Tesla REMOVING Their Ultrasonic Sensors
ฝัง
- เผยแพร่เมื่อ 11 ต.ค. 2022
- Mike Lane weighs in on Tesla's decision to remove ultrasonic sensors from their vehicles, offering pros and cons from safety and cost perspectives.
If you like the content you're seeing, please consider visiting our website or subscribing to our Patreon group for exclusive content.
PATREON GROUP: / munrolive
MUNRO LIVE LINKEDIN: / munro-live
MUNRO MERCH: teespring.com/stores/munro-li...
MUNRO LIVE WEBSITE: munrolive.com/
MUNRO LIVE TWITTER: / live_munro
MUNRO LIVE INSTAGRAM: / munrolive
MUNRO LIVE FACEBOOK: / munroliveteam
MUNROLIVE TIK TOK: / munrolive
MUNRO & ASSOCIATES MAIN WEBSITE: leandesign.com/
Outro:
Music from TH-cam Audio Library [AKA YAL]
Music provided by RFM: • Video
#Tesla #Sensors #EV - บันเทิง
Ok, Mike Lane is now a favorite of mine! Easy to understand, well spoken, broke it down for us and even showed us charts/graphics. More of this please!!
+1. More of this,pls. Visual + Qualitative + Quantitaive + critical. Explained very well. Sorry Mr. Lane, already subscribed ;). Thanks.
Mike says, "What we have here is the front fascia from a Model Y ... " when that's clearly the back one. I do like the new opening graphic though.
Does it mean that if Tesla add a nose camera all bases are covered ? So does it mean all those sensors are useless ?
Well said. I really enjoyed listening to him even with the terrible graphics! 😂 JK. They were fine. I would add that they could easily add a low front facing camera like the one that’s on my f150 integrated into the “ford” emblem and that would eliminate any blind spot and would give total coverage. But that’s a cost to be considered.
...and he even brought his own PowerPoint slides!
One problem with removing ultrasound sensors is that ultrasound doesn't get fooled by fog, rain, bright lights and obstacles that are about the same color as their background the way that cameras do.
ya but they cost money and people will pay for self driving even if it doesnt work anyways so its win win for tesla
@@jebes909090 While ultrasound sensors may cost money, Increasing camera count, resolution, frame rate, dynamic range and processing power enough for FULL self-driving to actually be capable of FULL (unattended) self-driving isn't going to be cheap either.
I bet the current FSD hardware isn't even half as powerful as is going to be necessary to get officially recognized as L3.
Maybe not by rain, but wet ultrasound sensors are a problem, my car (Toyota) does give false alarms when the ultrasound sensors are wet.
@@PeterK6502 I drive a Volvo semi and the sensors dont work worth a shit in weather. They are absolutely useless in snow. We have had 2 trucks get wrecked from the truck panic braking on ice due to the sensors being messed up by heavy snow.
I really think people overestimate the quality of data that comes from USS. They honestly really suck. I know, when I played around in a university project with some Arduino sensors, they weren't the same quality as car sensors but I really overestimated how good they work. They simply don't. They are sooo easy to disturb. Room corners, weird materials, too close or too far distances... I'm honestly surprised how well they seem to work in cars - they probably do a ton of postprocessing to somehow eliminate most of the misreadings (and in narrow but wide enough driveways, they still suck, because they only have a quite wide detection angle and don't give back precise data poins about positions)... And my experience with them is without weather conditions such as precipitation. First of all, I have a huge respect for the people that make them kind of work properly at all. But at the same time, I'm not convinced that USS are the future. They are relatively primitive and so is their output data. However, I'm not sure the way Tesla is choosing is the right one - at least without real camera coverage in front of the nose.
I absolutely hate this change. I use the USS function with the distance call out on a daily basis. Not having that function for accurate proximity would be a major downgrade in user experience.
A camera and computer, let's say at night, trying to calculate the distance of an object from the car.
You're going to have to keep that camera clean.
Someone thought the yoke was a good idea too.
@@djsmithe Can you estimate with your eyes the distance to an object at night? Those cameras past the capabilities of human vision. They should not be compared to phone cameras or DSLRs in the sense that they have to deliver a beautiful picture. When you crank up the ISO on the newer SONY A7 shooters for example, the picture is grainy, but its WAAAY past a human can see.
Oh and many users actually like the yoke.
@@konrad7592 So you tink that this is not just another cost saving from Tesla? LOL. Those cameras have nowhere near the quality you are talking about and guess what, for distance you don't need good quality you need somthing to understand the picture. Tesla may have that tech but based on the self driving scam they are running I would be careful to believe anything they say.
@@konrad7592 The iPhone 14 Pro has a LiDAR sensor. LiDAR is something Tesla doesn't want to use. Too expensive for a Tesla but not for iPhone.
Some auto focus cameras use infrared light to calculate distance and focus the camera len. DSLR cameras take a huge set of pixels and with a computer, trys to make an image sharper, not calculate distance.
Apple's Measure app does a good job measuring distance but it's not perfect.
You are backing up your vehicle. You are on concrete paving stones. Behind you is a brick wall. Will Tesla's computer realize there is a brick wall there or think it's a continuation of the concrete pavers. We all know how accurate Tesla's computer is at identifying objects. (sarcasm)
@@Drkacsabaszk Oh for sure it is a cost saving measure, I never said it wasn't. But what's wrong with that? The best part is no part. Bumper replacements are gonna be cheaper, There's less a future self driving AI has to rely on, the car looks cleaner,...
Those cameras are more than enough to retrieve useful data from. Not pretty data, but useful data. Grab ANY camera, crank up the ISO to the max and you'll be ALWAYS able to tell what's on the picture, regarding necessary objective for traffic situations. It's not gonna look nice at max ISO but it also doesn't need to. And if humans can retrieve useful data from it, a future AI will DEFINITELY be capable of that too.
More of Mike please. He is a smooth knowledgeable operator with a great knack for the camera.
"Ferrets on amphetamines" got me going and so did the blond lady filming with an iphone.
I like Mike's humor, he's a good add.
Thanks
"Ferret on amphetamines."
@@losergamer04 I feel like there must be a story behind that one.
@@losergamer04 best bit in the vid :D
@@MunroLive But can you believe the snake oil sales man? the man who lied and claimed all the roofs had solar tiles WHEN NOT ONE TILE EXISTED! "its really not that hard" some ignorant guy who doesnt know shlt!
I’m liking the variety of presenters on this channel. Mike is no exception. It keeps it interesting. Don’t stop! Who else you got hiding I wonder. I also wonder who is doing your hiring if not Sandy.
That would be Pann
“If you’re gonna criticize my graphics, you’ve got to subscribe. It’s the law.” I love this man.
I’m most disappointed that they didn’t get the camera only features working before they pulled the sensors. This tells me they had a supply side problem and decided to pull them. It was probably in their plan to do it but it’s poor customer focus to roll out the change this way.
Tesla is poor at building customer trust. Their take it or leave it mentality will eventually work against them.
Yeah - a very Tesla move. But the owners easily forgive. I say that having experienced that myself: I bought a Model S just after they dropped MobileEye and switched to an in-house FSD design. Not only was there no self-driving when I got it, but no cruise control of any kind, the windshield wipers were manual fully-on or fully-off (!), and the headlights were completely manual but with no physical controls. This lasted the first 6 months of ownership. Yet I still have the car 6 years later, and I still love it.
we dont know that for sure, what if they just wanted to add $10m profit per year by cutting back on costs?
@@robb412 I really hope so
Mr. Lane is a smooth operator. Loved this piece. Thx !
Our pleasure!
I live in an arctic climate with a lot of snow, rain, cold temperatures and dark days during winter. I'm highly sceptical about Tesla removing the ultrasonic sensors, as cameras tend to be less reliable than the sensors in the wintertime (e.g., dirty, covered with snow). I'd be particularly concerned with backing the car, as snow often get clogged up at the back of the car when driving in snowy conditions.
Maybe the teslabot was made to get out your car and wipe all the blinded cameras before you can reverse into your space🤣🤷♂️
the fix will be to clean your cameras every xxx miles.
FSD today cant even handle mild rain and will automatically turn off.
@@realshompa
it can handle mild rain lmao
@@ogzombieblunt4626 I have experienced, *several times* having Autopilot declare itself unavailable due to inclement weather conditions (don't remember the exact wording of the six-point-font notifiers that appear for a few seconds). And this included a time or two when the rain was so light that the auto wipers didn't even come on. It doesn't happen all the time, but it does happen; I wouldn't be so quick to dismiss @realshompa's observation.
Loved this style of presentation on one single topic. Short, clear, concise.
Acoustic sensors are less affected by dust, fog and snow than visible light sensors, but give less range and spatial resolution. Ideally both types are used to give complimentary data. A buyer paying over $100K is not going to be very impressed that Tesla comprised performance to save $100 in parts.
Mr Sandy, this guy is awesome! Sense of humour betters even yours!
That’s a strong statement. 😂😂
I like Mike. “Don’t like my graphic, you have to like the video to comment - It’s the law” 😂
🤗 THANKS MIKE …we would NEVER criticize your artwork …BECAUSE YOU GIVE US SO MUCH USEFUL INFO easily understood 🤔 and we also THANK THE MUNRO TEAM …for all their hard 😅work 🤗💚💚💚
I’m concerned too Mike. My feeling is leave them in for redundancy if another part of the system fails.
If the cameras fail the car should stop
@@paulmeynell8866 How do you determine the cameras failed ? That's the problem. You can't determine inaccuracies.
@@alejandroperez5368 there is a camera calibration mode that you run .
Then having multiple cameras you can compare images.
It’s unlikely the cameras will fail but you air on the side of caution.
We never worry about the 60 small computers in normal cars or any of the sensors.
Do you drive along worrying the airbag may go off for no reason? The tech is good and very robust.
And humans are always going to make thousands of times more mistakes than a computer!
@@paulmeynell8866
That's handy if you're driving, camera has failed your now stuck on the side of the road in the arms of Tesla repair centre waiting for parts.
@@Robert-cu9bm you can still drive yourself.
And considering the average diesel car has all the same sensors and up to 100 computers how often do you here of them breaking down? It’s a very rare occurrence
The production quality has gone up pretty good.
Thank you so much!
No it hasn't. I was in the showroom last week and the ultrasonic sensor covers were a different shade and the rear door didn't fit . LOL
@@kkal1183 He's talking about the production quality of Munro Live's videos....
@@kkal1183 Then don’t buy one.
@@kkal1183 Imagine caring about the color of a sensor that is no longer there. lol
(that’s the rear fascia)
That was a straight up business presentation. Very concise. I loved it.
Totally agree! Eloquent and simple. I liked his sense of humor as well.
my concern is the fact that camera can get blocked by snow and dust where sonics would not be as impacted
Software monitors the image for dirt or covered camera, if the image is moving but part of it isn't the software warns the driver etc......
@@mohammadwasilliterate8037 warn who in a robotaxi?
@@mohammadwasilliterate8037 Like it does today every time it rains & it intermittent lowers the speed of the car due to limited visibility.....Fantastic isnt it....
Sonics can still be impacted by slush and ice build-up in the winter. (my personal experience)
@@Clearphish Totally right. I live in the nordic countries. Last winter i had 3 days where the radar would get that much slush it wouldnt work. The last 3 weeks after beeing on "Vision", i have multiple times in rainy conditions wondered why it felt like the car was lowering its speed until i checked the screen and "Reduced speed due to limited visibility" showed up. No other cars with cruisecontrol does this and from 3 times a year to every time it rains? Guess if the cams for parking will be the same or not ;-) Its a spreadsheet costsaving world he lives in.
I love this guys dry sense of humor, would love to see him and Sandy do a joint review.
EXACTLY! One time there was a little kid randomly sitting in front of my vehicle and I nearly ran her over because my SUV at the time was so tall. Fortunately, she stood up at the last possible second when she heard the car ENGINE start - imagine if it was a Tesla. Tesla NEEDS NEEDS NEEDS to conceal the new 5MP ultra wide cameras in all 4 head/tail lights if they're going to replace the sensors. This would look better and give customers the added benefit of seeing what's in front while creeping out and seeing more while in reverse. What I mean is, if rear cameras automatically activate to see rear blind spots when we initiate reverse then why don't front cameras automatically activate to see front blind spots when we initiate drive?
Wow Mike was so easy to listen to and concise with his speech. Would love to see him do more breakdown videos
"Ferrets on amphetamines" might be the greatest part-sentence I've ever heard.
Loved the presentation! He took a complex issue and made it simple for us laypersons....and he had flipcharts. I love a good flipchart! There is something more personal about that than computer graphics.
Thank you for laying out the parts cost savings in detail. And due to my education by Sandy I understand the importance of eliminating production steps, and thus costs. But I still have large reservations about relying on a purely vision based system - I want technology that can see *better* than vision.
Finally someone properly explains this case, thank you.
I wish they will increase the USS and give coverage next to the wheels to avoid curb rash. I would pay +$1500 for all the USS + some extra near wheel added, Come on , they do not add noise while driving like radar was supposedly adding , it is for low speed and for parking , why would you get rid of them. They are an essential tool. (P.S. maybe also adding them to the automatic opening doors so you do not "ding" them.
Mike is a great addition to the team
That is a Rear Facia though ;)
Great presenter. My takeaway is that less is more. Less for the customer, more (profit) for the corporation. I’m skeptical about 100% reliance on cameras, particularly under conditions of restricted visibility such as snow, fog, and heavy rain - not to mention those pesky Tasmanian Devils in the driveway.😅
Photon counting.
They have millions of miles of data and realistic simulation. If they remove it, it’s because they don’t need it.
The best part is no part. One less thing to fail, to pay for, to manage.
@@alexforget data and simulation don't mean much if they were made with the sensors there.
Very professional and informative presentation Mr. Lane, well done thanks!
One of my favorite members from Munro Team so far. Looks like the very cool and prepared teacher every engineering students would like to have
great video & loved the slides. Mike is now in my top 2 fave presenters along with Sandy himself!
Love the presentation Mike. Great analysis by the team at Munro.
Much appreciated!
More of this guy, please! Great job, Mike!
Great video! Very clear and easy to understand. I like how you educated without taking a side. Nice!
Good job Mike. The Munro team is really something.
James Douma said the cameras will fix an object (ie parking block or curb) at a distance and then software will use wheel rotation to know precisely where the object is relative to the car.
That is pretty much how the cerebellum works to keep people walking.
Seems like the Camera need to be more aware while car is "resting" now ala centry mode, incase a bugger have put a concrete block infront of the car during the night! :)
Impossible! Using software to map the world? Can't be done. Lmao
@@jonasgranlund4427 ultrasonics wouldn't always pick that up either though. Depends on height and distance away. I suspect that telsa will get a system that is as good in at least 99% of cases and that is worth the x million per year.
What if traction is not 100%? The wheels could slide but the car wouldn't know. I would also be concerned about tight parallel parking where the car might never see what's in those 3 feet of blind spot.
Mike is awesome. I like his tone, demeanour and communication in layman's term
Thanks Sandy for Mike Lane's detailed analysis of the ultrasonic sensors and what costs would be associated with removing the sensors and ancillary wires etc. and modifying the software to use 5 Megapixel cameras.
Worked in engineering for 33 years and never had one Product Manager tell me they were happy with the cost of the connectors in our product. :) It was always the first thing to get addressed during the follow-up cost reduction project.
After tying to find ipx7, x8 connectors for a motorbike project i understand, $40 just for a 15 pin d-sub PCB mount connector. Its expensive. The labor cost in large scale manufacturing must be huge.
Buy cheap connectors and you get exactly what you paid for.
Pennywise, pound foolish.
Don’t even talk to me about the Porsche Panamera Chassis ECU connector, it is the worst connector, with the worst terminals and the worst terminal extractor tool ever conceived, even worse, every manufacturer has succeeded the part because it is so shit which means the fragile extraction tool is now unattainable, short of having a company custom machine a 0.1mm precision hardened steel tool.
@@gtjack9 you had me at German connector design. :)
Well done, Mike! You made it so accessible to all of us to understand. Great graphics too!
yes! easy break down
Enjoyed the presentation.
Perfect Presentation Mike. Thanks
Thanks, Robert!
Excellent job, Mike. Very informative and my kind of dry humor.
This is great info and cost analysis that really consumers should know about what they are paying for. Much appreciate Munro Team!
In this case the customer gets screwed. Do you really believe they will reduce the price of the car?
@David Johnson This is such a cold take. It’s spicier to believe computer vision has gotten better and doesn’t need USS. Reduce processing/production, reduce waste.
@@taylor92493 No, what that means is that they removed sensors that actually provided some functionality. They could detect small things, down low, in the blind spot that neither cameras or humans can see. That way, your cat, your dog, or your child doesn't get run over. As for the full self driving, in my opinion....
Vaporware that will never work! By that I mean you will NEVER be able to take your hands off the wheel and ignore what is going on around you. Don't believe everything your leader tells you.
@@taylor92493 This is correct.. most people wont bother watching or wont understand AI day, but what was presented sure sounded like they made these parts unnessesary.
@@patreekotime4578 I watched AI day and that joke of a robot they brought out.
1) The point here is that these cameras could detect small objects in the blind spot of the cameras. No amount of software will cover that.
2) In my opinion FSD is a scam because the customers who paid $15,000 will likely never see the advertised benefits! At least the removed sensors were good at detecting a pet or small child. They removed the only component of FSD that had any real useful value!
I now need to subscribe... this is the way.
My concern is failure modes. If one of the cameras gets covered up or experiences solar glare, that’s a large portion of the vehicle that’s effectively driving blind
This is about parking, not driving. At higher speeds, the USS are useless anyways.
Thank you for the technical review. Best thing I did was subscribe when Sandy first took apart the model 3. Keep up the great videos.
Thanks for watching!
Wow experts for each segment! What a Chanel and what a company. Glad you gave us all a look inside your minds
Another solid tech summary from another superb presenter, Sandy's team has impressive depth!
Definitely curious about that blind spot. It can probably memorize a curb position before approaching just fine, the problem as mentioned is something running in front of it once already in its blind spot. Maybe it's not an issue 99.9% of the time, but is that enough for true robotaxi where they want to get many many 9's over.
Ever since this announcement I've been paying close attention to what my cameras can see when I back up. There are also very large blind spots on the rear corners of the car that cannot be seen by either the side repeaters or the rear camera. In fact, the way I pull into my garage, there is a box on my floor that is NEVER see by those rear-facing cameras.
Not really. If the new obstacle came by side it will be picked by side cameras. If the new obstacle came by front it will be picked by the front camera. The only problem will be if the new obstacle raises from ground.
And piezoelectric sensors do not detect all obstacles ... we never ever see some damages because some rounded obstacles that diverge acoustics echos. For another point the vision could fail if the obstacle is a glass wall...
lets say the car is 'off', then an object moves into the front blind spot, then the car is turned 'on' (cameras become active) now does the car have no sense that the object is there? to truly always be aware of every blind spot the cameras would have to always be on
@@bpeterson4016 Sentry Mode already exists. The car can already detect motion near it and record that motion, and it doesnt have a huge negative impact on battery life. I would guess that if the car sits for too long it would go into a true sleep mode and after that period the owner would have to drive it out of park. But if you are driving every day or every other day, theres absolutely no reason the car cant be watching 24/7 and be aware of the environment around it the whole time.
very well done.
I wouldn't be surprised if this has to do with the global supply shortage, and or the desire to cut costs.. I don't believe that they do this only because they think that the camera solution will be better. Cameras can be fooled...
They tried replacing a standard rain sensor with a camera-based system years ago and it still doesn't work properly. It only works somewhat satisfactorily during optimal lighting conditions.
At night it's almost completely useless. Seriously, the most used function on my Model S is probably the windshield wiper button...
I also prefer the adaptive cruise control on my 2012 model Audi Q5. It doesn't phantom brake like the traffic-aware cruise control on my Tesla.. I don't dare to use it if I have anyone behind me, because it will slam on the brakes when you least expect it to. It's also extremely uncomfortable for the passengers who get very scared every time it happens.
I also believe that there should always be some redundancy. In an airplane, there's always more than one system to ensure safe operation. While in a car that is supposed to be able to drive itself in the near future, your life would be in its hands. Shouldn't there be at least SOME redundancy and not only cameras? I live in Norway and during winter the cameras are constantly blinded either by snow, fog, or the sun itself since it will be basically right in front of the car at certain times of the year here up north..
I don't think there's much redundancy. Flying is definitely not similar to driving
@@Spawn223311 Flying and driving is similar in one key way though: If you crash, you could die. Imagine going 100 km/h in your car, which is driving itself. Then the camera fails to see a parked semi truck in the middle of the road, and you slam into it. Now if it had radar or lidar as a redundancy, it would have seen it. Are you OK with taking that risk just to save costs?
The problem with camera-vision is that it can't understand anything it hasn't been trained to see before. under certain light conditions and such, it can easily get confused. There are many examples of this where cars have tried to steer off the road because of shadows on the road being misinterpreted.
@@Spawn223311 Yeah, driving is MORE dangerous.
I’ve worked for the largest legacy automaker for the past 24 years in QC, but I’m all in on TESLA, that should speak volumes. Thanks for the great video. Cheers.
I’d start by saying that the camera situation could be extremely easily mitigated by having a camera mounted centrally on the front bumper, akin to VAG “Top View” systems that use a forward facing camera in the grill. This would cover the extremely short range whilst also allowing for a side view when emerging from a junction with very limited visibility to the sides.
It could also replace their wide field camera mounted next to the monocam in the windscreen.
It’s easily solvable, it’s whether Tesla are willing to take this risk and account for the lack of short range data in some other way.
I was thinking the same thing.
But why? Camera would cost same as the sensors, they would need harnesses, wires etc. Would need some kind of washing system to clean the camera. All of this just increases the cost, while they are trying to save on everything.
Makes no sense, so it will not happen.
You can get a Job@Tesla and make this your pet project there. Lets see what happens to your idea!
@@tomasvidziunas5543 Because the cameras can do this anyway and Tesla no longer has to develop and support 2 systems.
With the Cyber Truck and Tesla Semi, the blind spots in the front will only get bigger, so I think another camera with a 210° lens in the front bumber is the best solution. Which would also need another port on the FSD computer and more processing power. So they need to switch to HW4 for this.
Good thing I subscribed to Munro way back. I wouldn't want to break the TH-cam law!
LIT animation at the start of the video!
Great to see another colleague of your team. Great video.
Glad you enjoyed it!
Well done Mike! Great video
Thanks 👍
Very clear and to the point. Enjoyed talk.
Great presentation. Easy to understand. Thank you
Glad it was helpful!
The key here is that the software MUST be able to track and REMEMBER what it has seen that is now hidden, and make alerts / plans based on this. This is significantly more difficult than just using a stereo pair of cameras to calculate depth in a visible area. It also assumes that whatever those objects are - that they do not move or change shape while hidden.
Imagine a child in a snake suit creeping below the camera in its blind spot before the cars front bumper and laying doggo and waiting for the innocent owner to start his happy day in the factory! And all because of missing Ultrasound sensors!
@@wolfgangpreier9160 They could also just add 1 new camera down there. Ultrasonic sensors just tell you theres something there not what it is, so a camera in its position would be better when it comes to running over babies
Clear, factual, analysis unclouded by emotion. Thank you. Interesting point about Tesla knowing the implications already by using the mountains of captured real world data to answer the what if questions everyone tends to think up. I never had those USS systems on the majority of the cars in my life and the only one I’ve scraped on a wall in the blind spot was one of the ones with them! Go figure.
Thanks Mike. Concise and extremely relevant points.
Your graphics were great!
they really need to add low front bumper mounted cameras left, right and center where those sensors were not only to give a better picture of close objects during parking but to provide blind corner views during unprotected turn crossings while driving. this would allow the car to not have to creep so far forward while trying to assess cross traffic. The pillar camera position is really too far back to make for good cross traffic detection.
@@JT_771 if cars are going to auto start and drive, and they have no sensor to tell if a child is lying in front of the rear wheel, this is going to be disastrous. The camera can not see enough in front of the car and can’t do a walk around like a human.
Two wide angle cameras in the headlamps would give depth perception too?
A front bumper cam at least could help giving 360o birds-eye view feature, to avoid high curbs/parking blocks in front before you scrape the front bottom lip on it :( USS didn't reliably detect those either, so adding one low cam might actually be an improvement... and would of course cover up the triangular hood blind spot he mentioned..
@@TwentyHz could you please tell me how sentry mode works and then revisit your statement?
@@clinton8432 I think that sentry mode can be disabled to prevent vampire battery drain when camping or away on a long vacation etc.
One more camera in the middle of the Tesla logo would perhaps eliminate the dead spot. I would personally also like to see the logos on all cars go black, that would make an extra camera blend right in. Plus, it’d be kind of cool, like the car has an eye.
Keep it up Munro Live, I’m really enjoying this content. And I like Mike, him and Sandy should do a video together some time.
I want a red light band from left to right and it should make aswosh sound. I love my cylons!
Nice work. I will respect the law!
On my now 8 year old Ford Taurus, I have rear ultrasonic sensors which have never failed to warn me of "bad things about to happen" while the rear facing back up camera, if not wiped off every time I go past the rear of the car looks like a view through an unwashed beer bottle. When clean it is fantastic. Living in Western NY, with rain, snow, snizzle {snow and drizzle}, salt, the camera is a dirt magnet.
I'm a bit worried about this change. Last week in the dark on a busy freeway I nearly changed lane into a black car that had its lights off. I couldn't see it over my shoulder or in my blind spot mirror, and Tesla's AI didn't show it on the display either. But the collision avoidance alert triggered when I was just about to hit it - and I'm pretty sure the ultrasonic sensors are what detected it - it showed as a red wall proximity alert on the UI. Tesla is very forward thinking: they often release products that rely on technology that doesn't exist yet!
Mr. Lane is the man!
Mike Lane is awesome! Great graphics and presentation!
Graphic were perfect! Keep up the well reasoned content!!!
I can see this working with persistence but if the car is off and something is moved into that location then there may be an issue.
This change leverages Sentry Mode which has the cameras on all the time already.
A 180 degree wide angle cam in the front of the hood wood be good. No blindspot and better visibility in intersections.
Wow, really enjoyed Mike's informative and engaging analysis!
Graphics were AWESOME 😆 I couldn't do that 👍🏾
Also electrical savings are made by removing sensors, which is always a plus in an EV.
In power ? It'll be so minimal as to be unimportant. A watt maybe ?
That wiring harness and the connectors look pretty robust, but I don’t know?
Back in September Elon mentioned there was a priority rush to get Smart Summon etc. updated to the latest FSD beta. So it looks like they were trying to get it ready before the USS removal but didn't make it.
With Vision-based park assist, the car would not only be aware of physical objects but could also account for things like parking lines on the ground, available door clearance, or maybe even identify fire hydrants and no parking signs as spaces it isn't allowed to park. Smart summon could then be used to tell the car "go find a parking spot".
As for the blind spot, it'd be pretty difficult for something to get into that area without 1 of the cameras seeing and tracking it before it gets into that space.
Yep, it seems that they'd be building a map that stays in memory to be able to account for blind spots. It may have to make some wild assumptions for obstacles in front though.
@@stevemulcahy5014 The biggest problem is that you either have to turn off the graphics processing of the forward facing, wide field camera and store that map and assume that nothing will enter that blind spot while turned off, or you have to waste battery power running graphics processing constantly, even while locked and parked so that obstacles that enter the blind spots can be tracked in and out.
It’s a risky cost saving exercise imo, especially given all they need is a VAG style grille mounted front facing camera which are extremely inexpensive.
@@gtjack9 Your camera's are always active if you have the Sentry mode active. So this would have no impact.
Only 8 years since Dear Leader Musk said that FSS is here. Image all Tesla owners that have paid thousands and thousands of dollars for FSD and it is not even near being released. The latest beta still needs interventions at least once every 5 minutes. This can be one of the most extensive product scams in history.
@@pipooh1 Thats a lot of lost energy & thereby extra expenses for the customer during the lifetime of the car.
I love your Graphics❤👍😃
Nice review! Thx
Great info and presentation 👍
Glad it was helpful!
Very well done! I Like Mike
We do too!
@@MunroLive If we on the outside were truly smart we would send the lot of you Munro'ies into Government and fix this whole mess were in.
well done, short and concise, just what we need.
Nice analysis!👌🏼
Is he the Christopher Walken of the car world? Smooth.
Came to the comments to find this! Definitely looked like Christopher Walken and had a very similar delivery style! 😆
In terms of the gap.... I wonder if tesla vision, utilising prerecorded footage like on sentry cam, captures the last couple of seconds of footage and premaps obstacles in the 1ft x 3 ft blind spot at the front?
Yes. Another youtuber suggested once cameras determine distance, wheel rotation will calculate distance as they roll forward with great accuracy.
What if an object appears in the blind spot after the car is turned off ?
That would help but not for every scenario. However, Tesla's engineers are not stupid; they are certainly aware of those issues and have thought of a solution.
Great video and research. Thanks!
Glad you enjoyed it!
Mike you are great! I would like to watch you more!
James Dauma explained the solution is: the cameras detect something 3 feet ahead, and then count how many wheel rotations it has left before it touches it.
problem when you are getting out of park
@@sathwickborugadda4744 this
I am guessing they plan on modifying the front and rear facia molds to remove the sensor holes so it is possible they may add a new hole and place a camera somewhere along the front facia to capture the three feet and beyond and give Vision even more depth. As always with Tesla ya never know until they show it off.
This is my guess as well. There’s still edge cases that the current camera setup will have no way of picking up. Say you back into a parking spot and come back some time later. Car would have no way of knowing whats directly in front of it in that blind spot. If fsd is ever going to be realized need to solve that problem
This would be good because maybe then we could get 360 Birdseye view! They really need this feature, I have it on far less advanced cars and its great. From what I've read and seem to believe they can't do it with current camera positioning. Mainly they are missing a camera way way up front.
Well done video! I wish all class can be this informative & entertaining!
Glad you enjoyed it!
He explaines really well. To the point and clear
I think the ultimate goals of Tesla vision is to be the only sensor on the car. In theory, it's possible because NN has crazy scaling potential , but I don't know if Tesla can do it safely and comfortably. We'll have to wait and see.
Vision has to be the end state. After all, that's all humans use to drive cars.
@@skenzyme81 people have also been using ultrasonic sensors to supplement said vision…
Mike is awesome. Keep em coming. I cannot wait to see how Tesla will solve this! Their park assist features are long over due for an overhaul.
This one may determine whether I stay with Tesla or move to the BMW i4 😅
Nothing could push me to an i4 after seeing the interior looked over on Munro Live. Really the only things I'd trade my Y for would be an R1S, maybe R1T, which seem to be actual steps up in interior quality while still being capable from the infotainment angle. I just can't give business to the dealership model.
This issue is already solved. Explanation was in last years and this years "AI Day". Occupancy Network ( See: th-cam.com/video/fCsfLCc4UzQ/w-d-xo.html ) Just needs fine tuning which they must be close to completing.
Improved paint! I could see the great Camera person’s reflection! Always great stuff and food for thought!
I hope aspiring engineers learn as much as I do.
Great presentation Mike! Thanks for putting up the camera distance drawing. It is good to see actual numbers assigned rather than just hearing "cameras won't do what the USS does" which is too broad of a statement.
I suspect their voxel mapping capabilities, combined with vehicle vector information, will allow for rather precise mapping of geometries within the blind spots. This should be especially true during low speed maneuvers such as parking. Any moving object entering the blind spot (dog, person, etc.) will enter from a perimeter area, thereby allowing detection from the vision system. The software could then predict the object's path/position within the blind spot or perhaps just wait for the object to pass through the blind spot and re-enter a camera's field of view (just as humans do) before continuing. Evil clowns popping up from storm drains while inside the blind spot will not be accounted for. But that's okay because evil clowns have it coming.
Good analysis
But the human can walk around the car to check if there's anything in blind spot, for parked car , unless it keeps FSD running, it will miss any object or person in blind area, like if someone is playing hide and seek & hide in front of Tesla in parking lot.
I be curious to know if they're adding new radar sensor in front or maybe camera in front like in the back to give 100% coverage 🤔
Except non-moving items (such as a rolling stool in a garage) would not be detected by the cameras, but are identified by the ultrasonic sensors. I frequently use summon to move my S forward and backwards in my garage when I'm cleaning it. I once forgot to move my rolling stool and I could not figure out why the car would not move back. I went to investigate and found the stool, not the car, was the culprit. Sad to see Tesla abandon this feature. I'll hang onto my AP1 car as long as I can.
I was about to post about the same thing. They are already building pointcloud 3d world around the car all the time anyway. They do not have to see the curb, they know it's there before hand and how close the car has moved to it. Because it's on the active 3D model that was built when approaching the curb or any stationary object.
The nearby resolution can be increased while moving at parking speed by not building far away model at that time.
@@gelisob You can’t assume what’s in a blind spot when the car starts up in a static situation.
When in motion you’re absolutely correct, the camera can keep track of obstacles that enter a blind spot and act accordingly but the implementation of summon isn’t possible based on the information provided on this video.
You can’t assume that the area in front of the bumper is clear when summoning from a garage, otherwise the camera would need to be carrying out analysis constantly, even while parked and locked. From a power utilisation point of view, it isn’t feasible to run a high power graphics processor constantly without greatly reducing range.
Mike looks like a good older real engineer