Kind: captions
Language: en
it's not rush hour there's no
construction no accident but boom
traffic jam
seemingly out of nowhere
why
researchers call this a phantom traffic
jam but it's not supernatural
it's mostly caused by humans these are
jams that occur even without the
presence of an accident without the
presence of say
on-ramp so no seeming cause here's how
they happen when vehicles are bunched up
too closely together then a
small
sort of tap on the brake pedal of one
car at any point in time could cause a
driver behind it to also tap on its
brake pedal a little bit
this continues down the line of cars
each one pressing the brakes just a
little more until there's a car that
actually has to come to a complete stop
and then all the vehicles behind it also
have to stop and then we have this
formation of a traffic jam that may have
just started from this little little top
on the brake pedal
this is different than just regular
congestion although they do happen
during heavy traffic
these phantom traffic jams are actually
super energy inefficient and more
dangerous than typical congestion
phantom jams are bad because they force
drivers to drive stop drive stop and
that's where all the energy and
efficiency comes because every time you
stop you have to re-accelerate and
that's how you burn fuel when a vehicle
is coasting or decelerating it's
emitting much less than if it's sort of
you know blasting the the engine but
there's a secondary impact of phantom
gems and that's the fact that
they increase the likelihood of
accidents
so if phantom traffic jams are mostly
created by humans could the solution be
non-human drivers
no not that kind this kind self-driving
cars also known as automated vehicles
could they actually eliminate some
traffic jams and make the roads safer on
average if a driver stops and drives and
stops and drives the travel time would
have been exactly the same if the person
had driven at the average speed the
whole time and so that's one of the
things that traffic flow smoothing is
trying to achieve in the case of
automated vehicles i would say we
actually have this additional
opportunity that while we're helping
passengers get to their destination we
can do a few other things along the way
we can help smooth out traffic we can
help to make the roadway safer we can
help to make other roadways more energy
efficient reduce air pollution
experts say that having all autonomous
cars on the road would be the best way
to keep traffic moving smoothly but can
self-driving cars prevent phantom
traffic jams if they're sharing the road
with humans so in the case of phantom
jumps the intention of the human is
usually to catch up with the car in the
front which is something relatively easy
to model for a machine and which is also
something relatively easy for a machine
to counteract
so to counteract the start of a phantom
jam the autonomous cars would figure out
the best following distance and speed to
avoid making any instability in traffic
worse
but on the road humans aren't always
predictable or let's be honest paying
attention at least not compared to other
autonomous cars humans are complex and
also the variety of goals they have when
they go on the freeway is quite wide in
some ways automation is easier when the
human is taken out of the equation would
a robot slow down to look at an accident
on the highway
probably not and a human's main goal
probably isn't to prevent a phantom
traffic jam even if that human is a
professor who studies the dynamics of
transportation systems i'm definitely a
person who if i see a lane that's open
i go there and i sort of move in and out
of sort of traffic and so that can
actually be optimal sometimes though the
definition of optimal may vary between
drivers and between human and machine
there's a notion of sort of being
socially optimal
versus you know optimal for yourself
this is sort of the benefit i think of
autonomous vehicles is that if you can
look a little bit beyond yourself cars
are doing things that are sort of
optimal for the sort of global system as
opposed to sort of a local system so for
researchers like kathy wu the goal is
for the automated cars to be safer
better drivers than humans not replicate
human driving if we want to do better
we as humans
should not be telling the automated
systems what to do or it will just you
know reflect back at all of our
imperfections that we can directly
specify this objective and what was
remarkable for us to find is that the
system could actually identify
strategies
that were effective in
achieving some of these objectives that
are different than how humans drive
researchers are still testing traffic
control technology in autonomous cars
and there are still a number of barriers
to accessing these kinds of vehicles the
aspect of cost and access is something
that concerns me i don't have an
answer i think that what we'll see early
on at least is automated technologies in
more niche scenarios so we'll see them
probably in automated trucking or
freight in low-speed environments like
college campuses in fair weather
locations like arizona like california
in
shuttle buses on fixed routes and so
sort of in these restricted domains this
is still rather narrow but this i think
will allow the designers and engineers
to test out the technology adjust for
risk gradually build consumer acceptance
and trust one of the things i like to
tell people about autonomous vehicles is
that it's perfect if we can get it right
but we should also understand the things
that sort of arise that we as a society
need to sort of come to before we sort
of release them all on the road