hi folks lately we've seen a lot of
impressive improvements from Tesla's
autopilot system one of the major
components of this system is the radar
and because of this radar sensor working
impressive accident avoidance and driver
assistance features we got traffic where
cruise control autopilot
collision-avoidance a pedestrian
avoidance unfortunately zombie avoidance
but that's easily manually overridden by
a sliding on the accelerator whole
different topic anyways I thought we
could dig a little bit deeper into this
radar piece of hardware which is a
amazing tiny piece of hardware that at
that one of my viewers was able to dig
up the Boche manufacturers features in
specs paper so let's take a look at this
critical component i got four pages i
have written note this was a little bit
of a research project for me and that I
learned a lot
so let's dig it dig into it the sensor
is manufactured by Bosh Bosh also offers
a scalable design and can modify the
hardware to suit the needs of the
specific met vehicle manufacturer and
the vehicle requirements because Bosh is
able to tailor the sensor to the auto
manufacturers specs which of course
Tesla most definitely has had some
changes done and of course we also lack
information on what those changes might
may have been
by Tesla we're just going to have to go
based on the publish documentation from
Bosh and one thing although this call
touched on this a little further
one thing I can tell you is testable
claims of distance that the vehicle can
see is considerably further by about 90
meters compared to what box claims are
well touch on that at the end closer to
the end so let's hit up the stock
information here one of the sensors of
multimodal with four independent
received channels and dim digital
beamforming multimodal means that it
varies range and resolution depending on
the target be informing means it is
designed to detect objects within a
certain range digital being forming is
the amplitude phase variation is applied
to signal after analog downconverter /
digital down conversion the received
signals from the antennas are first
passed from the ADC and DC before
summation of the operation being forming
is used for directional signal
transmission and reception it has the
ability to change both the amplitude and
phase with helps that has the ability to
change both amplitude and phase with and
helps with power variation as well as
being steering in the dirt desired
direction
ok let's see what you got
my handwriting was crap so i retyped
everything up
oh my bullet points and notes i got a
lot of them here summed up this all
allows the sensor to be configured into
pit with independent antenna for
different directions improved accuracy
and adjustable field of view depending
on the situation by focusing the radar
at a narrow main lobe with angle of plus
or minus 6 degrees the system can react
two vehicles of long-range and perform
well at higher speeds as well as
maximize interference
excuse me as well minimize interference
from vehicles in adjacent lanes so think
of that as being like okay
autopilot or traffic where cruise
control you're in a clustered situation
you got lots of vehicles around you by
focusing the radar think of it like
focusing flashlight you can zero in on a
specific section or object or in this
case the lane directly in front of you
it helps keep the vehicles on the side
out of that field of view that's
directly in front of you so it doesn't
interfere basically okiya help you keep
proper distance and react quicker to
what's directly in front of the vehicle
the elevation antenna has a plus with a
plus or minus plus minus 42 degree
opening achieves a close-range detection
zone capable of detecting a pedestrian
stepping out onto the road from behind a
parked car as as example as a early
stage of detection due to the high level
of integration of sensor control
functionalities Josh was able to achieve
compact design allowing he's a placement
within a really wide range
a vehicle design so first gen autopilot
it was a sensor on the bottom of the
bumper nothing too special about that
then they realized okay that placement
really kind of sucks gets a lot of slush
snow there so they moved it up and now
it is actually behind the t emblem the
intestinal insignia on the front bumper
of the Refresh design model s cars and
of course behind the tea on the X and
most likely who sees me behind the tea
on the model 3 equipped with a
self-calibration system the sensor
searches for reference points and first
journey and calculates sensor access
deviation from dynamic driving access
axis not access but axis access
there was a system software compensates
for this deviation during learning phase
some features or functions may be
disabled until complete you'll mostly
see this metric house how working out
with autopilot two point oh I with the
autonomous hardware but in autopilot one
point I know when my vehicle was
delivered pretty much deal well no
autopilot in cruise control did work
however a lot of the parking features
and other miscellaneous things i had a
pop-up warning seeing these are not
quite ready yet system is still being
calibrated while blog just as example
and at one point this past summer the
system had to be recalibrated now it's
self-calibration system not automatic
calibration system so when my system had
to be recalibrated actually had to have
service go in and they read set and then
they actually took the car and drove it
back and forth on the freeway in some
city streets a number of times to
recalibrate it
solid-state components or in other words
no mechanical moving parts allows the
unit to have a high tolerance for
vibration that's a plus on wisconsin
specifically Milwaukee roads where we
have summer no excuse me you have winter
then we have a pothole season and then
road construction season
the radar also has radome radome heating
available of which my investigation show
Tesla to have chosen based on my thermal
imagery you can see that video under my
account
radon heating allows the unit to melt
off slush ice and snow that is in direct
contact with the radar sensor so that
build-up unfortunately if it melts out
little cavern can only melt what really
what's physically touching it
so then you got a little cavern of slow
snow slush and ice and you can lose
access to the radar anyways which was a
major one of the major reasons they
raise the center of Tesla raised the
sensor up higher
it's about a foot higher and the
refreshed Model S's than it was on the
original styling the radar uses
frequency band of 76 277 gigahertz due
to triple carrier frequency at 77
gigahertz the sensor only needs one
third of the antenna surface of the
other older existing 24 gigahertz
sensors to cover the same field of view
at the same resolution benefits of this
means significantly smaller allows
measuring speed based on the Doppler
effect which results in three times more
accuracy than the older 24 gigahertz
style units
this allows sensor to intervene at an
earlier stage to reduce speed in
critical situations operation the radar
is designed to detect objects and
measure their speed and position
relative to the movement of the vehicle
which it is fitted to do this one the
radar sends out a wave in the 676 277
gigahertz range via transmission and
tennis two waves are reflected by
objects in front of the video
well relative speed and distance between
the vehicle and other objects is
determined based on the Doppler effect
and the delay frequency shifts between
the sent and received signals are
generated and comparing the amplitude
and faces of the radar signal measured
by the for antenna it is possible to
infer the position of the object
what is the Doppler effect the Doppler
effect is an increase or decrease in
frequency of the radar as the source and
observer move to oral way from each
other
this causes a sudden change in pitch as
example I think fire truck coming
towards you with its siren blasting we
then as it's going ok I guess I really
don't need to do the sound effects so
anyways as the fire truck with its iron
blasting comes towards you the pitch
goes up and then as its driving away the
pitch goes down say they happened with
radar elevation antenna can generate an
additional upward elevation being
allowing the sensor to measure height of
all detective objects to help reliably
classify relevant objects and determine
if a vehicle can drive under or over
them special signal processing
algorithms are used this allows the
system to cope with complex traffic
sexual situations and break more
reliably even in the case of stationary
objects predictive emergency braking is
also available if peb determines the
distance of the preceding vehicle is
becoming critically short at a vehicle
speed above 18 miles per hour hence the
18 mile per hour speed been made
mom for traffic where cruise control and
autopilot it will reap it will prepare
the vehicle for potential emergency
braking if the driver does not react
the system uses audible visual warnings
followed by a short jerk of the brakes
this system initiates partial breaking
and if driver also applies breaks it
will calculate the pressure required to
stop and apply additional force to avoid
collision if peb detects unavoidable
collision and operator does not
intervene it can initiate full braking
that test that
thank you Mike Anthony if collision is
unavoidable even with a eb it will do
its best to mitigate the severity of the
collision the sensor is capable of
supporting traffic workers control up to
93 miles per hour 150 kph and intestines
case also autopilot with a max relative
speed of 50 miles per hour or 80 kph and
works in stop-and-go environments if in
case you're not understanding what
relative speed means ok think of it I'm
going down the freeway at 60 mount 60
miles per hour and a vehicle passes me
going 65 so even though i'm traveling 60
he's traveling 65 are relative speed my
speed in relation to the other vehicle
is a five mile per hour difference so
we're traveling five miles per hour
relative or think of it a two-lane
highway country highway i'm driving 50
the person coming from the other
directions driving 50 as we pass each
other are relative speed is actually 100
miles per hour for each traveling then
away from each other at 100 miles per
hour
I don't think I explain that last one
quite right you get the idea
heading distance indicator can warn the
driver if safe following distance is not
being maintained the units also
incorporate sensor data fusion here's
where a whole bunch of magic happens
they're after me lucky charms wrong
serial sorry sensor data fusion combines
the benefits of different sensors and
measuring principles in the most
effective way possible providing data
that individual sensors working on their
own would be unable to generate this
fusion increases the measurement range
reliability and accuracy video sensors
along with software algorithms are data
fuse with the radar in Tesla autopilot
without the camera car cannot do
autopilot without the radar Crocker to
autopilot needs both and the data fused
together in put it by both to be able to
do autopilot and not kill anybody
this fusion also allows for automatic
emergency braking with pedestrian
protection now
Tesla has been updating software in
really starting to activate incorporate
more and more of these features into
their system actually they're all there
it's just now their programming them to
all work in synchronous asynchronous
synchronicity close enough
one example of this is i recently posted
a video showing my car driving down a
road that had no lines autopilot was in
fact active and doing all the steering
and all the hard work it was using a
combination of the radar the autopilot
camera and the 12 sensors sonar sensors
surrounding the vehicle in order to
maintain a radar Elaine in order to
maintain driving in its Lane even
without those lines on the road you
previously to Tesla activating these
features which is don't do that i was
testing things out
don't do that prior to this that would
have not have made possible so Tesla has
really started incorporating not just
this fusion of data between all the
sensors and program programming that
into the autopilot system but this
fusion is also gonna be incorporating
data from the GPS sensor and the fleet
learning and just mashing it altogether
uses as much of that as it can to to
control the vehicle every extra little
bit of data it can
helps so let's just go over Josh
statement of customer fat benefits one
digital beamforming two independent mode
for height measurements three
cost-effective design for tailored
solutions five self-calibration 6 can be
concealed behind the bumper seven no
moving parts
aight sensor data fusion nine small size
10 high speed and flex relay enter can
in flex really interfaces and 1176 277
gigahertz that covers be sense are the
Tesla radar sensor and as long as we got
it i'm gonna have all this in the
description box below so it's easier for
you guys to reference and access and as
long as we got it using all recycled
paper you know
no throw stuff away print on the back
side then recycle it
the technical features now a Boche has
available both a forward-facing and
rear-facing radar device radar sensor i
am only talking to be talking about the
and have only been talking about the
forward-facing I left out for the
rear-facing because as of right now
Tesla does not use the rear-facing radar
not in current autopilot 1 point 0
matin autopilot 2 point 0 as far as what
we're aware even autopilot on the model
3 and model x do not use the rear radar
only front-facing so frequency range 76
277 gigahertz detection range point 3 6
260 meters so there's that hundred and
sixty meters
that box states and i believe i read
that Tesla systems can see seen up to
250 meters so that is a 90 meter
difference now as i just read with the
with the the fusion data fusion sensor
data fusion how it can change the many
jobs uh combines different sensors and
measuring principles and the most
effective way now this is radar only can
see if 260 meters but it's quite
possible if you combine both the radar
and the camera together the autopilot
camera that you could get that further
distance of sight from the radar or from
the system feel the view horizontal main
antenna plus or minus 6 degrees provides
a hundred and 60 meter distance viewing
distance plus or minus 90 degree gives a
100-meter viewing distance and plus or
minus 10 degrees gives 860 meter fuel of
you are viewing distance elevation
antenna plus or minus 25 degrees gives a
36 meter plus or minus forty-two degrees
gives you 12 meter measuring accuracy
distance point 12 meters for speed
accuracy is point 11 meters per second
and angle accuracy is 0.3 degrees plus
or minus object separation capability
distance point seven two metres speed
points66 m/s and angle seven degrees the
cycle time
maximum or the fastest cycle time is 60
milliseconds so I'm going to touch on
that real quick and interject here that
means was my and then the calculator
let's see
so
so according to Bosh the radar sensor to
actually refresh and take a sample every
60 milliseconds that means it's actually
capable of sixteen point six seven
refreshes per second in firmware 8
according to Tesla firmware 8 has bumped
up the refresh rate from the radar 210
refresh rate 10 Ryo 10 times per second
so that's 100 milliseconds so it looks
like Tesla still has a considerable
amount of room to go that they could
increase the Refresh speed is low as 60
milliseconds i have just to interject
again my thoughts on why they may not
have done this on the one point or cards
with firmware eight might be just the
sheer processing power in these systems
these are all tegra3 systems now the
autopilot two-point old cars have
computer systems the Frankie forget the
exact model by Nvidia but they are
approximately 40 times faster than the
autopilot 1.0 and classic models
computer systems that is fast and that's
all for fully autonomous driving so it's
quite possible Tesla might use more of
that data from the radar in the future
and that will add another 6.7 frames for
per second of data of radar data has
frequency modulation FM CW there's
another 1 i'm going to interject done
max number of detected objects is 32
sorry trigger guide to the galaxy fans
it's not 42 apparently the magic numbers
32 now from where seven with autopilot
and I i went back in when firmware it
was released I touched on this now I
always knew and firm or seven that the
radar would see a maximum about 55 other
vehicles on the road at one time for the
release notes of press release of
firmware eight Tesla said they bumped
that up by five fold so 5 10 15 20 25
wasn't six-fold they bumped up 5464 they
said they bumped it up somehow I think
you'll see the five or six fold and
whatever it was my guesstimation was
that a Model S's radar community tech
approximately thirty objects at a time
that was back that was before I knew
anything about this based on previous
calculations and observations with the
car so it looks like i was only two
objects off once again though customized
radar
how could be processing power we don't
know because Tesla will not tell us or
release that information see weight of
the sensors a hundred 90 grams always
practically nothing
and here is the amazing one it does so
so much with only 4.5 watt power
consumption so it is extremely
energy-efficient now i don't know if
that 4.5 what does or does not also
incorporate the power consumption for
the radio meeting my guess is no
eating generally is going to take a
considerable amount of power and yeah
that's quite a bit there so I'm that is
it
I I know this one is the long one
30-minute video here and it was long
probably boring but if you're someone
who was interested in learning about the
radar systems that Tesla was using hope
I was able to teach us something
please don't forget to Like subscribe
below and if you have any questions
comments please put them in the comment
section below
I do my best to answer however i'm up to
like a thousand comments a day on my
videos I can having a real hard time
keeping keeping up
all comments and hopefully some of you
out there and fewer land if you have
experienced or have knowledge or any
kind of smarts dealing with the radars
please help out and respond whenever you
can
that helps me to help you more so that
covers it for this evening
more to come have a happy new year
smile cheese
For more infomation >> Perch Pro 2017 - EPISODE 1 - Kanalgratis.se (with French, German & Dutch subtitles) - Duration: 39:19. 
For more infomation >> BMW wymiana znaczka na masce - Duration: 1:35. 
For more infomation >> New Year's Giveaway! - Duration: 0:44. 
For more infomation >> Best #AbrahamHicks § Better than she planned § Daily #LawofAttraction Video #Reality #Create Posts - Duration: 7:29. 
For more infomation >> Star Wars Rose Parade - Duration: 3:06. 
For more infomation >> HOLOGLASS // Hologram Glasses - Duration: 4:01. 
For more infomation >> Dreams - Duration: 3:56. 
For more infomation >> VTS 01 1 - Duration: 21:33. 
For more infomation >> Funny Cartoons for Kids
For more infomation >> Victor Socaciu - Un veșnic început - Duration: 3:41. 
Không có nhận xét nào:
Đăng nhận xét