Tesla Plans To Launch a Robotaxi Network In 2020 (techcrunch.com) 205
A reader shares a report from TechCrunch: Tesla expects to launch the first robotaxis as part of broader vision for an autonomous ride-sharing network in 2020, CEO Elon Musk said during the company's Autonomy Day. "I feel very confident predicting that there will be autonomous robotaxis from Tesla next year -- not in all jurisdictions because we won't have regulatory approval everywhere" Musk said without detailing what regulations he was referring to. He added that he is confident the company will have regulatory approval somewhere next year. Tesla will enable owners to add their properly equipped vehicles to its own ride-sharing app, which will have a similar business model to Uber or Airbnb. Tesla will take 25 percent to 30 percent of the revenue from those rides, Musk said. In places where there aren't enough people to share their cars, Tesla would provide a dedicated fleet of robotaxis. Musk also said at the event that all new Tesla vehicles are now produced with its custom full self-driving computer chip. The remaining step for Tesla's full self-driving mode to work is the software, "which Musk says will be 'feature complete' and at a reliability level that we could consider that no one needs to pay attention, by the middle of next year," reports TechCrunch.
Great... (Score:4, Funny)
Re: (Score:2)
April 1 (Score:5, Funny)
Even his April Fool's jokes are 3 weeks behind schedule.
Quite (Score:3, Insightful)
The suggestion that fully autonomous driving that doesn't occasionally headbutt the scenery or a parked truck will be with us in a year is just a joke. Presumably Musk has been smoking one of his "special" cigarettes again.
Re: (Score:2)
Also foolish: letting your personal Tesla be used as a taxi.
First rule of the pizza delivery biz: if you do it with an expensive car it costs you more than you earn. Deliver pizza with a cheap beater.
You won't be laughing when... (Score:2)
... Some prankster sends all of Elon's million taxi army to your house to pick you up. A physical DDOS creating a gridlock so bad it will take days before the last taxu is cleared from your driveway.
Re: (Score:2)
Fuck you and the horse you rode in on.
Baby steps (Score:3, Interesting)
Re: (Score:3, Insightful)
In the past 3 or so years we've gone from pipe dream to Tesla's happily cruising down highways. This field is developing faster than computers in the 90s.
This is Slashdot. Maybe you would prefer another site: https://www.luddites200.org.uk... [luddites200.org.uk]
Re:Baby steps (Score:5, Insightful)
It does work on all roads and all weather. (Score:5, Informative)
It doesn't work on all roads, it doesn't work in all weather, and it needs a safety driver.
If you watched the event video you'd know that Tesla's approach, with the sensors they have, does work on all roads and in all weather.
Indeed one of the problems they have with LIDAR only solutions is exactly that - t does't work in all weather, other sensors are important besides just visual.
They are very, very close to being able to not have any driver, and the way they train they have large amounts of real-world performance data to show regulators that they can drive as well as, or better than any human - even from cars that have not turned on Autopilot, it's still running the models for self driving and evaluating deviations from actual driving.
Tesla's approach of having a vast fleet of cars contributing to real world training is in fact the only way self driving cars will be able to work on all roads, and all weather... I don't think the camera/LIDAR debate is as clear as he makes it but there's some visual data LIDAR does not give you so he might be right there.
Re:It does work on all roads and all weather. (Score:5, Insightful)
Almost having full self-driving is like almost reaching orbit: eventually it'll crash and burn.
Re: (Score:2)
I agree that their approach does seem more logical, especially their ability to create learning data. But my problem is that just because they have a system that can get even 90% of the way there, it does not follow that all they need to do is more training for corner cases and they will reach 100%. DeepMind did not create AlphaGo by simply writing one system and then feeding it ever more training data. They had to continually iterate and refine the approach itself. From the presentation it appears Tesla is
Re: (Score:2)
This is untrue. The Tesla system doesn't work if it can't see road markings due to say snow, unless it also has another car to follow in front of it. Clearly that's not a viable solution for a self driving taxi.
If Tesla was anywhere near where they claim to be then they would be inviting journalists to film their own journeys on routes they select, not limiting the demonstration to a carefully pre-prepared run in ideal conditions.
Re: (Score:2)
Re: (Score:2)
You don't seem to understand how Tesla's self-driving AI works; it mimics the behavior of human drivers, based on the same visual cues we use: wheel ruts left by previous traffic; shoulder geometry (slope, ditch, guardrail, etc); mile markers and other signage; light poles; etc... and in most cases we drive at reduced speed to increase our margin of safety.
Of course, a smart robotaxi fleet operator would suspend service during hazardous weather conditions, even if they have a robust solution that performs b
Re: (Score:2)
Also I think it is a stretch to think that the taxi service would stop operating. I think they would likely surge price the fares and keep going as less people want to drive, then claim that it was the fault of the weather when there is an accident. For them to stop running would require them to have some sort of compassion.
Re: (Score:2)
Re: (Score:2)
If you watched the event video you'd know that Tesla's approach, with the sensors they have, does work on all roads and in all weather.
Even if their approach works, their implementation doesn't.
they have large amounts of real-world performance data to show regulators that they can drive as well as, or better than any human
Ok, lets enter them into the World Rally Championship. Should be an easy win, according to you.
They are very, very close to being able to not have any driver
My six year old Mercedes can do that. I mean, it's only competent at holding position when parked, but that's closer to a Tesla's capabilities than they are to not having any driver.
Re: (Score:2)
A lot of cars will happily cruise down highways pretty close to being self-driven thanks to distance-sensing cruise control, lane assist and automatic lane centering.
Yet I've had people tell me even these things are so totally dangerous and that self-drive will never happen. I honestly think they've just never driven with them and just don't understand it.
Re: (Score:2)
Re: (Score:2)
Have you ever tried driving in such a place?
The lane detection in my car isn't anything special, but it can see the lines pretty much as well as I can. It's also okay at following the tracks of previous vehicles even though I doubt very much it was designed with that in mind. It wouldn't be hard to train a more sophisticated system to drive slowly and judge the location of the road based on side of the road markers and ditches.
And if you don't have any of those things, the human is just guessing too. I have
Re: (Score:2)
The lane detection in my car isn't anything special, but it can see the lines pretty much as well as I can.
That's the problem, in order to see them everywhere they need to see them better than you can. After all when it comes to self driving, the human is the problem remember?
Tesla already does just this (Score:2)
Yes, you drive split across two lanes where all the other cars are also driving. Gives you more room to slip and slide without hitting anyone and it's the only place the snow has melted. No autonomous car is designed to handle something like this
If you watch the video, Tesla is doing exactly that already.
The key is the train from all sorts of real world video of real drivers driving a Tesla in snow, including when lanes are covered - they train based on where drivers actually drive, and also human annotati
Re:Baby steps (Score:5, Informative)
Let's be realistic what Tesla have actually done so far.
They have some nice driver aids. The driver is required to pay attention at all times. There are many situations it cannot handle. Tesla has not been able to demonstrate any level 5 autonomy.
The recent demo was staged. Carefully selected route, unknown number of attempts to make it work. That is a very, very long way away from arbitrary routes with no human safety driver.
The idea that they are going to reach full level 5 by next year is laughable. It also lacks credibility as they have made these claims before in 2015 and 2016. Even the recent demo was basically the same as they showed in 2015, so how much progress have they really made in the past 4 years?
Re: (Score:2)
This is not level 5.
Level 5 is all conditions, no matter what.
This is NOT what Tesla is talking about - it's not what Waymo is doing either.
As long as the car is safe, it's okay if it needs help in a few cases.
Re: (Score:2)
As long as the car is safe, it's okay if it needs help in a few cases.
So the "robotaxi" will have a human safety driver ?
Re: (Score:2)
Re: (Score:2)
Could be a remote driver.
How many vehicles will this remote driver be responsible for? If it's more than one why wouldn't they just be in the vehicle anyway?
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
They were talking about a remote safety driver.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Even level 4 is way beyond what Tesla is capable of, or likely to be capable by next year.
Re: (Score:2)
I am also in the camp of thinking it will be many years before we see full autonomous driving, but your comment is contradictory. You state that, "Tesla has not been able to demonstrate any level 5 autonomy." But then you go on to criticize the level 5 demonstration Tesla provided. Even if all of your concerns with the demonstration are accurate, and I believe they are, the demonstration is still a compelling example of progress on level 5 autonomous driving.
Re: (Score:2)
The demo was not level 5. They had a safety driver, so all they needed was level 2. It wasn't an arbitrary route either, it was pre-selected, so that limits it to level 4 max.
The best anyone has demonstrated is Waymo with level 4. They have shown cars with no-one in the driver's seat, navigating around the limited area they confined it to. So Tesla's demo isn't even as advanced as what Waymo showed off last year.
Re: (Score:2)
"The driver is required to pay attention at all times."
This is, of course, crazy. I'll be surprised if there are as many as 3 humans in 1000 who can actually remain constantly alert for extended periods of time while monitoring an autonomous vehicle. Even if their life depends on doing so. Which it might. Further, except at very low (parking lot) velocities, I doubt that even one of those hypothetical constantly alert driver will always have enough time to figure out that the vehicle is about to do some
Re: (Score:2)
For Reporting Year 2018, Tesla did not test any vehicles on public roads in California in autonomous mode or operate any autonomous vehicles, as defined by California law. As such, the Company did not experience any autonomous mode disengagements as part of the Autonomous Vehicle Tester Program in California.
Sounds like they are a *very* long way away...
Re: (Score:2)
Re: (Score:2)
In the past 3 or so years we've gone from pipe dream to Tesla's happily cruising down highways. This field is developing faster than computers in the 90s.
This is Slashdot. Maybe you would prefer another site: https://www.luddites200.org.uk... [luddites200.org.uk]
Cool, you first then?
Re:Baby steps (Score:4, Insightful)
Re: (Score:2)
Wow, this may well be a very apt comparison. After the 90s so many computer companies were either defunct or on life support. So, your prediction that Tesla may be gone soon may become a reality.
By the way, self-driving cars have been around since the 80s. (https://www.politico.eu/article/delf-driving-car-born-1986-ernst-dickmanns-mercedes/) So, after more than 30 years I would have expected self-driving cars that are more capable than what Tesla is offering.
Re: (Score:3)
Re: (Score:2)
Here's a similar video from 2016
https://www.youtube.com/watch?... [youtube.com]
Re: (Score:2)
And here's a similar video from 2009 [youtube.com].
No Subject (Score:5, Funny)
If he doesn't facetiously call it JohnnyCab I'll be disappointed.
Lessons from Boeing (Score:4, Interesting)
I hope the software developers being pushed to do this take a lesson from the 737 MAX fiasco. The reality is that despite all the other folk involved in that plane's design, somewhere a handful of embedded software devs will have actually written the MCAS code, and if they did not protect their liability well enough (documenting the scope of their involvement in the design, objecting to the implementation being requested) they are going to be thrown under a bus by management once lawsuits and possible criminal investigations start. Remember how quickly VW blamed the diesel issue on a 'rogue' developer.
I think Tesla has done great things, but on driverless (and anything AI in general) Musk seems to have totally unrealistic ideas (the sentient AI that is going to take over blah blah blah). He also has a terrible track record, having promised full self driving on the previous hardware, and now this new hardware is supposed to be it (if a version of the software is not demonstratable yet, how can he even know what hardware will be required?). Personally I think he runs a bit of a hype ponzi scheme, but has slipped so far behind his achievements that his claims are becoming ever more detacted from reality. This is dangerous stuff.
Anyway all the best to them, but again I just hope no young ignorant devs get their lives ruined because of this guy's hubris.
Re: (Score:2)
I hope they don't "learn" from Boeing. For self driving cars perfect should not be the enemy of good. That is a very different story for the airline industry.
Re: (Score:2)
Re: (Score:2)
Re:Lessons from Boeing (Score:5, Interesting)
Musk uses the same technique as other bad bosses in high pressure environments. Set ridiculous deadlines and make public announcements, to try to force engineers to deliver. Steve Jobs was very much the same, and to be fair it does sometimes work. In some areas Tesla, Space X and Apple all exceeded expectations.
But it also leads to burn-out and good engineers leaving when they realize they are being asked for the impossible. Tesla's autonomy division has suffered greatly from talent leaving over the years. And while the worst that happens at Apple is an expensive screw-up, with self driving cars people are going to get killed. People have been killed.
Tesla has a great system here - shadow models (Score:3)
I think Tesla has done great things, but on driverless (and anything AI in general) Musk seems to have totally unrealistic ideas
I am sure that Musk has some unrealistic ideas, but self driving is not one of them.
Something the summary didn't mention but that I got from the video of the event is that Tesla cars with self driving computers can run side by side shadow models of potential updates. So Tesla has real world feedback about choices the various models involved in driving would have made in real worl
Re: (Score:2)
They just demonstrated the ability last week of being able to recognize traffic lights. Traffic lights! The cars still don't recognize street signs. Audi and Volvo production cars have had this ability for years. But they are going to be FSD running "shadow mode" simulations next year? Right. He definitely caught the tech bros hook-line-and sinker though. You guys are drooling over it and your imagination is running wild.
Re: (Score:2)
They just demonstrated the ability last week of being able to recognize traffic lights. Traffic lights!
Like Musk said, people have trouble understanding what exponential improvement really means. Your statement is a case in point.
The cars still don't recognize street signs.
Your source for that? Presumably they do in that same example car that drove the route with lights.
But they are going to be FSD running "shadow mode" simulations next year?
I agree that next year is over-optimisitc, and Musk even pref
Re: (Score:2)
Something the summary didn't mention but that I got from the video of the event is that Tesla cars with self driving computers can run side by side shadow models of potential updates.
I think this is an under-reported facet of the presentation. Definitely under appreciated.
Specifically, they have the user fleet reporting constantly back to the mother ship. They accumulate hundreds of thousands of incidents every month about what the driver-less car would have done in all those real-time situations had it been in control.
Real. World. Data. All over the world. It is as if they are already selling Class 5 automated cars except that when an "accident" happens there is no price to pa
Re: (Score:2)
What you say was the initial assumption - that the pilots did not switch off the elevator trim motor. But the analysis of the black box showed that they did. However, once they switched to manual trim, they did not have enough authority to adjust the elevator, so they switched the motor back on, which again enabled MCAS.
Re: (Score:2)
So why is the entire fleet grounded?
Something doesn't add up here.
Same plan Elon told us 3 years ago (Score:3, Interesting)
Elon said the very same thing 3 years ago, in 2016 when he unveiled AutoPilot 2.0. All cars have everything they need to drive themselves, software coming soon. Tesla ride sharing network coming next year. Even a demo video that is pretty close to the one they released now was available in 2016. None of that came true. Definition of insanity (doing the same thing and expecting a different result)?
Re: (Score:3)
Re: (Score:2)
I think it's the definition of a con man.. he said it last time and they paid him, so he will say it again and again and again.
Well, considering he seems to be this way also when there's no obvious benefit I think most of all he's drinking his own kool-aid. The solution is just around the corner, the finish line over the next hill, why is everything taking so long why aren't we on Mars by now. I mean there was probably a lot of plausible naysayers Paypal couldn't take on the banking industry, Tesla couldn't take on the car industry, SpaceX couldn't take on ULA and so on.
I think he's reached the point where he just walks down the li
Re: (Score:2)
Musk on 15 January, 2015 [youtube.com]:
"I really think we'll have mainstream cars capable of full autonomy in five years or less. But, proving that it's safe and getting the regulatory approval is likely to take 2-3 years after that."
Let's do the math. 15 January, 2015, plus five years = 15 January 2020. "Mainstream cars capable of full autonomy" by 15 January 2
Re: (Score:2)
"He added that he is confident the company will have regulatory approval somewhere next year. "
Next year, not 2022-2023
Re: (Score:2)
Yes, if anything their timeline has become more aggressive. The GP is trying to argue that they've been delayed.
Tesla will launch a robo-taxi service. (Score:3)
People will still be saying, "cant spell felon without elon" and pat their own back for their creativity.
Tesla may be onto something (Score:5, Insightful)
1. Lidar and high precision mapping doesn't really help towards a generalized solution to the problem. I think they are probably correct here.
2. They have a special purpose accelerator specifically designed for this problem, with redundancy built-in at every level, and the necessary performance for the task. This is a tossup for me, and mostly only matters from a power consumption perspective.
3. They are not trying to explicitly program rules, as this is an intractable problem. Instead, they want to interrogate the fleet to provide high-quality data so that the neural network can train itself. This again, seems like a sound principle to work on.
4. They have the infrastructure (i.e. the fleet + software) to gather high quality data to feed the network. Large volumes of similar data alone is not enough, as you run into the overfitting/sparse data problem. They've built out the software infrastructure to gather a diversity of high-quality examples so that the neural net can learn in a very generalised way.
5. Don't under-estimate exponentials.
The fourth and fifth are still the biggest stumbling blocks IMO. The number of bizarre cases are so many, it's not clear to what extent even that kind of infrastructure can gather sufficient data to solve the problem. However, one thing that is clear is that, at least, they do have no. 4, whereas the competition is not even close, which gives them a significant leg up. As for Elon's predictions on the timeline, he seems to be relying heavily on no. 5, but it's also not clear whether skynet is actually learning at a geometric rate, and to what extent no. 4 will scale to allow it. Considering the rate at which auto-pilot is reported to be improving, they are probably expecting a good outcome.
Re: (Score:2)
5. Don't under-estimate exponentials.
Problem is that the amount of training data only grows linearly (or maybe a little faster thanks to growth of fleet)
Training data still not fully used (Score:2)
The training data they have only grows liberally but they also have vastly more of that data than they are currently using, so they can still have exponential learning growth for a while before they are constrained by training data.
Re: (Score:2)
You can have a lot of data in total, but most of that data will show the same common cases. The problem is that the car also needs to know what to do in extremely rare corner cases, like a bunch of protesters blocking the road, a fallen traffic light/stop sign, a crashed truck that caused a giant paint spill, some directions handwritten on a sign, a hay truck catching fire, or a tornado approaching the road.
Re: (Score:2)
It doesn't necessarily. It just needs to know when to pull to the side of the road and stop. Much like a safe human driver.
For personal vehicles, have some manual controls so a person can take over. For an autonomous taxi, a remote driver could connect and assess the situation. But in many of those rare corner cases the best approach for a human driver is also to stop and wait it out. So few do.
Re: (Score:2)
It doesn't necessarily. It just needs to know when to pull to the side of the road and stop. Much like a safe human driver.
If there's a tanker truck leaking fuel, pulling over to the side and waiting is probably not what you want to do.
But in many of those rare corner cases the best approach for a human driver is also to stop and wait it out
No, you also have options of turning around, going slowly around obstacles, keep driving but slow down/speed up/change langes, follow directions, roll down window and talk to people. Stopping and waiting is only a good strategy if you can see that the problem is going to resolve itself.
Re: (Score:2)
I didn't say it's the *only* solution. I said it can be the best approach. Most notably in the "but can it drive in the middle of a blizzard" example people love to give. Anybody who's actually done that knows that when you can't see the road, the best thing to do is to pull over and wait it out.
With sufficient training data, there's no reason why a self driving car couldn't learn to take most or all of the alternative actions you suggest. In their presentation Tesla showed examples of overturned vehicles
Re: (Score:2)
They can request the exceptions (Score:2)
The problem is that the car also needs to know what to do in extremely rare corner cases,
Tesla knows when autopilot is on and a driver intervened, so they automatically get rare cases that way.
Also they can see the exceptions on request - for instance, they can ask the fleet to show them images of traffic lights, which would include the case you mentioned where the traffic light had fallen.
They can also presumably request images of extreme visibility loss from weather and train for those cases as ell...
Tru
Re: (Score:2)
Re: (Score:2)
How would you possibly know "they also have vastly more of that data than they are currently using" anyway?
Because currently they query for items, they don't get back feeds from every car 24x7. If they needed more data they would be making stronger use of the transmitted data from cars.
It really, really, really helps to talk about this if you actually watch the video where the technical experts (not Musk) talk about what they are doing.
Re: Tesla may be onto something (Score:2)
What the extra compute power allows for is more shadow models to be able to run alongside real ones, so it allows for a greater range of model evaluation and improvement (and maybe just plain more models as well for different subtasks)
Re: (Score:2)
Number 4 is the most compelling thing about Tesla. It's a clever strategy, which I suspect they must have planned in advance. No company would put a bunch of expensive equipment in their product unless they expected to get something out of it.
With machine learning, training data coverage is really by far the most important thing, and Tesla is credibly far ahead of the competition in that aspect.
Re: (Score:2)
Re: (Score:2)
1. Lidar and high precision mapping doesn't really help towards a generalized solution to the problem. I think they are probably correct here.
In the sense that you still have to solve the problem of how to boil driving down to rules or an AI you are correct, but lidar has massive advantages in the much harder and more problematic part of the overall problem - mapping the environment around the car.
Lidar doesn't get blinded by the sun and drive straight in to trucks. It provides direct measurements of distance and shape. It isn't fooled by reflections of objects in windows. It's also immune to lack of adequate lighting.
Consider the common case of
Zero advantage (Score:2)
idar has massive advantages in the much harder and more problematic part of the overall problem - mapping the environment around the car.Lidar doesn't get blinded by the sun
Only one camera at a time would, and there are precautions you can take to reduce such blinding of camera lenses.
It provides direct measurements of distance and shape.
Tesla gets that as well, if you watch the video you see how - for one thing they have much greater stereo separation than humans, For another they ave radar and ultrasoni
NN are real, and they are spectacular (Score:2)
You guys definitely think like tech guys who grew up during the boom
I grew up way before the boom. You absolutely think like someone that grew up before the boom and has since ceased to learn about what technical advances can and cannot do.
train the biggest "neural net" and have the fastest processor
Huge clue you don't really know how any of this works.
If it were possible, it would be demonstrated already
You see simpler forms of it demonstrated every day by Autopilot in non-preselected conditions. Impro
The event video is long but extremely interesting. (Score:2)
I was just going to skim but ended up watching the whole 3 hour video of the event (well after skipping over the hour of marketing video on the YouTube.
The talked in great deal about the chip, and about the very substantial gains over hardware they could deploy (a CPU + GPU combination allowing for 17FPS analyzing, the custom chip something like 2100 FPS).
They also talked quite a bit about the software, and the approach to training the self driving models. They debunked a number of things I saw posted on a
Re: (Score:2)
I disagree that a computer - any computer, even all the world's supercomputers - is capable of sufficiently understanding the context of any limited-view set of any resolution video image of a scene to resolve it to a sufficient approximation of a 3D representation that I'd be willing to direct movements of a vehicle outside of my control at 70mph.
"Another winter of training data" sounds like a nice annual-delay, when you could just put a bunch of Tesla's into some snow-covered country and get that data to
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
LIDAR is as they said a shortcut, but doesn't help with the real world work required to understand the context of your surroundings.
That's insensible. Of course it does. It tells you where things are, even when an optical sensor would be blinded, and when you'd need multiple optical sensors to make a determination of range. It also is much better and more precise at determining range, in general.
The correct solution is a combination of sensors, probably at minimum including cameras, lidar, and sonar, and also ideally including radar. Putting all of that information together in one model enables much more rapid detection and identificati
Link to the source (Score:5, Informative)
Come on slashdot, just link to the actual source. It is this 3 hour long super in-depth must-watch talk full of interesting details by some technical people people from Tesla:
https://www.youtube.com/watch?... [youtube.com]
(fast forward the first hour; it is some promotional screensaver)
Spoiler: it is awesome. And confidence inspiring. They are absolutely going to pull this off.
The elephant in the room (Score:2)
In places where there aren't enough people to share their cars, Tesla would provide a dedicated fleet of robotaxis.
I'll give you three guesses why Tesla would be trying to figure out some way to generate revenue from a big bunch of its cars ASAP....
This seems like a much more candid indicator of ongoing demand than the usual earnings call puffery.
Re: (Score:2)
Ignorant of what? How to keep your brand in the paper at all times? Free advertising? Getting additional investor money? The only people who are ignorant are those that don't understand why CEO's come out promises of unicorns and world peace.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Anything even remotely complicated isn't using ladder logic!
There's much better visual descriptions than picturing everything as relays
You choose different state machines with the behaviors you need, to properly define the states, and transition triggers / actions that make up your application. This application is then used by a similar state machine controlling multiple applications as a part of a larger overall task.
The propagation of signals, and
priority of the actions taken on them is predefined. This
Re: (Score:2)
You use Simulink or similar to define the system behavior, and that generates the code / logic.
Here is what runs the engines and vehicles I helped develop, this has been in production for about 10 years now.
Everything, the engine controls, emissions, boost, throttle, shift, fly by wire, GPS navigation, autopilot, hybrid systems. Everything is a Simulink model.
http://mcs.woodward.com/suppor... [woodward.com]
And as far as industrial controls, the last 5 years, I've been watching the industry leave the people who only under
Re: (Score:2)
You clearly did not watch the autonomy presentation where they did an (extremely) deep dive into the hardware. Everything is redundant - power supplies, processors, external communications, servos, etc, etc.
You want to write a hardware-level neural net in ladder-logic?
Re: (Score:2)
Why would I pay for that level of redundancy? Like I give a shit if my car stops sending data to the manufacturer.
Worse still, given you have to pay extra for the driving assist features anyway, why should someone not paying for that software have to pay for redundant hardware they're not even going to use, and on top of that act as a testbed for the driving assist features they can't even use?
Tesla has a fucking horrifically abusive business model.
Re: (Score:2)