Quote:
Originally Posted by AndrewJ
Gday Gary
Whilst i can still see these sort of cars becoming common in the CBDs, i think the deep learning would have a hard time a few km outside the CBD. ( Be interesting to see if it could be leveraged for PEC training :-) )
That said, i had also read ( a few years back ) that research was being done for CBD/freeway type operations where a lot of the sensors would be fixed on/in the streets, thus reducing the load on individual transport units, as it would be easier to map the flows. Do you know if more research been done on that??? I can easily see that as a near term possibility, vs a truly autonomous car that can do everything by itself.
Andrew
|
Hi Andrew,
Thanks for the posts.
With the recent advances in in-vehicle sensors, including LIDAR and
cameras, the increase in computing power and the increasing sophistication
of the software, the major players are aiming for Level 4 and Level 5
autonomy and attempting to avoid the necessity for guidance infrastructure
to be built into the roadside.
The cost of these sensors is also dropping, in particular with companies
targeting semiconductor LIDAR's that will only cost a few dollars compared
to the thousands of dollars they cost today.
The first trials of modern autonomous vehicles took place in the country
and not the city, starting with the 2004 DARPA "Grand Challenge"
held in 2004 on dirt tracks in the Mojave Desert.
Back then, not one of the 15 teams that qualified for the final race managed
to finish the 132 mile course.
The following year, in 2005, five teams finished the course.
So in that short span from 2004 to the present, the advances have
been impressive, with the Google cars alone having clocked up well over
2 million miles.
The decision by Google to aim for Level 4/5 was arrived at early
in their trials when the in-car cameras reportedly captured an
engineer horsing around with a notebook computer rather than
being attentive as the backup human driver.
This is similar to the Ford decision to skip Level 3 after multiple
incidents where the engineers would nod-off in the car.
Even when Ford put a second engineer in the car to watch the first,
they were both nodding off.
So the expectation is that the general public would become similarly
blasé when sitting in an autonomous car so the computer better be
capable enough to be in control at all times.
With the majority of car trips taking place in towns and cities, it
will be no surprise if the uptake of autonomous vehicles is initially
stronger in more populated areas.
Several companies are testing autonomous trucks, so it might transpire
that they become a more common sight on rural highways in the future.
A couple of years ago I drove the 857km from Cooktown to the Tip
of Cape York. There was no shortage of gravel roads, potholes,
corrugations, rocks, bulldust and the occasional roadtrain.
By June 2019, even that frontier will be tarmac all the way and so that
drive is likely to be less challenging to a human or computer than the
8km of Pennant Hills Road here in Sydney's north from Hornsby to
Parramatta which is congested and challenging at any time of the day.
When we watch science fiction movies, we seem comfortable with the
notion that in the future we will be accompanied by seemingly intelligent
robots that can acts as scouts, descend to a hostile planet and have
reaction times that can outmanoeuvre any threat.
So in order to make such a future a reality, one might argue
we have to start somewhere.
No doubt the development of autonomous vehicle technology
will not be without incident, unfortunately including some serious accidents.
But if advances in autonomous vehicles begin to reduce accidents and
start to help save lives, then that has to be a goal worth aiming for.