Does anyone know the life expectancy of the Tesla Model S and the Cybertruck?

[ajdelange]

No, not weird at all. Even though it is called a test drive you aren't really testing anything - you are being given a demonstration. Real testing requires operating the car for thousands and thousands of hours under a variety of carefully controlled conditions.

Suppose I asked you to flip a coin 10 times in a row and repeat the experiment 20 times then tell me how many times you got 10 heads in a row. You'd probably tell me not to waste your time because there is no way you can get 10 heads is a row. But there is. The probability that you will on a particular run is 1//1024, a small but finite number. The probability of getting 10 heads one or more times in 20 experiments is actually 1.9%. If you repeated the experiment 10,000 times you would get 10 in a row approximately 10 times. So it is with the autopilot. You drove it a few miles. If the probability of the autopilot doing something alarming in each mile driven is 1/1024 the probability that you will have 1 or more alarming events is 0.97%. But if you drive 10000 miles it is 99.994% i.e. almost certain. In a 500 mile trip it would be 38.6%. This is difficult to explain to the layman so no more mention of probabilities. Were I you I would just consider that there are a fair number of people out there driving Teslas in real world conditions who are reporting alarming actions on the part of the autopilot and conclude that at least some of these are not making this up to advance FUD.

I will also note that Tesla does not map the region (that's Waymo's approach). Telsa collects data from the cameras and uses that to train the neural network. Tesla knows where it is, of course, through the GPS receiver. It could use that knowledge. For example everyone knows that the first rule for driving in Boston is "Never make eye contact". I wonder how you program that into an autopilot.

Advertisement

 

[BillyGee]

You don't need to explain statistics and optimism bias to me. My experience was my experience and anecdotal, I'm aware of this. It was, however, on an open road with live traffic, so it wasn't controlled conditions and I was the one driving.

I'm also aware that it's not true FSD yet. I've watched the drive along videos. Clearly you need to keep your hands on the wheel and be ready to help it since it's driver assist, not a driverless system.

I've also seen several accidents get prevented by autopilot including ones involving my work truck. I foresee it as a net gain, even if it has a small amount of glitchiness in its current state, it's still a safer driver than most people on the road, including me after 8 hours on the road. I anticipate the new software and continuous improvements are going to put AP at much better capabilities by the time the CT launches.

This is difficult to explain to the layman so no more mention of probabilities.

You don't seem to understand how condescending this line is, get over yourself.

 

[ajdelange]

I'm sorry your sensibilities have been offended but only a layman would offer an anecdotal experience as the basis for thinking that the widely reported experiences of many Tesla drivers are weird.

Given a fair coin can you tell me the probability of getting more that 3 but less that 7 heads in 20 tosses? If not then you are a layman when it comes to this kind of analysis. If you can then you are not and I apologize. I am sure that I would be equally at sea if you pushed whatever it is you work on in front of me and told me to fix it. I would readily admit that I didn't understand and not take offense but you are you and I am I.

More to the point after this discussion I thought I'd see where autopilot is with the current software revision and so took a little run out to the country yesterday. It seemed that the number of hisses and gasps from my co pilot were about the same as the last time I used it. In about 180 miles of driving there was only one incident where there might well have been a crash had I not taken over. As in previous experiences with it the "Navigate on Autopilot" auto lane change feature only worked briefly and then not for the rest of the trip. So why did I order it for my CT and my new X? Because it is useful when traffic is light and you have a long drive. Tesla's statistics show that Tesla drivers using it are 8X less likely to have accidents than drivers who do not use any form of autopilot. This says that on average while it may scare the hell out of your passengers it is actually safer than not using it at all. On average. I apologize in advance for pointing out that that term has special meaning that layman won't fully appreciate. Here that seems to mean that if you are able to catch it in those situations where it is likely to do you a mischeif it is actually a lifesaver. But that is hardly a set of circumstances in which one can be relaxed.

Again the message I want to get across is that those who fancy full self driving is going to chauffer them about starting on the day they take delivery of their CT's are going to be disappointed.

 

[BillyGee]

I'm sorry your sensibilities have been offended but only a layman would offer an anecdotal experience as the basis for thinking that the widely reported experiences of many Tesla drivers are weird.

Did it not occur to you that someone can recognize that they're giving an anecdotal example to counter your anecdotes? "Most people" and "weird" don't make a trend.

I would readily admit that I didn't understand and not take offense but you are you and I am I.

I do understand, I just know that it's still astronomically small and most of the issues come from unusual use cases that the AI is still being trained to assess. I have a background in physics and statistical mechanics, so all you're doing is being condescending when you talk like this. You may as well be tipping a fedora and smoking a vape talking like that.

I'm not expecting a robot chauffer, I'm expecting something to take the edge off of long drives which I do regularly for work.

 

[Jhodgesatmb]

Weird, I tested it in heavy traffic and found it worked fine, if not better than I would have. As far as the weird behavior, I think the sensor suite on the CT will be a bit more fleshed out and they're dropping new software in a few months. I'm hoping it's near ready by CT launch only because it's the most taxi g part of my job as is.

It's also worth noting that I live in the Bay area of San Francisco, I imagine their mapping and ai scans are pretty thorough around here.

I live in the Bay Area as well, and I have used AP in my wife’s M3 to drive to Ukiah and back and it worked surprisingly well, even in weekend traffic around Santa Rosa. I was impressed. On 2-lane highways I am still edgy with it, but on the separated highways I definitely get to take my eyes off the road for moments at a time, I can relax my forearms more, and my legs, and was definitely less tired at either end of the trip.

 

[ajdelange]

I live in the Bay Area as well, and I have used AP in my wife’s M3 to drive to Ukiah and back and it worked surprisingly well, even in weekend traffic around Santa Rosa.

I think it works surprisingly well too despite its warts.

I was impressed.

I'm impressed enough that I have or have had it on all the Tesla's I've owned or ordered (including the CT).

On 2-lane highways I am still edgy with it, but on the separated highways I definitely get to take my eyes off the road for moments at a time, I can relax my forearms more, and my legs, and was definitely less tired at either end of the trip.

That's my experience with it too except that I, and my passengers get edgy on divided highway too when the traffic is heavy especially since I've had to take over several times to avoid collision with barriers or other vehicles. I've got a new X on the way which will have the new hardware. Maybe it will be more reliable.

 

[ajdelange]

I have a background in physics and statistical mechanics, so all you're doing is being condescending when you talk like this. You may as well be tipping a fedora and smoking a vape talking like that.

Well now that I know that you are conversant with probability theory I can simply explain in terms that you will understand why I responded the way I did and hope this will mitigate the offense. Using Bayes theorem, recognizing that the a-priori probability that a reader in this forum would know statistics is low and that the probability that a person who knows statistics would find that an event commonly reported is weird based on the fact that in his single observation it was not seen is also low the product of those probabilities, the liklihood that you are familiar, based on my observation and what I know about you a-priori, is low.

[Edit]Fun to put some numbers to this. Let's say the a-priori probability that someone in this forum is statistically conversant is p(s) = 0.03 and that the probability that someone who is statistically conversant would call a widely reported phenomenon wierd based on its absence in a single observation is p(w|s) = 0.03. So in a population of 100 there are 3 statisticians and 97 non. Assuming all the non statisticians would call the single observation weird but only 3% of the 3 statisticians would we'd expect 97.09 declarations of weird per 100 for p(w) = 0.9709. Thus the probability you are statistically conversant based on the observation and assumptions is
p(s|w) = p(w|s)*p(s)/p(w) = 0.03*0.03/0.9709 = 0.000926975 and the probability you are not is 99.91%. Thus a Bayesian (I am one) would respond as I did.

 

[Bathbunny]

Does anyone know what the criteria Tesla is using for the 1,000,000 mile battery rating is? What I mean is, a million miles based on how much its driven, frequency, state of charge and discharge between cycles, stuff like that. If one was to drive a 100,000 miles a year, and the battery last 10 years, there's the million miles. But I'm making a presumption that the rating is not based on 100,000 miles a year use.

Yes and no. Main damage to batteries is charging, so the 1'000'000 miles figure is a generic shorthand for a certain number of charge/discharge cycles. Most of the testing Tesla has relied on up to now is from a research lab in Canada that has studied Li-ion batteries for many years and that has developed technology to conduct accelerated testing of real batteries rather than rely on computer simulations. So their figures regarding batteries are solid; translating them to miles driven on a particular car model is trickier, mostly because of all the variables you mention.

 

[ajdelange]

Discharge is as damaging. It is current flow in either direction, Some of the lithium ions that are supposed to intercalate with the graphene in the anode during charge or with the salt crystals in the cathode during discharge wind up reacting irreversibly with the SEI and are thus lost to the battery for the rest of it's life.

The "research lab" in Canada was Jeff Dahn's lab at Dalhousie U in Nova Scotia. He now works for Tesla. I don't know if that means he's left the U to become a direct employee or is under a contractual agreement with Tesla but he's now their boy,

 

[Rix]

Yes and no. Main damage to batteries is charging, so the 1'000'000 miles figure is a generic shorthand for a certain number of charge/discharge cycles. Most of the testing Tesla has relied on up to now is from a research lab in Canada that has studied Li-ion batteries for many years and that has developed technology to conduct accelerated testing of real batteries rather than rely on computer simulations. So their figures regarding batteries are solid; translating them to miles driven on a particular car model is trickier, mostly because of all the variables you mention.

Thank you for answering my question, its interesting to see how this comes out in the end. The variables I mentioned will come into play...at least a little. Thanks again for chiming in.

 

[Ehninger1212]

Thank you for answering my question, its interesting to see how this comes out in the end. The variables I mentioned will come into play...at least a little. Thanks again for chiming in.

half of this thread is worthless non-sense.. But yes the 1,000,000 miles is based of the number of charging cycles.

 

[ajdelange]

Does anyone know what the criteria Tesla is using for the 1,000,000 mile battery rating is? What I mean is, a million miles based on how much its driven, frequency, state of charge and discharge between cycles, stuff like that. If one was to drive a 100,000 miles a year, and the battery last 10 years, there's the million miles. But I'm making a presumption that the rating is not based on 100,000 miles a year use.

Consider a CT driven 1,000,000 miles. The CT is going to take about half a kWh to go a mile (I hope it's really going to be less than that but using a half makes the math simple). Thus the CT would have taken 500,000 kWh out of the battery. That 500,000 kWh had to be put into the battery in order for it to be taken out so the charge/discharge load on that battery would have been 1,000,000 round trip kWh.

Now we didn't say whether we were taking about an AWD with a 100 kWh battery or a TrMotor with a 200 kWh battery, In the former case the charge/discharge load is 1,000,000/100 = 10,000C but in the latter it is 1,000,000/200 = 5,000C.

I hope this makes it clear that the "million mile battery" is a product of the marketing department - not the engineers. If you have a battery that endures 5,000C of charge/discharge that can be tied to a distance but it takes specification, as you suspect, of battery size, and consumption per mile. As the capacity of the battery also degrades as a function of charge rate and depth of charge/discharge cycle those too have to be specified.

Battery scientists measure degradation against C or cycles (one cycle from 0 to 100% is 2C). Marketers try to express that in miles.

 

[Ehninger1212]

No matter how you put it.. its a lot of freakin miles! at the rate i drive it will take me ~56 years to reach 1,000,000 miles. I will be 84 years old when i get my next cybertruck.

 

[ajdelange]

No matter how you put it.. its a lot of freakin miles! at the rate i drive it will take me ~56 years to reach 1,000,000 miles. I will be 84 years old when i get my next cybertruck.

Well that just illustrates why it is so silly to rate the battery life in miles. Don't look at the miles. Look at the slope of the degradation curve. Assuming I have a battery that degrades by 2% in the first year with a rates life of n*C now if it is replaced by one of the same size, C but with degradation rated at 10*n*C it is going to show me 0.2% degradation at the end of a year. There are lots of reasons to want that multiple of C to be as big as possible (relting residual value in the battery after vehicle end of life) but during life Tesla's complaints about battery degradation should go down by a factor of 10.

 

[Bigvbear]

I think a lot of it depends on how you treat the battery. I know there was one guy who always supercharged, never trickle charged (because he lived by a supercharger) and fried his battery quickly.