Kind: captions
Language: en
today we have ameliafe Rizzoli he's the
CTO of new Tata me one of the most
successful autonomous vehicle companies
in the world he's the inventor of the RR
T star algorithm
formerly a professor at MIT directing
research group that put the first
autonomous vehicles on road in Singapore
and now he returns to MIT to talk with
us give him a warm welcome
oh thank you Lex
it's a great opportunity is a great
pleasure to be back here I spent 15
years of my life here at MIT first as a
graduate student and then as a faculty
as a faculty member and this is where
autonomy the company essentially was
born and we did a lot of the research
that led us to you know to start this
company and eventually you know develop
all this technology what I will talk
about today is a little bit about you
know our vision on autonomous vehicles
why we want to have autonomous vehicles
you know some of the guidelines you know
on the technology development why we are
doing things in a certain way let's get
started but and you know I really would
like to tell you you know a number of
stories about why I started doing this
and why I think this is an important
technology why we ended up starting this
company so you know I've been a faculty
member here for 10 years I mean I was
happily working with my UAVs and I was
in Aero Astra at some point around 2005
mm yeah something you know there was
these DARPA Grand Challenges that
sounded cool right so I started working
on on cars as well but they are that
were that I was doing was mostly you
know I was working on airplanes and cars
to make them fly and drive by themselves
because it was cool you know just look
you know no hands you know it drives and
as it controls guys roboticist that's
all I needed
right but then in 2009 there was this
new project that was starting in the
team that was you know getting together
to write a proposal for a project on
future urban mobility in Singapore okay
now telling you the whole story but
essentially you know I got interested in
that project just because I wanted to go
to Singapore okay and then I you know
then I called the person who was putting
together the team and okay yeah thank
you for your interest but you know what
do you think that you bring to the table
and you know we had just done the dark
urban child and so well you know I know
how to make autonomous cars
so what this is a project on future
urban mobility so what do cars have to
do with with urban mobility autonomous
cars you know what would they had to do
with mobility and you know there was the
phone call the five minute phone call
that changed my life okay because she
asked me this question that actually was
Cindy Bernard who is now a chancellor
right and then I had to come up with an
excuse right so why well imagine they
have a smart phone and then a smart
phone app and then you use this app to
call a car the car comes to you you get
on the car drive wherever you go want to
go step off the car and the car you know
goes to pick up somebody else it goes to
park or something right so this was two
thousand in nine uber twas Travis
kalanick and a couple of guys and black
cars in San Francisco right so and
essentially she bought it so I joined
the team and and we started this
activity but you know the important
thing is that I started thinking about
you know there was something an excuse
that they made up in those five minutes
okay
but you know what kind of sounds like a
good idea and I started thinking more
about this and I started thinking more
about why do we want to have
self-driving vehicles okay so the number
one reason that you typically hear is we
want to have self-driving vehicles so
that we make roads safer
okay a very large number of people die
on on the road road accidents every year
what these people do not realize is that
most of those people are actually you
know fairly young like in their 20s and
30s okay ompletely they you know what
people usually say is that you know
Sebastian Thrun and you know back in the
day he gave all these TED talks up
talking about his best friend from where
he was young who died in a road accident
right and then he made a mission for his
life to reduce the road accidents right
but and I mean so any idea is that you
most of the road accidents are due to
human errors you remove the human you
remove the error
right and then you save lives okay so
this is this is typically the number one
reason that people mention when they
talk about why you want to have some
tiny vehicles second reason is
convenience
essentially if the car is driving by
itself you can do other things you can
sleep you can read you you can text
legally to your heart's content you can
check your emails or so and so forth
right this is also great third thing is
you know improved access to mobility you
know people who cannot drive me because
and I have some physical you know
impairment or maybe they are too young
they're too old already too intoxicated
to drive right so then you know if
computer can take them home
another thing is increase efficiency
throughput in in a city as cars can
communicate beyond you know visual range
for example another one is reduce
environmental impact okay now these are
all fantastic reasons you know why we
may want to have some driving vehicles
the problem with me is that if you think
about this these are all you know good
reasons but these are all ways that you
take the status quo
you know how cars are used today and you
make it a little bit better maybe a lot
better but you do not make it different
okay and really that is what I am mostly
what I was mostly interested in can we
you know use this technology leverage
this technology to change the way that
we think of mobility okay so how do you
compare all these different things okay
so this is you know quick back of the
envelope kind of calculation that you
can do in on your own you can question
the numbers but I think that the orders
of magnitude are right okay so you know
the first thing is okay so fine we heard
that a big reason for self-driving cars
is to increase safety you know save
lives great now how much is your life
worth well to yourself to your loved
ones your friends your family is
probably you know priceless
- the government is what about nine
million dollars okay so this is what is
called the this is what is called the
cost of a statistical life there was a
report that was released a few years ago
probably you know there is an update now
but I haven't seen it
the economic cost in road accidents the
United States is evaluated to be about
you know 300 billion dollars a year the
societal harm you know of road accidents
is another you know all the pain and
suffering is evaluated to be another six
hundred billion dollars a year so what
we are getting to is about almost 1
trillion dollars okay it's a big number
okay but let's look at where the other
effects are okay what is the cost of
congestion is an estimate hundred
billion dollars a year the health cost
of congestion of the extra pollution so
another fifty billion dollars a year so
you see that these are a little it just
a small change right the next effect is
actually important right so what is the
value of the time that we as everybody
in society will get back from not having
to drive okay simple calculation what I
did is I multiplied one half the median
wage of workers in the United States
which is an embarrassingly low number
multiplied by the number of hours that
Americans spend behind the wheel okay
and what you get is about you know what
was it about 1.2 trillion dollars a year
so something that you may notice is that
the value to society of getting the time
back from having to drive is actually
more than the value to society or
increased safety okay of course it's a
little bit cynical okay so take it with
a grain of salt and a grain of salt but
you start seeing you know how these
things compare and what you may notice
from this pie chart is that you know
there is still half of that is missing
what is the other half
the other half is actually the value
that you provide to society to you know
all individuals okay by essentially
making car sharing finally something
that is convenient to use affordable
reliable okay so for me car sharing or
you know vehicle share in general is a
concept that everybody loves but nobody
uses okay or not as many people as we
would like to you know use this kind of
services examples when I was you know
here at MIT I really like using hub way
you know the bicycle you know sharing
but you have to be very careful you know
if you wait too long in the afternoon
sorry there are no more bikes on campus
right or maybe very often you cannot
find a bike or maybe you cannot find a
parking spot for your bike so then you
had to buy somewhere else and then work
so that defeats the purpose of of using
that bike same thing with with cars
right so typically with you know car
sharing systems what you have is either
you have a like a two-way which is
essentially hourly rental right or you
have a one-way but in one way system
then the distribution of cars tend to
get skewed right and unless the company
you know rip repositions cars in some
you know clever way then the year you're
not guaranteed that you will get a car
where you need it and you're not
guaranteed that you will get a spot a
parking spot when you don't need the car
anymore okay if you think of that these
are both like a friction points you know
for using vehicle sharing and these are
both pre friction points that are
actually addressed by if the car can
drive itself okay so if you bring in all
the economic you know benefits of a a
car sharing system that actually works
that's something that we estimate it to
be you know it's about two thousand
dollars a year so you see that this
actually it has a like a big chunk in
this in this pie chart okay and that is
using an estimate of what we call the
sheriff factor of four meaning that one
of the shared vehicles can essentially
substitute for for in privately owned
vehicles okay
there are some studies that you know get
to this sharing factor up to ten and of
course the benefits are even more now
every time I see inter write a round
number like that I get suspicious right
you know ten is a little bit too
convenient to be true right but any so
that's something that you can find in
the literature so so this is really
where I think that the major impact of
of autonomous driving or certain cars
will come from now if you I think also
there is a lot of confusion in the
community in the world about what a
self-driving car means now what I'm
doing here I just listed this you know
five levels socially six levels of
automation you know these are the
Society of Automotive Engineers levels
okay so level zero is not a mission
that's your you know great-grandfather's
car right driver assistance level one
there is for example cruise control or
you know some simple single channel
automation partial automation you have
you know something like for example
lane-keeping and cruise control but you
still require the driver to pay
attention and intervene conditioner
automation level three
a driver is a necessity it's not
required to pay attention all the time
but needs to be able to intervene given
some notice okay and you know that some
losses I think is like ill-defined
concept and then you have level four
level five that are like a higher
donation essentially no driver needed in
some condition that is level four and in
all conditions that level five okay now
my first reaction when I started seeing
these levels and you know there is also
similar version by Nitza
is that listening to me you know a
horrible idea and the horrible idea
in the sense in because they are given
numeric levels so you have level zero
one two three five whenever you have a
sequence of numbers you are led to
believe that these are actually
sequential right that you do level zero
then you do level one thing you do level
two three four five I think this isn't
like an enormously bad idea because I
think that level 2 and level 3 that is
anything where you require the human to
pay attention and supervise the
automation and be ready to intervene
with no notice or with some ambiguously
defined you know like a sufficiently
notice they just go behind you know go
against human nature and you know this
is especially painful for me as a former
aeronautics and astronautics professor
where we saw in the airline industry
that as soon as Auto Palace were being
introduced and everybody thought that
accidents would go down
actually there were more accidents
because now you have new failure modes
induced by auto pilots okay you have
multiple fusion Pylos flu situation
awareness pylos lose the ability to
react in case of an emergency okay so
the idler and Industry had to
essentially educate itself on how to
deal with automation in a good way and
think of pile you know pilots are highly
trained professionals which is not the
same that you can say about your
everyday driver right so how do you
train people who probably you know you
know the last time they said with a with
an instructor in a car was you know when
they were 16 right how do you train
people to use the automation technology
in and do it safely right so I think
that you know distantly that front very
scary on the other hand I think that you
know the full automation when the car is
essentially able to drive itself does
not rely on a human to take over isn't
it that in a sense is easier
and you know this is what we are doing
and but you know the point is that not
all it is easier but I think that is
essential to capture the value of the
technology now if you think of it so how
do you realize the value of these
self-driving vehicles okay so the first
thing that people say is safety I think
it is true that eventually
asymptotically self-driving cars will be
safer than their human driven
counterparts however at what point can
we be confident that that is the case
are we there yet not sure okay
so so how do you demonstrate the
reliability of these self-driving cars
so we know you know they've driven that
cars for three million miles right so
with a readily small number of accidents
if I remember correctly only one was
their fault right but um actually humans
drive for many you know many times that
without accidents or so how do you
really make sure that even though the
number sounds impressive it really
doesn't have that much of a statistical
significance right and then every time
you make an update to a change to your
system to your software you really have
to validate again right so I think that
making the case for safety is actually
is a very challenging issue and we may
not be positive that these self-driving
cars are actually safer than the human
counterparts you know until you know a
really long time from now
okay so safety for me remains kind of an
questioned open question at this point
how do you get back the time value of
driving if you had you know at least I'm
speaking for myself if I have to
constantly pay attention to what the car
is doing excuse me but I rather drive
myself
okay because you know if the car is
driving and you know this is the paradox
right so the better the car drives the
harder it is for me to keep paying
attention right and this is where the
whole problem is right so there would be
very hard for me not to fall asleep or
you know not to get distracted so if I
want to get that time back
really you know the car must be able to
drive itself without requiring me to pay
attention captioning again you know is a
you know in order to make car sharing
really convenient and reliable and
sounds fourth you need the car to come
to you with nobody inside and Ford it
for that you need level four or level
five okay anything else just doesn't cut
it you know everything else for me is
just a nice gadget that you have on your
car that you show off to your friends or
to your girlfriend okay so that's about
it
right is it's not that useful so my
point is that level four or five
automation is really essential to
capture the value of this technology and
in fact the one game-changing feature of
these cars is the fact that these cars
now can move around with nobody inside
that's really the game-changing feature
okay good and you know this is you know
really what we like to do now there are
many paths that you can go after this
target okay I usually show this this
fear okay so on this figure what I show
on the horizontal axis is the scale or
the scope of the kind of driving that
you can do okay so on the left is like a
small you know pilot maybe a closed
course on the right is driving
everywhere okay on the you know like
complex environments right mass
deployment and so forth on the left
there is on the vertical axis is the
level of automation okay now really what
we would like to do is get to the top
right corner right so we have millions
of cars driving all over the world
are completely you know completely out
in a completely automated way okay
what I see is there are two different
paths that the industry is taking okay
what I show here is what I call this is
the OEM path okay
so this is the the automaker's right so
they're used to thinking of production
production of cars in the orders of many
many millions okay
and essentially what they do is they
make a lot of cars and they are adding
features to discuss you know advanced
driver assistance systems and so on so
forth right and essentially they're
following these levels 0 1 2 3 4 5 ok
and you know today you can buy cars
which even though they claim fully
autonomous you know package for $5,000
plus another $40,000 or something in the
fine print this is level 2 rights or
level 2 or level 3 so you know Tesla
said is I think the bau-t who the new
Audi a8 is a 8 they're coming out with
this we just kind of feature Cadillac I
think as a similar thing okay
the problem with that I seen that you
know you had to cross this this red band
okay
this red band where you're actually
requiring human supervision you know of
your automation system another path
where people are following is this other
okay so this is what we are doing what
way more you know where these are where
all the indications or that Weimer is
doing of course they're not telling me
exactly what they do similar thing for
uber right so essentially what they're
doing is they're working on cars would
be fully automated from the beginning
and they start with a small you know
maybe geofence application and then
scale that update operations outright
but always remaining at the full you
know High full automation level okay
another thing that is important that you
know people make a lot of confusion and
don't seem to realize the big difference
is the following when people ask me when
do you think that we will see autonomous
vehicles everywhere on the city aware
you know autonomous vehicle would be and
would be common I guess I'm okay but you
know what do you mean exactly by that
right because if you ask me when you see
that you will be able to walk into a car
dealership and get out with the keys to
a car that you know you just push a
button it takes you home that's not
happening for another 20 years or at
least okay on the other hand if you ask
me when you will be able to go to some
new city and some on one of these
vehicles that piece you up and takes you
to your destination the other thing is
happening within a couple of years okay
what is the difference there is a big
difference between autonomous vehicles
self-driving cars is a consumer product
versus a service that you provide you
know to two passengers okay so for
example what is the scope you know where
do these cars need to be able to drive
okay if it's a product and I pay you
know ten thousand dollars for it then I
want this thing to work everywhere right
so take me home you know pick two to be
into this little alley you know drive me
through the countryside on the other
hand if I'm a service provider and I'm
offering the service I can decide you
know I'm offering this service in this
particular location and by the way I'm
offering this service under these
weather conditions and maybe under these
traffic conditions okay so just the
problem becomes much more much easier
what are the financials right so if I
have to sell you in autonomy a car with
an autonomy package how much can i cost
you know what would what are my
cross-country constraints on that
autonomy package if I sell it to you you
know first of all the cost of the
autonomy package must be comparable to
the cost of the vehicle okay
you know you will not buy a $20,000 car
with a half a million dollar autonomy
package right also you can do so another
back-of-the-envelope calculation that it
is okay so let's say that what is the
value to you as the buyer of this
autonomy package let's say that the
value to you is the fact that now
instead of dragging you know for the
rest of you know for the next 10 years
you can have the computer grinding for
you what is the value of your time as
you are not driving right so do a quick
calculations again you know total number
of hours that Americans spend behind the
wheel median wage or in a value of time
what you get is you know what I get is
that you know the net present value of
the drivers time over the next 10 years
is about twenty to twenty thousand
dollars okay so then you know a rational
buyer will not pay more than that you
know to buy this autonomy package right
so now you're constrained by twenty
thousand dollars okay
or actually if you want to make a profit
out of it you know your constraint your
autonomy package cannot cost more than a
few thousand dollars okay on the other
hand if you're thinking of this as a
service then what you are comparing to
is the cost of providing the same
service using a carbon-based life form
like a human behind the wheel okay so
now you want to provide 24/7 service you
need to hire at least say three drivers
per car okay then the cost is comparable
of the order of hundred K a year okay so
now I'm comparing the cost of my
automation package to something that is
going to cost me $100,000 a year over
the life of the car okay so now the cost
of the Atlanta computer or that fancy
radar or something doesn't matter that
much okay so I have much more freedom in
buying the sensor that I need
infrastructure for example people talk
about maps HD maps right now again if I
want to sell it as a product I need to
enter have to sell it I want to sell it
on a globe
scale well global could mean older the
United States for example or all of
Europe then I need to have maps HD maps
of the whole of Europe or the continent
or any other stays or whatever I want to
sell the you know the cars if I'm
providing a service then I only need to
map the area where I want to provide the
service and by the way how do how does
the complexity of the maps scale with
the customer base that you're serving if
you think of a uniform people density
okay so then actually they land you
think that the complexity and the cost
of generating Maps scales with the
length of the road network then the cost
of the maps scales with the square root
of my customer base meaning that will
become negligible as I serve more people
okay so HD maps yes it's a pain in the
neck to collect them and to maintain
them but it's much less of a pain in the
neck that actually open it in the
logistics of a fleet serving the
population of a city okay and servicing
and maintenance you know how would you
calibrate your cameras and your sensors
you know that's not something that you
would do as a normal consumer right oh
we are not used to that when I was
little I was used to my father you know
he was tinkering with the car all the
time you know checking the you know the
timing belt or changing the oil or you
don't do any of that nowadays right so
you just sit in the car switch it on if
the yellow light you know Check Engine
comes up into the dealership right
that's all you do now imagine that you
know now you have if you want to use
your autonomy package you had to
calibrate the sensors every every time
you go out or you know you have to
upload you know like a new version of
the drivers and these are that so you
don't want to do that on the other hand
in the service model I had the
maintenance crew that can take care of
it in a professional way okay
so big difference between the two models
so there are a couple of important
takeaways right so one thing is that the
cost of the autonomy package is not
really an issue
really the cheaper I can make it the
better it is right but that is not
really the main driver in particular if
you need a lighter sensor for example to
detect a big truck that is crossing your
path by the ladder sensor okay so that
is not making the difference and maybe
you can save some lives okay any
reference to other things is intentional
the other thing is HD Maps the people
worry about you know 12 you know very
much today from my point of view HD Maps
my expectation is that HD maps within a
few years will be a dime a dozen
okay what is complicated what is
expensive now in generating all these HD
maps the mapping companies need to put
these sensors on a car on you know and
send these cars around now imagine that
I have a fleet of 1,000 cars with these
sensors on board and these cars are just
driving around the city all the time the
generating gigantic amount of data that
I can just use to make and maintain my
HD maps so I think that you know
especially from the point of view of the
operators the providers of these
mobility services very easy to collect
data to essentially make you know make
and maintain their own Maps okay so if
you need HD maps that's fine because as
soon as you start offering this service
you will be able to collect all the data
you need to generate this a generate and
maintain these maps oh by the way this
is showing an animation showing you know
like a simulation of a fleet of I think
it's a couple of hundred vehicles in
Zurich in Switzerland right so that's
where I was based until a few days ago
and as you see in essentially you have
vehicles that going through
go through most of the city you know
every few hours okay I think that for
example the uber fleet goes through 95
percent of Manhattan every two hours or
so
cos advantages you know of course you
know the you know most of the cost of
you know taxi services nowadays is is
the driver you know it's about half of
course you remove the driver from the
picture you don't have to pay them of
course the automation costs you a little
bit more servicing cost you a little bit
more but you see that you know you still
have you know you you know you can get
like a really significant increase in
the margin right meaning that you can
pass some of those you know savings to
customers right but also you can make a
very strong business case however this
is also misleading now if you think of
it okay so typically what the reaction
that you get is the following
oh my goodness now you make this thing
and then all taxi drivers all truck
drivers would be out of a job okay
and in fact one day I was summoned by
the Singapore Ministry of Manpower okay
and I was terrified
oh my goodness they're gonna shut me
down because they're afraid that that
will put all of their taxi drivers on a
State on a street in the sense of being
unemployed turns out it was the opposite
what most people do not realize is that
actually mobility services worldwide are
actually meant power-limited okay in
Singapore they would like to buy more
buses but they don't have enough people
who are able and willing to ride the
buses okay this is true pretty much the
same to for tracking same for Tarsus now
this is another back-of-the-envelope
calculation that you can do on your own
now imagine so as we know you know Ebers
be widely successful you know very high
valuation a lot of this valuation is
predicated on the fact that everybody in
the world will eventually use uber right
or something similar now something that
people don't think about is the
following now if everybody in the world
you
uber for their mobility means how many
people in the world need to be drivers
for uber do the calculation what you see
is that one person out of seven must
drive for uber if uber is surveying the
whole world do you see that happening no
way right so people still need to be you
know teachers doctors you know policemen
firemen you know or you know some people
need to be kids you know so that is
something this cannot happen how are we
facing these paradox in a sense right so
you know today what you have is people
who drive around but what is happening
today is that we are all doubling up as
drivers for ourselves and in fact we do
spend about one-seventh one-eighth of
our productive day behind the wheel
very often ok so you know for me you
know did the big the big change is will
be more on the supply of mobility rather
than on job loss I mean of course if you
increase supply of mobility you know the
the cost of mobility will have to you
know we will you know probably go down
wages for drivers will go down right so
that is that is a that is a that is an
issue but you know maybe other you know
baby balance by like a added value and
service or other things that you can
imagine another thing about truck
drivers you know something that they
recently learned 25 percent of all job
related deaths in the u.s. are actually
by trucks drivers ok is the most the
single most dangerous industry that you
can be in so maybe if you can take some
of those people out of those trucks and
maybe supervise remotely control a truck
sitting in their office instead of
sitting in the truck you know that that
may be actually benefit to them back to
the question of when we lot on most
vehicles arrive and you know in a sense
this is what you know what our
prediction our vision is right so what
we will see is that what we think is
that
you have a fairly rapid adoption of
self-driving vehicles in these mobility
as a service model okay as a fleet of
shared autonomous vehicles that people
can use you know to go from point to
point right rather than all of course
eventually you know people will be able
to buy these cars and maybe own them if
they really want but you know that is
something that is much later in time for
for a number of reasons some which I
discussed okay so this is you know what
we expect in terms of the timeline for
this now what is the state of the art
for autonomous technology today you do
see a lot of demos for from a number of
companies you know doing a number of
things right but but a lot of the things
that you see are not too much different
from this video I don't know if any of
you recognizes this video but you know
look at the cars this was actually done
by LSD commands in the late 90s in
Germany okay no fancy GPUs no it was
just a cameras and some you know basic
computer vision algorithms but
essentially he was able to drive for
hundreds of miles on the German highways
okay if you're not showing
something that goes beyond that you have
not made any progress you know over then
over the past 20 years okay yeah you're
using fancy deep learning and GPUs and
things nowadays but you're doing what
people were doing 20 years ago you know
okay so you see arena clearly there's a
lot of hype in these things but you know
if you see something like that I don't
think it's very impressive
okay people people you know knew how to
do that for for a very long time
something that I find a little bit I may
be biased clearly right but this is
something that I find a little bit more
exciting this is actually footage from
you know our daily drives in Singapore
okay this is four times in real time we
don't drive that fast
okay but essentially what we're doing in
Singapore we are driving you know in you
know public roads
normal traffic what you will see is not
so but you know do we have you know
construction zones intersections traffic
you know you know of both sides we will
get to a pretty interesting intersection
has a red light will turn to green in a
second human mind in Singapore they
drive on the left right so making the
right turn is what is hard because you
had to cross traffic right and here you
have in a lot of traffic and you know
the car is making the right decision in
all of these without any human
intervention right so I think that in
this day and age if you're not showing
the capability of driving in traffic in
an urban situation like that you're not
really showing any advance over what
people were able to do 20 years ago okay
and you know I mean as you can see if I
saw the intercessions other cars
pedestrians you know all kind of like a
crazy interactions you know you know the
cars park in the middle of the street
that you had to avoid go to the other
lane you know things like that okay so
this this is what you had to do every
day and you know this is what we are
doing every day in Singapore we are
doing every day here in the c4
if you're aware of botany we are driving
you know cars we are allowed by the city
of Boston to drive our cars autonomously
in the Seaport area so what are the
technical challenges okay so actually
this is a slide that I did I'm fairly
reusing from a talk that I'm not chakra
the founder and CEO of mobile I gave
here at MIT a few months ago okay so
this is what he said okay so it's not
what I say what he says is that the big
challenges are sensing you know
perception it's mapping and then is what
he called driving policy right that I
will call more like a decision-making
okay now what he said is that sensing
perception is a challenge but is a
challenge we are aware of and then we
are making rapid progress on getting
better and better sensing perception
algorithms okay second it's HD maps what
he said is that it was a huge logistical
nightmare so he didn't want to deal with
that you know like mobile I tries to
avoid that from my point of view as I
said you know for me it's the maps it is
a replay in the neck to get those maps
but in a few years
maps will be a dime a dozen okay so
we'll get all the mapping data that we
want and we need so the big problem is
during policy
okay the remaining problem is drawn in
policies so how do you do it not and you
know this is a typical example of things
that we encounter in you know in any
color urban driving situation so you
will see a video so this is a case where
we are at the traffic light we are
stopping the traffic you know the light
turns green we are making the turn this
is a pedestrian crossing the street wait
for the press tree and go through it and
then we see that there is a truck that
is part in the middle of our lane so we
need to go to the other lane which is in
the opposite direction there is a model
excuse me a motorcycle coming so we had
to handle all that kind of situation
right so how do you write your software
in such a way that your car
is able to deal with this kind of
complicated situation by itself okay and
my point is that you know this is not
really about negotiation is not about
policy why do you have rules of the road
my claim I have not proved it
mathematically yet but my claim is the
following
the touching the rules of the road were
introduced exactly to avoid the need for
negotiation
when you drive okay when you're walking
as a person you just walking down the
hallway you know walking down the
infinite corridor and there is a person
come in the other direction
there's always that awkward moment right
away you're trying to linger I go left
I'll go right
right with cause you you don't do that
right so in cars the side everybody go
right or in other places everybody go
left period and you don't negotiate that
okay you get to an intersection the the
the light is red you stop you know
saying I'm putting I'm really in a rush
you know do you mind if I go no you
don't do that right so it's red and you
stop okay so the rules of the road have
been invented by humans in order to
minimize the amount of negotiation and
you know and you know in particular okay
so this is a slightly I mean this is
actually very old video but I kind of
like it so now our car is a little bit
more aggressive but you know what you
see here is this case you know this is
how the car behaved in that particular
situation so you see it's raining red
light turns green there's a pedestrian
crossing our path
so we heel to the pedestrian you see
that there is a you will see that there
is a truck that is parked on the on the
left lane in the middle of the lane so
we had to go around it but this is a
motorcycle that is approaching so we had
to be careful in going to the other lane
okay so we squeezed through the through
the motorcycle you know we try to go
very slowly next to squishy targets
right but then as soon as we pass the
truck the truck driver decides to get
moving okay
so then what we do is we wait for the
truck to get you know to get going and
then go back to our lane now imagine
writing a script you know
or you know if then else if there is a
track but the truck is moving and then
do this and this the network so you know
what to do that right so how do you
handle this kind of situations okay so
the industry standard
you know this approach to this was - and
by the way this is what we did at the
time of the dark urban challenge okay so
we had a lot of if-then-else statements
or you know finesse test machines or
some logic that was encoded by you know
some furnaces machine kind of kind of
things the problem with that is it's
very hard to come up with this logic and
is essentially impossible to debug it
and verify it right so I spent many
miserable months sitting in the naval
airbase in Weymouth right so here in a
rental car just plain interference with
our autonomous car trying to adjust all
these logic and parameters and things so
I vowed that I would never do it again
I was just miserable experience I'm
happy to say that actually we did come
up with a much better way of doing it
and you know by the way this is a video
from the Caltech team at the dark urban
challenge as you can see they're trying
to go to an intersection they decide to
go then for some reason they decide not
to back up out of the intersection so
the director of DARPA you know Tony
Taylor at the time he was there he went
like that so they were out of the race
okay so as soon as CCO saw that what
happened here there was essentially a
bug in the logic Caltech you know very a
team of very smart people very capable
dedicated people work on these for
months they didn't catch this Bank this
Bank they were out of the race right so
it's very easy to make mistakes and it's
very hard to find those bugs okay so as
a reaction to that you know there is
this new
is it possible to cut the sound thank
you
so now what people what you hear people
saying is well there are too many rules
of the road it's impossible to code all
of them correctly so let's not do that
just feed the data you know feed the car
a lot of data and let the car learn by
itself how to behave okay
okay and this is what you see you know
you know there are a number of circuits
and other efforts that are trying to use
all these you know deep learning or
learning approaches to to get to the
fore end to end driving of of cars okay
so you see a video from Nvidia okay
understand this is a course on deep
learning for cars right but so so I
don't want to sound too negative on the
other hand I will try to be honest in
what I think ok so you know there are a
number of problems right so that's what
is happened to us right so one of our
developers you know you know super
bright lady from you know you know
Caltech and you know the first version
of the code for dealing with traffic
lights essentially the reaction that you
know that that they had for for the
yellow light was if you see a yellow
light speed up what the heck oh this is
what my brother does okay so there is
always the danger that you learn the
wrong thing okay did the wrong behavior
in a sense of course there are some
situations in which accelerating when
you see a yellow light is actually the
right response but it is not always the
case right so there are some other
features of the situation that you need
to examine right also the other thing is
as a cartoon right so you know you want
to be able to explain why the car did
something and I would say that more than
explaining because now you also see
articles in which people say Oh
a fun way of explaining why they do not
not for him to carve decided to do
something right I want to show you is
some okay so these are the noodles that
were activated just saying that you know
what if I do an F in a fast MRI of the
brain and they see what neurons what
areas of the brain are activated when I
watch a movie then I know how the brain
works no I have no idea okay the point
is that yes you want to trace the reason
the cause for why they can't behave in a
certain way but you also want to be able
to revert the cost right so you want
that information would be actionable in
some sense right so you want you want to
know that okay this happened because of
this reason and this is how I fix it
okay and the other thing that you know
society that is hard to do with purely
based learning algorithms on the other
hand you can let me actually skip that
in the interest of time okay the reality
is the following that it is simply not
true that there are too many rules of
the road in fact any 16 year old in the
states can go to the DMV get the booklet
study the booklet do a written test and
be given a learner's permit okay and
actually this is what we require of
every single licensed driver in the
United States okay we don't say just
drive with your dad or mom for a few
thousand miles and that will give you
the license no we ask them you know show
me that you study the rules and you
understand the rules okay so how many
are the rules of the road actually went
to an exercise of counting okay and what
they did I can do like a cluster them so
essentially you have rules on who can
drive when and where what can be driven
whenever you know at what speed in what
direction
who yields to whom right how you use
your signals active signalling how do
you interpret the signals that you see
on the road right and where you can park
away you can stop that's essentially it
you know this is this these are all the
roads okay so not that many it's
collected twelve categories what is true
is that the number of possible
combinations of rules and the instance
instantiation of the rules given the
context of you know the scenario where
other actors are pedestrians are and
where other cars are that is a humongous
number okay
so you don't want to code you don't want
to be to essentially any generative
model that gives you what is the right
response to all possible combinations of
rules and instantiations of actors that
is something that is just coming up
totally you know intractable you just
cannot do that but the point is that not
only it is hard to code the good
behavior what to do in every one of
these situations I claim that is also
hard to learn the good behavior because
now you have you need to have enough
training data for every possible
combination of rules and instantiations
good luck with that
okay on the other hand it is very easy
to assess what is a good behavior and
that's why I was showing this slice on
np-hardness right so what is the problem
that is np-hard the problem is np-hard
where if you have a non deterministic
system that is generating a a candidate
solution then it is very easy to check
whether or not that candidate is
actually a solution of your problem and
that's something that you do in
polynomial time okay so in a sense what
what I claim is that if you have an
engine
that is able to generate a very large
number of candidates and all you do is
checking and then you know what you do
is checking whether or not each one of
those candidates is good with respect to
the rules then that's all you need and
turns out that you know the algorithms
that I worked on during my you know
academic career
where exactly generating that very large
number in our TRC star these are
algorithms that work by generating a
very large graph exploring all potential
trajectories reasonable trajectories
that a robot a system that can take and
then what you do is you check them for
you know whether they satisfy the rules
or not you see that is very different
from giving the rules generates
something that satisfies everything
rather than given a candidate check
whether or not this candidate satisfies
the rules the generating the rules the
generating candidates given all the all
the constraints is a combinatorial
problem checking a single candidate for
compliance with a number of rules is a
linear operation in the number of rules
so that's something that you can do very
easily okay and then essentially what we
have in our cars today we are using
these formal methods okay so essentially
we write down all the rules in a in a
formal language you know so you know
very precise you know like your syntax
and then what you can do is you can
verify whether your trajectories satisfy
all these rules written in this language
that is automatically that can be
automatically translated into something
look like a finite state machine by
computer okay but there's not something
that you do by hand it's something that
is done automatically and then what
happens is that what we have is we
generate trajectories these trajectories
are you know you can think of these as
trajectories that now are not all the
trajectories in the physical space and
time but are also trajectories evolving
in this logical space telling me whether
or not and to what extent I am
satisfying the rules okay and that's all
there is
okay so this is um you know for example
regular little example so you know
initially what we are doing is work so
this was very early days on Deuteronomy
where we're still working on a research
project with industry with customers so
our customer in this case wanted us to
do an automated parking application and
then what you see on the left is our
planner eager planet that is just trying
to to park the car right avoiding
hitting other cars but you see is kind
of ignoring the fact that you have lanes
and direction of travels right so you're
putting the rules and what you see is
what is on the on the right where now
what the car is doing is not only
finding the trajectory to go park but it
does so obeying all the rules that are
imposed on that particular parking
structure okay something that is very
important and you know this is something
that we as humans do every day is to
deal with infeasibility okay so very
often you're doing your planning you're
trying to plan your trajectory you have
a number of constraints and well sorry
but turns out that there is no
trajectory there's no possible behavior
that you can do that will satisfy all
the rules so what do you do the computer
time sorry does not compute unfeasible
still driving this car I need to do
something right so you do need a way of
dealing with infeasibility
the way that we approach this problem is
being having this idea of hierarchy of
rules okay and my claim is that all
bodies of rules generated by humans are
actually organized hierarchically
typical example is the Three Laws of
Robotics by asana right so the first law
of robotics is a robot will not harm a
human right or cause a human to come to
our second law is a robot will obey a
human orders by a human a human unless
they violate the first law and the third
law is a robot will try to preserve its
own life or preserve itself unless it
violates the first two laws right
same thing in in when you drive right so
there are some rules that are more
important than others right so for
example do not hit people do not hit
other cars and then lower priority level
is to be driving your lane the lower
priority level is maybe maintaining the
speeding or something like that okay and
then what we do is come up with now we
have this product graph of trajectories
in the physical and logical space on top
of that we can give them a cost right
what we need is a essentially a total
order what we use a lesser graphic or
drink okay when we have violating an
important rule even by a tiny amount is
much worse than violating a less
important rule by a large amount okay so
that gives a total order structure for
the cars and then essentially what we do
is we solve a shortest path problem on
this graph okay which is exactly what
you do the robot is one on one when you
try to do you know do any kind of motion
planning okay and well you know this is
in a collection of a few interesting
things so here we need to go to the
other Lane but you see that there is the
other vehicle coming so technically we
could not go to the other Lane but you
see that you know as long as it is safe
to do so the car will go into the other
Lane okay you know and again you have
like a lot of you know like a difficult
situations that the car was able to
handle by itself without any scripting
or without any like a special
instruction for that particular case
okay so what is
here the problem here is that okay so
you can do all of this right and but
then you know assuming that everybody is
running this minimum violation planning
you know everything will be okay the
problem is that humans introduce a lot
of uncertainty in the whole thing okay
now you can think of disease asking the
question so when I was young in a if
that is two years ago I thought that I
take all the rules of the road and you
convert them to this formal language you
put them in your software and you're
done and then and then you go and look
at these rules of the road and then you
see that they are a mess
okay these rules are just not the sound
theory in the sense that not complete do
not cover every possible case and are
not consistent