sensors are actually covered by ice or rain from a passing truck, etc.

None of this stuff will be overcome for 30 years.

The human needs to be the responsible party so he can be blamed for the accident, not the car company.

 

Page 1 / 2   >   >>
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed