Tesla’s Autonomous Coast-To-Coast Trip Now Some 3 To 6 Months Away

Tesla Autopilot


Tesla Autopilot

Tesla Autopilot 2.0 Update – On the freeway in heavy rain

During this week’s Q4 earnings call, Tesla CEO Elon Musk spoke about the missed coast-to-coast autonomous drive, reasons behind the delay, and his thoughts regarding the technology.

According to Musk, Tesla could have pulled off the planned coast-to-coast autonomous drive prior to the end of 2017, though it wouldn’t have been good enough. Some will surely say that this is just another of Musk’s excuses and that these delays are getting ridiculous (like Woz, who’s done believing anything Musk says), while others will argue that Tesla needs extra time and should be cut some slack. The CEO says that the tech may be ready in as soon as three months, but it could take up to six.

Essentially, Musk explained that it’s important that the Full Self-Driving technology is capable of working in any location, regardless of conditions. This is much different from using a pre-planned route and attempting to aim for cooperative weather. He shared (via Teslarati):

“We could have done the coast-to-coast drive, but it would have required too much specialized code to effectively game it or make it somewhat brittle and that it would work for one particular route, but not the general solution. So I think we would be able to repeat it, but if it’s just not any other route, which is not really a true solution.”

Musk did admit that the process has been slow, but as usual, he assures that Tesla is making progress. He elaborated that what seems like slow progress now will eventually be exponential. Once the automaker releases the tech in full capability, Musk believes people will be truly wowed. He said:

“It’s also one of those things that’s kind of exponential where it doesn’t seem like much progress, and suddenly, ‘wow.’ It will seem like well this is a lame driver. (Then,) like okay, that’s a pretty good driver. (Then,) like holy cow, this driver’s good.”

The CEO went on to answer questions about LiDAR. As most of you are probably already aware, Musk doesn’t believe that LiDAR is needed in order for Tesla to achieve fully autonomous driving. Not only is it expensive, but it’s also unsightly and has weather-related limitations. Musk asserted:

Tesla Autopilot

Tesla Model X With Enhanced Self Driving Hardware/Software drops of its “driver” and parks itself.

“In my view, it is a crutch that will drive companies to a local maximum that they will find very difficult to get out of. They’re going to have a whole bunch of expensive equipment, most of which makes the car expensive, ugly and unnecessary. And I think they will find themselves at a competitive disadvantage.”

Musk believes that the successful combination of advanced radar, multiple forward facing cameras, and near-field ultrasonics is the key to successful self-driving tech. He reminded that this mix of equipment enables the car to remain aware through any type of weather. These vehicles should be able to drive without cause for concern in heavy fog, dust storms, and blizzard conditions, even when a human driver would be unable to continue driving. Interestingly, however, he did admit that he could be proven wrong, though he’s pretty sure he’s right. Musk said:

“Now perhaps I am wrong. In which case, I’ll look like a fool. But I am quite certain that I am not.”

While Musk has built up a reputation for being wrong when it comes to delivery timelines, profits, and the like, he’s proven his prowess on the technology front on several occasions. Many people didn’t think SpaceX could land and reuse rocket boosters, or that Tesla would actually follow through with an electric semi truck, or that he’d follow through with a company that is actually digging tunnels for electric car sleads … and the list goes on and on. However, continually pushing back these timelines breeds increasing skepticism.

We can only hope that Musk is right and the automaker’s Full Self-Driving capability is a success, sooner rather than later. The feature will be available on the Model S, X, and 3 (following software validation and regulatory approval), but we have no way of knowing when it will become available in relation to the upcoming cross-country demonstration.

Source: Teslarati

Categories: Tesla

Tags: , ,

Leave a Reply

68 Comments on "Tesla’s Autonomous Coast-To-Coast Trip Now Some 3 To 6 Months Away"

newest oldest most voted
(⌐■_■) Trollnonymous

It’ll get pushed back…….lol

Bring on the FLAMES!!!!

Who cares!!!!???? Build the darn Model 3’s. Who cars about boring machine…Build the darn Model 3’s. Who cares about SpaceX…Build the darn Model 3’s.

Speak for yourself! I care about space X! ( But not as much as the Model 3. )

But yeah, autonomy and the Boring Company I could take or leave.

I’m the opposite. Who cares about the damn Model 3?!! Get us to Mars and quit messing around with cars.

I don’t believe that for a second

Tesla did test autonomous cars *at all* on public roads in California in 2017. Until that resumes, don’t hold your breath.

*did not

Tesla don’t need to test vehicles on public roads, they got all the data they need to test their AI from S, X and 3’s.

“Tesla don’t need to test vehicles on public roads”
What? Will you drive a Tesla only in simulator?

Not simulators, shadow mode. My car is secretly pretending to take actions while I drive, then uploading what the results would have been, back to Tesla.

Play with the stuff on closed test tracks all you want, shadow in real life. No need to risk hitting a motorcycle or something, like GM did. They can play back data from human caused crashes in Teslas and tell us if their AI would have avoided it. WAY better data than just driving a pre-programmed route and avoiding cars along the way.

Making Actions different from Driver behaviour can induce new risks. Others might behave different If you behave different. This will never been seen in shadow mode.

You overstate the effectiveness of shadow mode. In shadow mode, the car is not under autonomous control.

An autonomous car will always react differently from a human; if it did not, there would be no purported safety advantages from autonomous cars.

Are you seriously proposing that a coast-to-coast drive will be the first time Tesla will try AP in California?

“Are you seriously proposing that a coast-to-coast drive will be the first time Tesla will try AP in California?”

😆 😆 😆
It looks like you are trying to insinuate exactly that!

But of course, we can’t expect consistency or logic or reasoning, or actual facts, from a Tesla Hater cultist like you. That would be contrary to your FUDster agenda!

Tesla has been very clear that they are intentionally testing in other states (and countries) besides California.

You actually know this, you commented on it previously. But I guess when you wish to prove yourself intentionally ignorant on a daily basis, you have to dig deep, eh?

Where have I heard this story before…..oh yeah, almost exactly a year ago.

Said 3-6 months 1 year ago.

Jan 2019 Mysk will be saying 3-6 months for coast to coast autopilot. Lol

I give it 50/50.

I will give it 0%

Tomorrow’s technology, tomorrow.

It’s déjà vu all over again.
—Yogi Berra

The question was:

“At what point will “Full Self-Driving Capability” features noticeably depart from “Enhanced Autopilot” features?”

If you have some reading comprehension then it is clear Musk wasn’t talking about full self driving capabilities here.

More fake news.

More precisely: Musk wasn’t promising full self driving in 3-6 months.

In my view, it is a crutch that will drive companies to a local maximum that they will find very difficult to get out of.

Did He mean/ say maximum or maxim?

Tesla’s new motto: driving with eyes open is for sissies. Jk

Elon is going to get some derision for being late again (and deservedly so) but he seems spot on on this one. I appreciate that they opted to not waste time gaming this to save face.

Maximum capabilities, they can’t get past because of hardware and how they’re programming things. Tesla won’t have a maximum, they can keep growing and doing better. No keep the car home because of fog, rain, snow, or dust blowing around. No need to pull over and wait for things to calm down, or a human to pick you up. No artificial maximums, Tesla will adapt, see through all of those things with it’s radar, and press on, unlike other cars with LIDAR.

“Did He mean/ say maximum or maxim?” I think what Elon meant was that using LiDAR makes certain things (most notably, building up a virtual reality “picture” of the car’s surroundings) quite easy, so that makes it easy to achieve a “local maximum” in that part of developing autonomous driving. Elon apparently believes the ability of radar to penetrate fog and rain slightly or somewhat better than LiDAR makes LiDAR a dead-end technology, despite every other company developing self-driving cars coming to rely more on LiDAR than anything else. But I think Elon reveals his true motive for stubbornly resisting use of LiDAR when he says: They’re going to have a whole bunch of expensive equipment, most of which makes the car expensive, ugly and unnecessary. Is it “expensive”? Well, the price has been dropping like a stone, so that’s rather outdated. Unnecessary? Well that’s rather the question, innit? It’s circular reasoning to use your conclusion for one of your premises! That leaves “ugly”. And I think this is the real reason that Elon — probably overriding the objections of his designers, which he has a habit of doing — this is the real reason Elon is so stubbornly resisting… Read more »

Lidar is getting smaller and less expensive on a Moore’s law curve. Plus, the new ones are solid state. No moving parts.

Roof mount will give way to mirror mount or incorporated into shark antennae.

Musk is right about the range. I think he is wrong about everything else.

“Did He mean/ say maximum or maxim?”

Maximum. “Local maximum” is a concept from mathematics and is commonly applied in AI. The meaning in this context is that once committed to Lidar they’ll eventually develop a technology suite that does the best possible job using that suite — but that it won’t be as good as a different solution that would be developed with a different sensor suite. The “locality” in the context is defined by the sensor suite — you will tend to explore only solutions that fit within it.

“Good” in this context should be broadly construed to cover price/performance ratio — so an expensive solution that drives just as well as a less expensive solution would be considered inferior.

Note I’m not advocating (or denigrating) Musk’s position — just answering your question about maximum/maxim.

You know it’s easy to be a couch critic, but given the military’s use of all weather laser systems since the late 90s, and the fact that Google and several others have drastically driven down both the size and cost of the LIDAR units, Musk may not be right here.

LIDAR does a really good job at making our 3d world navigable to a computer.

Aesthetics is certainly important, but then to a large extent our perception of cars is shaped by necessity. Little LIDAR bumps in key locations on the car, but well integrated into the body work will probably be totally normal in a few years and no one will pay them any more mind than we do side mirrors or side yellow markers now.

LIDAR and camera’s also share the same weakness, it’s a weakness I’m reminded of every time I have to go wipe off my rear view camera – so I don’t see a win there for camera’s.

We’ll see, but I’m betting on LIDAR because it’s currently already working for at least 3 or 4 other manufacturers.

Musk’s logic is simple: Humans don’t have LIDAR and humans can drive. Therefore, computers don’t need it either.

Humans dont have Radar, Humans dont have Ultrasonic, Humans can Turn their Heads.

Do future cars now need a turning camera and nothing else? Probably not.

A turning camera, plus all the other human sensors and nothing else plus software is all you need. ?

The software is the hard part.

Humans aren’t that good at driving. If they were, the we would have no need for self-driving cars.

The goal of autonomous cars is not to mimic human drivers; it’s to be better than human drivers. One of those ways will be by using active scanning with LiDAR, which is almost infinitely better at scanning the environment than using passive scanning with either cameras or the human eye.

Video cameras — at least the ones used in Tesla cars — can’t “see” at night any better than the human eye can. Contrariwise, LiDAR doesn’t care if it’s light or dark outside. That’s only one of the ways LiDAR is superior.

“Humans aren’t that good at driving. If they were, the we would have no need for self-driving cars.”

Say what?!???!!! I want self-driving cars because 99% of the time driving a car is mind-numbing. Nothing like you see in the commercials.

Lasers aren’t LIDAR although LIDAR uses lasers. LIDAR isn’t all weather and never can be.

Infrared-frequency LiDAR should be nearly as good as radar for penetrating rain and fog, because the EM wavelengths will be nearly as long.

Radar is also degraded somewhat in rain, as are LiDAR and optical systems such as video cameras and the human eye.

This isn’t a binary, either/or situation. All EM (electromagnetic) sensors are degraded by adverse weather conditions such as rain. Some are more degraded than others. The best solution is to figure out ways to minimize that degradation, not refuse to use LiDAR simply because visible-light LiDAR is somewhat more degraded in rain than is radar.

I hope he’s making sure a Tesla never crashes into another Tesla. That would be very bad.


It would be VERY BAD if a Tesla S, X, or 3 driving by AutoPilot crashes into the side of a Tesla Semi Truck turning in front of it to cross a divided highway.

“We could have done the coast-to-coast drive, but it would have required too much specialized code to effectively game it or make it somewhat brittle and that it would work for one particular route, but not the general solution.”

Because if there’s one thing Elon Musk would never do, it’s throw together a gimmicky one-off show pony for publicity purposes, if it doesn’t represent the real-world application of the technology. That would be like having a press event for a 90-second battery swap when the actual battery swap process takes much, much longer.

It should go without saying, but if Elon could successfully game the system to complete the coast-to-coast drive, he would have already done it. The fact that they haven’t means that they can’t even figure out a way to execute an autonomous coast-to-coast drive in optimal conditions.

I concur

I suspect they brainstormed the idea of gaming it and realized it would be nearly as hard as doing it right, so they opted to eat crow and do the right thing. Good for them.

Or re-shooting the autonomous “demo” over a 100 times before they managed to do it without intervention when going a couple of blocks in California. Imagine the number of reshoots they would have to make to go from coast to coast!

I see that serial Tesla apologist Get Real is as usual attacking the poster instead of the post itself. One can only assume because the post is full of truths…

It would be really nice if IEVs would try and curtail this behavior, on both sides. Their failure to do so has resulted in a somewhat hostile comments section.

I certainly do wish InsideEVs would put limits on people repeatedly posting lies which they obviously don’t honestly believe.

That would certainly curtail the endless stream of FUD from Tesla Hater cultists like you, DJ, and also Six Electrics, and Dav8or, and bro1999, and “Someone out there”.

It would also be nice to see a restriction put on “pushing the envelope” of extreme hypocrisy, such as GM fanboy Spider-Dan accusing Tesla of less than complete 100% truthfulness.

Of course, GM is never guilty of hype or exaggeration in their claims. For example, they’d never make an absurdly exaggerated claim like the Volt getting 230 MPG.

Oh, wait…


By the standard in place at the time that was truthful. Of course, just like NEDC, it was a lousy standard. The government came up with MPGe to fix it (which you don’t like either).

Not that GM is always truthful. They lied about how the Volt works, for instance.

I’m glad you mentioned this again, as the last time you posted it, it was late in a stale thread.

That 230 MPG figure that you call “absurdly exaggerated”? Yeah, I did 50% better than that in real world usage over 3 years when I had my previous Gen1 Volt.


I guess by your standard, a BEV gets infinity miles per gallon?

Quoting 1 fuel when you use 2 is ridiculous and is just attempting to hoodwink people who don’t understand the technology.

As we all know, electricity is neither free or entirely without environmental consequences. You can’t just ignore it.

Why, yes: a BEV, which uses no gasoline, therefore gets infinite MPG.

Of course, that isn’t including the times that a BEV’s owner chooses to drive another ICE instead because of range/charging concerns… something that doesn’t apply to a Volt.

If you are looking for a fuel-efficiency metric, MPGe exists and will meet your needs. But if you are looking for a “How much money am I sending to middle-eastern theocrats?” metric, MPG is perfectly valid.

Bro1999 traded his Bolt for an S60. Doubt if he posts any more Tesla FUD.

In particular, the “Musk lies” or “Tesla Lies” is nothing more then a meme being pushed by the short sellers from the financial gambling sites like Seeking Liars.

Of course it gets repeated now Ad Nauseam by the usual suspects shills, shorters and haters here which is not surprising since many of them probably also frequent Seeking Liars.

Posting verbatim information from Tesla’s own filings counts as FUD according to you so I’m glad you’re not setting the standard here.

I mean, it should by easy – we have the hardware, you just need 100,000 lines of programming code. Lets just hire 100 cheap coding monkeys, they will start typing, and in 3-6 months we would certainly should have been typed 100,000 lines and it will be done deal. Debugging can be done on the fly by our devoted clients, will just stamp it as “beta”. What people should absolutely have zero concern about, and I mean 0, is that Tesla will achieve it by December 2016… eee, scrap it, 2018!

Stepped exponential, that is the thing! Get ready for the disruption now! To infinity and beyond, GO TESLA!!!

+1 😀

I see that the Elunatic Musk troll idiot posts under another username here.

Actually, I could care less about the projected timeline, what Elon said when, etc. I’m more interested in the task itself. It’s not like he’s about to be scooped by another company. Have we heard anything about how the car will charge itself, or will there be human crew at the needed Superchargers? It would be cool to have the car charge itself so it could be driving continuously without being touched. I know about the snake-thing; if they use it, I suppose they’ll have to leap-frog the car to get it set up at each destination before the car arrives. Also, I assume that the trip will be at the speed limit, which is going to make for one boring trip for the tag-along support crew.

As for demonstrating any road, any where, I suggest a 10 car to 12 car run.

5-6 cars start out driving east from California, Oregon, & Washington State: ideally 2 Model S’s, 2 Model X’s, & 2 Model 3’s! Each on different roads: Freeway, State Roads.

6 more start out on the East Coast, similarly split up in road and car types, driving west! Allow news crews to ride, follow, & track them!

Succeed at that challenge, some measure of belief will grow! Do it with just 1 car, one way, not so much!

How much less could you care?

I’m interested in the same issues you raise, such as how it will be connected to Superchargers, and whether that will be handled without any direct human assistance.

But I wouldn’t be so sure that Tesla won’t be scooped by another company, on this coast-to-coast self-driving demonstration. With Waymo and Lyft both deploying test fleets of self-driving taxis, how far off are they from making a similar demonstration run?

GM already demo’d their Super Cruise hands free Coast to Coast on highways.

Not level 4.

If Elon threw us a bone on charging, that would have gone a long away. Him not mentioning could suggest either a working auto chargier is functioning..or human assisted charging will still be part of the trip

I am sure the fanboi types will go ballistic but I am skeptical Tesla will be able to pull off FSD with their current hardware. If you want some sense of the challenge of relying on the images from today’s camera sensor/processing technology grab the camera with the best video DR (dynamic range) you can find (probably a full frame digital SLR) and get someone to drive you around at night and during high contrast scenes while you record video. Then look at the recorded video and see how much data the human eye can easily see that is missing from the video. You will quickly see the difference between the 12-14 stops of DR the best cameras available today can see Vs the 20-24 stops the human eye can see. And keep in mind sensor size, DR and lens aperture size are all related. Judging by the lens sizes on the Model 3 I suspect the camera sensor size is no where close to that of a digital SLR and hence has a lot less DR. One work around would be to be to apply HDR (high dynamic range) processing by taking 3 or more exposures and combining them… Read more »

Great explanation! Fact based and understandable explanations are hard to find these days! How much does radar help to provide some of the missing information such as speed of objects in the scene and distance. To me, extra dynamic range and resolution helps our brains to distinguish subtle movement information and object separation for distance.

“requires a *LOT* of processing power. More than is available on high end desktop computers today *for each camera*.”

With that NVIDIA thingy, they’ve got how many TFLOPS?

Actually, the problem with trying to rely on using video cameras as sensors for a self-driving car has little or nothing to do with the video resolution. It has far, far more to do with the limited ability of software being able to do real-time optical object recognition, using the processing power found in one or two microprocessors. Neither the hardware or the software can come within orders of magnitude of matching the processing power of the human brain’s visual cortex.

Active scanning data from radar and LiDAR reflections don’t need anywhere near as much processing, and the data provided by active scanning doesn’t need processing by anything as slow and unreliable as optical object recognition software.

I agree that resolution isn’t the biggest problem in using video cameras. I think DR (dynamic range) is a far bigger problem because if the DR isn’t sufficient there won’t be any data for the object recognition software to process. And for those who aren’t camera nerds like I am each stop of DR is a represents a factor of 2. So, 20 stops Vs 12 is a difference of 2 to the 8th.
DR in sensors has improved by perhaps 3 stops in the last 10 years.

“Some will surely say that this is just another of Musk’s excuses and that these delays are getting ridiculous (like Woz, who’s done believing anything Musk says), while others will argue that Tesla needs extra time and should be cut some slack.”

What I say is that this shows what an extremely difficult problem it is to achieve reliable Level 4 or Level 5 autonomous driving, and that Tesla shouldn’t be wasting its time, money, and resources on developing this independently. Tesla should concentrate on what it does well, which is designing and building top-rated, desirable automobiles. Let other companies develop the self-driving tech; Tesla can license that from them.

But if Tesla does insist on going its own way on this, then they need to face reality and start using LiDAR. So long as Tesla stubbornly keeps trying to rely only on cameras and low-resolution radar, it’s going to continue to fall behind Waymo and other companies with advanced self-driving technologies.

All just my opinion, of course.

Must be coincident, that the date slipped into summer with nice weather and easy to drive road conditions.

Meanwhile in Japan, Mitsubishi develops self driving capabilities especially for difficult road conditions.