Chris Urmson: How a driverless car sees the road
Chris Urmson - Roboticist
Chris Umson is the Director of Self-Driving Cars at Google[x]. Full bio
invented the automobile.
for the first public test drive,
crashed into a wall.
reliable part of the car, the driver.
we've added air bags,
started trying to make the car smarter
a little bit about the difference
with driver assistance systems
a little bit about our car
and how it reacts and what it does,
a little bit about the problem.
on the world's roads every year.
are killed each year.
falling out of the sky every working day.
other than drive.
increased by 38 percent.
than it was not very long ago.
in America, which is about 50 minutes,
workers we have,
about six billion minutes
so let's put it in perspective.
life expectancy of a person,
who don't have the privilege
drive to work in the morning,
of piecing together bits of public transit
that you and I have to get around.
these driver assistance systems
and incrementally improve them,
into self-driving cars.
that's like me saying
one day I'll be able to fly.
something a little different.
about three different ways
than driver assistance systems.
with some of our own experience.
of a self-driving car
they were 100 Googlers,
them to use it in their daily lives.
this one had a big asterisk with it:
but it could still fail.
we let them use it,
was something awesome,
to bring a product into the world.
who came in and told us on the first day,
What are we thinking?"
"Not only should I have it,
because people are terrible drivers."
the people inside the car were doing,
and realizes the battery is low,
and digs around in his backpack,
the charging cable for his phone,
puts it on the phone.
65 miles per hour down the freeway.
it's kind of obvious, right?
the driver is going to get.
the wins we really need.
a little technical for a moment here.
and along the bottom
apply the brakes when it shouldn't.
and the car starts stopping randomly,
the car is going to apply the brakes
to help you avoid an accident.
the bottom left corner here,
it doesn't do anything goofy,
out of an accident.
a driver assistance system into a car,
of technology on there,
to have some operating properties,
all of the accidents,
along the curve here,
that the human driver misses,
by a factor of two.
dying every year in America.
that looks like this.
more sensors in the vehicle,
operating point up here
gets into a crash.
and we could argue
I could say something like "80-20 rule,"
to that new curve.
from a different direction for a moment.
the technology has to do the right thing.
is a driver assistance system.
to traffic accidents
is probably making decisions
between these two,
I'm never actually going to get there.
the system can handle uncertainty.
stepping into the road, might not be.
nor can any of our algorithms,
a driver assistance system,
that's completely unacceptable.
can look at that pedestrian and say,
and then react appropriately after that.
a driver assistance system can ever be.
the differences between the two.
how the car sees the world.
where it is in the world,
and aligning the two,
what it sees in the moment.
are other vehicles on the road,
over there is a cyclist,
if you look really closely,
is in the moment,
we have to predict what's going to happen.
is about to make a left lane change
what everybody's thinking,
how the car should respond in the moment,
quickly it should slow down or speed up.
just following a path:
pressing the brake or gas.
at the end of the day.
and the other boxes on the road,
and roughly where the other vehicles are.
understanding of the world.
on neighborhood and city streets,
new level of difficulty.
of us, cars crossing in front of us,
problem by comparison.
that problem solved,
to deal with construction.
forcing it to drive to the right,
in isolation, of course.
through that construction zone as well.
breaking the rules, the police are there
that flashing light on the top of the car
it's actually a police officer.
on the side here,
differently as well.
other people have expectations:
to yield to them and make room for them
stood in the road,
that this means stop,
we should continue.
is by sharing data between the vehicles.
sees a construction zone,
so it can be in the correct lane
deeper understanding of this.
that the cars have seen over time,
of pedestrians, cyclists,
what other vehicles should look like
we could take from that a model
to move through the world.
crossing in front of us.
and we anticipate
and around the car to the right.
coming down the road
to drive down the shape of the road.
going to make a U-turn in front of us,
and respond safely.
for things that we've seen,
lots of things that you haven't
through Mountain View,
in the DMV handbook
to encounter that,
with just ducks.
The car reacts to that.
anywhere other than Mountain View.
to deal with drivers,
jumps out of this truck at us.
with the green box decides
at the last possible moment.
the car to our left decides
blow through a red light
blowing through that light as well.
the vehicle responds safely.
who do I don't know what
pulling out between two self-driving cars.
with a lot of stuff there,
down pretty quickly.
with the cyclist again,
we can't actually see the cyclist yet,
blue box up there,
really easy to understand,
to turn that laser data and look at it,
at laser data, you can see
is that cyclist.
has turned yellow already,
in the imagery.
to proceed through the intersection.
his is solidly red,
is going to come all the way across.
were not paying as much attention.
and fortunately for everyone,
some pretty exciting progress,
to come to market.
in our simulators every single day,
that our vehicles have.
this technology on the road,
is to go through the self-driving
it's a really complicated problem,
in four and a half years,
to get his driver's license.
to making sure that doesn't happen.
I've got a question for you.
is pretty mind-boggling.
driver-assisted and fully driverless --
going on out there right now.
for example, Tesla,
that's kind of going to be a dead end
that route and get to fully driverless
is going to say, "This feels safe,"
and something ugly will happen.
and it's not to say
aren't going to be incredibly valuable.
in the interim,
to help someone like Steve get around,
to change our cities
these urban craters we call parking lots,
with huge interest.
CU: Thank you. (Applause)
About the speaker:Chris Urmson - Roboticist
Chris Umson is the Director of Self-Driving Cars at Google[x].
Why you should listen
Since 2009, Chris Urmson has headed up Google’s self-driving car program. So far, the team’s vehicles have driven over three quarters of a million miles. While early models included a driverless Prius that TEDsters got to test- ... um, -not-drive in 2011, more and more the team is building vehicles from the ground up, custom-made to go driverless.
Prior to joining Google, Umson was on the faculty of the Robotics Institute at Carnegie Mellon University, where his research focused on motion planning and perception for robotic vehicles. During his time at Carnegie Mellon, he served as Director of Technology for the team that won the 2007 DARPA Urban Challenge.
Chris Urmson | Speaker | TED.com