h a l f b a k e r y"Look on my works, ye Mighty, and despair!"
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Should emulate all of truck driver's driving tech:
a. 2 high resolution high speed CCD (camera input) - $4000
b. 2 wide range low resolution side cameras - $2000
c. 1 wide range low resolution back camera - $2000
d. Double digital spedometer - $1500
e. Control software for gas,
brakes, gears and stearing. - $5000
f. Software to do the following:
1. attain distance to next car
2. analyze and discern type of current and adjacent lanes.
3. read and understand road signs and traffic lights.
4. recognize advancing of side cars - and warn them of danger.
5. avoid collision with a suddenly stopped car or a person jumping into path, while considering: people or cars on sidewalk, and in adjacent lanes.
6. recover from sliding problems (ice or oil on the road)
7. receive a path and alternatives, and decide on path according to traffic.
8. allow for traffic to overtake when needed.
9. alert driver center when possible problem ahead, change to ultra-fast remote manual response when requested to, using a dedicated satellite control system.
Autonomous long haul 18 wheeler drivers needed
http://finance.yaho...t-nobody-wants.html
Tons of Trucking Jobs that Nobody wants [Sunstone, Jul 25 2012]
DARPA Autonomous Vehicle Urban Challenge
http://spectrum.iee...rpa-urban-challenge
[AusCan531, Jul 25 2012]
[link]
|
| |
Perhaps not in-depth, but rather in detail. IMHO halfbakery is not just for the fun, but also for getting ideas that are halfway there, to there. |
|
| |
I'm hoping some of the annotations will assist in thinking it out, as they usually do with other ideas. |
|
| |
The idea is fine in principle, it's just not helping or providing anything new in this field. |
|
| |
Having a camera (even if it's a really expensive one) doesn't equal being able to "see". There are cognitive processes that need to be developed that will recognise obstacles, roads, roadsigns, everything - and then still be able to determine a sensible course of action when the data available to it doesn't add up. Switching to manual control means someone has to be paid to dilligently sit in a control room, not going to the toilet or eating food, should they suddenly be immersed in a near-collision situation - and probably isn't an ideal solution. |
|
| |
Some of these processes are currently in development, and there are some very, very cutting edge systems that can tell the difference between a car and a mouse for example - but not at 100% accuracy, and probably not at a reaction speed neccessary to convince civil and political authorities to let loose 10 tons of autonomous metal onto the roads. |
|
| |
I drove into central London this morning, and had to make a number of (even if I say so myself) pretty high order negotiation-style decisions requiring my signalling my intention to other road users, and in turn, understanding signalling transmitted to me through eye-contact, hand gestures and occasional horn-useage. |
|
| |
If you can write the software that can do that, integrate it into a control system that can be interfaced with a vehicle I'd like to hear exactly how you'd do that - because it's that, not the "bits" that's the blocking problem here. |
|
| |
N.B. I'm not saying it's impossible, or not a good idea - I'm just responding to the statement that it's //Not too hard to accomplish// - because it probably is quite tricky indeed. |
|
| |
What happens when you cannot signal with hand waving etc. I'm not sure this is a total requirement. We'de have the truck taking the "safe side" and not advancing if the situation is not clear. |
|
| |
Intel's advanced imaging toolkit, or NI's Labview ImageView with a team of good students and software engineers can do it IMHO with existing technology within a year, with no need for any fantastic new discoveries. |
|
| |
Just an interesting 'aside': I was visiting in Australia.
I made eye contact with an elderly lady in the drivers seat, who had just come out of an underground parking lot. She nodded to me, and then picked up a newspaper in front of her eyes and began reading. I nodded back and began crossing. Just then the car STARTED DRIVING!!! with the lady still reading her paper! I yelled "Hey!" Then caught myself. She was the passenger, the drivers seat is on the other side, so the calm Australians can drive the wrong side of the road. |
|
| |
Can the system identify a: car, bicycle, motorcycle, cat,
dog, large bird, train car, fallen rock, fallen tree,
refrigerator, washing machine, dryer (I've seen those last
three on the road), a section of road with no markings,
poor markings, dual markings, nonstandard sign, blown
over temporary sign, incorrect signage? |
|
| |
And all of these in any possible orientation, velocity,
mass, material. |
|
| |
(For the record, I work in fields that use advanced image
processing, and we can't always find an edge we know is
there, so the answer is no) |
|
| |
Author clearly has no idea of the programming complexity he is trying to describe. |
|
| |
Automotive companies have armies of software engineers and we can't even program *them* in less than a year. |
|
| |
IMHO, armies of software engineers are harder to
program than armies of truck drivers. Most of the
time their job needs less of skill and more of
patience. |
|
| |
The author was VP R&D of 4th largest company in his
country, formerly at IBM Research, where he
developed the Multimedia part of the Java language
(bought by Sun from IBM after, as Charles Petzold
wrote, IBM wouldn't know what to do with this
fantastic technology), is an expert in DSP, and
software development management. He tinkered a
bit with robots. He also many times comes up with
stupid ideas that have nothing to do with reality, only
later realizing that his understandings in physics or
biology where totally unrealistic. |
|
| |