EAP HW2 - my experience thus far... [DJ Harry]

[NeverEnough]

How do you explain the jitter at a complete stop ? My guess is something to do with the 3 cameras and trying to make sense of the depth.

Looping through your video, most of the wiggle is next to the car and becomes greater the further back. It looks like there is a small wiggle up front and that is being amplified as it projects the line further back. Interesting... my hope is it will soon use side and rear camera data to improve line estimation which will greatly stabalize the estimation! However, I am not alarmed at the jitter at all. In fact, I would expect more jitter at a stop than when moving. The greater the speed the closer the forward lines represent the lines next to the car.

 

[oktane]

Started pretty patient, but am getting annoyed at the lack of information and communication.

Now starting to feel like a lot of others with HW2, that it's simply unbelievable that Tesla didn't have the AP1 functionality even close to parity on the new hardware before release on NYE. I think we were all lead to believe the software was done but going through validation. It seems it's not even close to being done, and there's a long road ahead to even full parity let alone EAP.

Even some small updates would go a huge way for the new owners waiting for the software update.

Agreed. I'm already getting very irritated over this, and will lose my *sugar* if not functional (AP1 parity) by the end of the month.

 

[chillaban]

Started pretty patient, but am getting annoyed at the lack of information and communication.

Now starting to feel like a lot of others with HW2, that it's simply unbelievable that Tesla didn't have the AP1 functionality even close to parity on the new hardware before release on NYE. I think we were all lead to believe the software was done but going through validation. It seems it's not even close to being done, and there's a long road ahead to even full parity let alone EAP.

Even some small updates would go a huge way for the new owners waiting for the software update.

I think this hits the nail on the head. It is looking undeniably apparent that when Tesla started offering the hardware 2 months ago, it really was only the hardware was remotely ready, and they probably had great confidence that the software is there, but it wasn't even close to production ready.

I hate to say it, but it looks like you AP2 folks probably have another several months ahead of you for even an early-access version of AP1 parity, and I'd be pleasantly surprised if the "enhanced autopilot" abilities to automatically change lanes to maintain set speed on freeways is available (widely) before the end of the year.

I also would not be surprised if the full self driving promises don't end up materializing, much like the original promise that AP 1.0 would do onramp-to-offramp and understand traffic lights.


The stuff I'm seeing right now in the videos, quite frankly, looks like it came out of Geohot's basement. It's a really loose approximation of what AP 1.0 was, even on the TACC front. It sure looks like Tesla relied a lot more on MobilEye's cameras than they wanted to admit. Even for TACC functionality.

 

[Saghost]

One example where we know of where they are NOT going with the same parts is battery packs where the 3's will get the new 2170 batteries while the S and X have 18650s and are not expected to change for at least a couple of years. In this case, the argument could be made that the 2170s are actually BETTER than the 18650s, but the driving reason Tesla is going to them is because of cell cost -- the 2170's are supposed to be MUCH cheaper coming from the Gigafactory than the 18650s are today. The expected higher energy density is simply a benefit of the newer design/advances in battery technology, but it is the cheaper battery cell cost that Tesla is counting on to help them hit the Model 3 cost targets.

This really isn't the right thread for this discussion, but since you brought it here: If the 2170 cells are much cheaper, why in the world wouldn't Tesla switch the other cars to them and take the extra profit margin?

Assuming they have factory capacity for it, of course - but even if they are limited in number of cells, they'd make more money switching the other cars to 2170 and building fewer 3s than continuing to build the S/X with 18650s and selling a few more 3s.

2170s are also likely to enable higher capacity packs - and they need a high discharge 120 or 130 kWh pack late this year or early next year to push back against Lucid and Faraday Future anyway.

 

[Saghost]

I think this hits the nail on the head. It is looking undeniably apparent that when Tesla started offering the hardware 2 months ago, it really was only the hardware was remotely ready, and they probably had great confidence that the software is there, but it wasn't even close to production ready.

We kinda knew this after the news items, blogs, and interviews this summer.

Mobileye thought they had Tesla over a barrel and tried to strongarm them, Tesla stood up and refused to play, then rushed out their own hardware solution.

It's looking like duplicating a decade of Mobileye IP is taking a little longer than expected, but it'll put Tesla in a much stronger position in the long run.

 

[chillaban]

We kinda knew this after the news items, blogs, and interviews this summer.

Mobileye thought they had Tesla over a barrel and tried to strongarm them, Tesla stood up and refused to play, then rushed out their own hardware solution.

It's looking like duplicating a decade of Mobileye IP is taking a little longer than expected, but it'll put Tesla in a much stronger position in the long run.

I agree. I mean, I completely agree in the long run, it will be the right way forward. I bet it will be during this year that AP 2.0 eclipses 1.0's abilities and 1.0 stops getting as many enhancements, and before long this initial period will be a long-forgotten memory. Just like how the first year or so of AP 1.0 was not at all interesting and most of us seem to selectively forget that it ever happened

 

[macpacheco]

This really isn't the right thread for this discussion, but since you brought it here: If the 2170 cells are much cheaper, why in the world wouldn't Tesla switch the other cars to them and take the extra profit margin?

Assuming they have factory capacity for it, of course - but even if they are limited in number of cells, they'd make more money switching the other cars to 2170 and building fewer 3s than continuing to build the S/X with 18650s and selling a few more 3s.

2170s are also likely to enable higher capacity packs - and they need a high discharge 120 or 130 kWh pack late this year or early next year to push back against Lucid and Faraday Future anyway.

1 - Because there isn't enough supply to migrate even 10% of MS/MX production right now to 2170
2 - Because current MS/MX packs are plenty good enough for 90% of customers
3 - Because 2170 GF ramp-up closely matches storage and M3 demand
4 - Because Musk and Tesla engineering has enough on their plate to create a 2170 battery pack and be forced to migrate the M3 to a 2170/18650 hybrid if they prioritize 2170 towards MS/MX.
5 - I assume third party 2170 production isn't anywhere near to supply all MS/MX right now
It a matter of another 18 months until GF is fully built and Tesla goes 100% 2170. If M3 is delayed for any reason, expect Tesla to migrate MS/MX packs to 2170 right away...

 

[alapd]

While I don't have the .185 update yet, I do have HW2. While driving today in the snow, my sensors bugged out a lot. I was stopped at a traffic light and the car kept telling me to stop. The side sensors were all over the place as well, and one point it said park assist was unavailable. While this was not a major issue while driving myself, I do imagine it to be a big issue with self driving in the future.

 

[drklain]

This really isn't the right thread for this discussion, but since you brought it here: If the 2170 cells are much cheaper, why in the world wouldn't Tesla switch the other cars to them and take the extra profit margin?

Macpecheco hit it right on the head. This has been discussed a lot on the main Tesla forums. The other reason is that Tesla has a contractual obligation with Panasonic until at least 2018 to buy 18650s from Panasonic's plants. So they have to buy the batteries anyway and have an established battery pack for the Model S and X. Elon has been public that the initial output from the Gigafactory batteries is for Powerwall units (he said that at the January 4 event as well). He (and JT) have previously said that the Gigafactory batteries will be cheaper than 18650s (based on the automated production line) and that the result will drive down costs enabling them to hit their price targets on the Model 3. A number of industry analysts have stated the believe the Model 3 price targets are aggressive anyway but that there is no way Tesla can meet them without getting the battery price down.

I think if Tesla could shift everything over to 18650's as soon as the Gigafactory output is sufficient, they would, but contractual obligations have to be met...

The relevant discussion is this thread: 2170 Battery in Model 3 First? | Tesla Motors

Of specific detail about the contractual agreement is Pungoteague Dave's post:

The issues are (1) contractual, and (2) volume/logistics.

Tesla has a contract with Panasonic that guarantees a minimum purchase of 2 billion 18650s in order to defray the expansion of their battery factory in Japan, signed and beginning in 2013, re-written and expanded in 2014, running until 2018. There are 7,104 18650 cells in every 85, 90 and 100 (perhaps tweaked in the 100 as no one has cracked one open and talked about it yet). Tesla contractually must buy 18650 cells until they produce about 282,000 MS/MX vehicles with them (after fall of 2014). That's at least another year of production. Further, after some concern in Japan, Elon gave a press conference in Tokyo where he said the gigafactory was for the 3 and power system, and that Tesla would continue sourcing the S and X batteries from Japan for some unspecified time. Tesla could rewrite the contract or shift the cell format being produced in Japan, but there is a contract and a political imperative to keep producing batteries for Tesla in Japan (remember that Tesla is a big retail brand in Japan, and are wedded to Panasonic). Panasonic's performance at its Japan battery facilities is a two way street for Tesla - Tesla needs Panasonic to be a good partner and cannot screw them, must consider their needs too.

Another issue is the existing form factor - Tesla barely got the X out to customers in volume a few months ago. Are they really going to do an update as substantial as a re-engineered battery pack for that car after one year of production?

The practical reality is that Tesla's future is about ramping up the 3 in the next six months to production levels that have never been reached with the combined S or X. They need every bit of gigafactory volume dedicated to making the 3 happen, and only when they've sold the first few hundred thousand does it make sense to think about converting the S and X - there aren't going to be enough cells to do it.

Thinking about it another way - the gigafactory is now about half its intended build-out size, and Tesla is already looking for gigafactory sites overseas. If Tesla knows today that it needs to at least double what it has already built in Nevada to get to the 500k cars/yr level in 2020 (perhaps an Elon wet dream - but the current stated goal), there's probably no way they have capacity for anything else except the 3 until the Nevada gig expansions and new battery factories come along - unless they can do a replacement deal with Panasonic to use the existing 18650 capacity in Japan to make the new cells. Even then, does Tesla have enough engineering capacity to do this many things at once?

I am betting we will see one more major refresh using mostly the same body structure and panels on both the S and X in early 2018 - a year from now (can they really even do that in the next year given the 3's priority), using the current packs with the cell and content enhancements in the P100 cars. That would take them to a full redesign in 2020 for both the S and X, or maybe phased over two years.

Remember that the battery pack in the S and X is a stressed structural member - it has a bunch of torqued bolts and is integral to the body's integrity. That likely gets redone only as part of rethinking everything for the next generation of Tesla's halo cars. My sense is that there's enough life left in the incremental 18650 enhancements, as we've seen over the past four years, that Tesla will be able to keep pushing that window with very little expense. Putting the 21700 cells into the older models will be lost on most people now that they can produce an 18650 version with well over 300 miles range, the factor people really focus on. The battery's internal technicalities interest us, but most buyers care only about what the car does, and which cell Tesla uses will be a distinction without much difference for most people.

 

[Ulmo]

Precisely ! Thanks for confirming that AP1 doesn't do this.
Here is a video of that odd subtle lane-dance at a complete stop. I don't have the equipment to record while on the move ( yet)

Probably the loud rock music is shaking the cameras.

 

[MarcusMaximus]

Probably the loud rock music is shaking the cameras.

On the bright side, play your music loud enough and your car will start dancing.

 

[Ulmo]

Most likely, although the drive to keep Model 3 costs down may mean it does not have the same equipment as the S/X but some lesser level.

Meanwhile, Faraday (and I think Lucid) are making their EV's with more sensors than Tesla.

Faraday Future FF 91, Lucid Air First Rides - Motor Trend

Faraday Future 91: Its complement of sensors includes lidar, 10 cameras, 13 radars, and 12 ultrasonic sensors.

Lucid Air: I don't know yet.

Tesla Model S: 8 cameras, 1 radar, and 12 ultrasonic

Looks like FF91 has 13 radars, the things that Tesla is able to see a huge amount of data with, whereas Tesla only has 1. I'd also like to know camera placement for the FF91: can it see around cars to the left and right? If they put it in those mirrors that hang out left and right, then likely they can. I think I'd be a lot safer in a fully programmed self driving FF91 than AP HW2 Tesla, simply because it has better sensor input, more knowledge of what is going on. Anyone sitting bitch knows they can't see around the vehicles in front of them as easily as the people on left and right (driver and passenger), and yet, that's where Tesla mounted their AP HW2 cameras. A FF91 will see red brake lights faster than the Tesla AP HW2, if I interpret that right. I'd like to know if the FF91 cameras are color aware; we know Tesla's don't see in color, so they'd probably not be able to distinguish a headlamp from a taillamp.

I'm a lot disappointed that we are seeing supposedly "self-driving" hardware start off with less tools than humans already have driving the car. We see color. We see depth. We see around things by being on the side of the car and able to look down the side of a column of traffic (especially if we weave out gently for a fraction of a second and weave back to let others see past us too so they don't hit us).

 

[Saghost]

Meanwhile, Faraday (and I think Lucid) are making their EV's with more sensors than Tesla.

Faraday Future FF 91, Lucid Air First Rides - Motor Trend

Faraday Future 91: Its complement of sensors includes lidar, 10 cameras, 13 radars, and 12 ultrasonic sensors.

Lucid Air: I don't know yet.

Tesla Model S: 8 cameras, 1 radar, and 12 ultrasonic

Looks like FF91 has 13 radars, the things that Tesla is able to see a huge amount of data with, whereas Tesla only has 1. I'd also like to know camera placement for the FF91: can it see around cars to the left and right? If they put it in those mirrors that hang out left and right, then likely they can. I think I'd be a lot safer in a fully programmed self driving FF91 than AP HW2 Tesla, simply because it has better sensor input, more knowledge of what is going on. Anyone sitting bitch knows they can't see around the vehicles in front of them as easily as the people on left and right (driver and passenger), and yet, that's where Tesla mounted their AP HW2 cameras. A FF91 will see red brake lights faster than the Tesla AP HW2, if I interpret that right. I'd like to know if the FF91 cameras are color aware; we know Tesla's don't see in color, so they'd probably not be able to distinguish a headlamp from a taillamp.

I'm a lot disappointed that we are seeing supposedly "self-driving" hardware start off with less tools than humans already have driving the car.

Actually, according to everything I've read, Tesla's cameras are white/white/white/red. Red is the only color they distinguish from the rest, and it can easily tell a headlight from a taillight. (Mobileye used this approach too.)

 

[MarcusMaximus]

Meanwhile, Faraday (and I think Lucid) are making their EV's with more sensors than Tesla.

Faraday Future FF 91, Lucid Air First Rides - Motor Trend

Faraday Future 91: Its complement of sensors includes lidar, 10 cameras, 13 radars, and 12 ultrasonic sensors.

Lucid Air: I don't know yet.

Tesla Model S: 8 cameras, 1 radar, and 12 ultrasonic

Looks like FF91 has 13 radars, the things that Tesla is able to see a huge amount of data with, whereas Tesla only has 1. I'd also like to know camera placement for the FF91: can it see around cars to the left and right? If they put it in those mirrors that hang out left and right, then likely they can. I think I'd be a lot safer in a fully programmed self driving FF91 than AP HW2 Tesla, simply because it has better sensor input, more knowledge of what is going on. Anyone sitting bitch knows they can't see around the vehicles in front of them as easily as the people on left and right (driver and passenger), and yet, that's where Tesla mounted their AP HW2 cameras. A FF91 will see red brake lights faster than the Tesla AP HW2, if I interpret that right. I'd like to know if the FF91 cameras are color aware; we know Tesla's don't see in color, so they'd probably not be able to distinguish a headlamp from a taillamp.

I'm a lot disappointed that we are seeing supposedly "self-driving" hardware start off with less tools than humans already have driving the car. We see color. We see depth. We see around things by being on the side of the car and able to look down the side of a column of traffic (especially if we weave out gently for a fraction of a second and weave back to let others see past us too so they don't hit us).

Neither of those actually exist yet. Comparing a real car you can buy today against ones that may or may not someday exist isn't really a fair comparison.

 

[DJ 240V]

EAP AutoSteer take 2 - easy traffic nothing crazy happened .

Based on Elon's latest tweet we will get another update tonight. It's clear that it's far from being ready.

 

[NeverEnough]

EAP AutoSteer take 2 - easy traffic nothing crazy happened .

Based on Elon's latest tweet we will get another update tonight. It's clear that it's far from being ready.

Your message seems contradictory. Judging from Elon's wording, I took his message to indicate a higher confidence than the initial rollout. My assumption is that this new update will fix/improve the reported problems and there is a higher confidence that the rest of the fleet will be activated throughout the week.

 

[Ulmo]

Meanwhile, Faraday (and I think Lucid) are making their EV's with more sensors than Tesla.

Faraday Future FF 91, Lucid Air First Rides - Motor Trend

Faraday Future 91: Its complement of sensors includes lidar, 10 cameras, 13 radars, and 12 ultrasonic sensors.

Lucid Air: I don't know yet.

Many, and well placed, it seems.

 

[drklain]

Faraday and Lucid are both one step above vaporware right now. I have more confidence in the work Mercedes and Volvo are doing getting to market before Faraday and Lucid ever do, and Tesla has a head start on all four. Comparing what Faraday and Lucid are talking about to the final product is like going to the Detroit auto show, looking at a concept car and expecting to see that on the finished product....

 

[DJ 240V]

Your message seems contradictory. Judging from Elon's wording, I took his message to indicate a higher confidence than the initial rollout. My assumption is that this new update will fix/improve the reported problems and there is a higher confidence that the rest of the fleet will be activated throughout the week.

That's what his message implies , but based on the maturity of what I'm seeing they need to be cautious. I'd rather have a delay than rush to keep a promise - especially when the stakes are this high. For e.g I would only use TACC ones I'm in a lane , set it to 7 car distance and disengage when I'm ready to exit . However if a larger populations pushes the usage under the false expectation of level 5 autonomy, Tesla will have a bigger problem on their hands. With rhe current state of AutoSteer , its more stressful for me. I'll only do it provide to data when I'm alone doing nothing else. I.e. No music no phone call etc ... and almost never with the family In the car . Its wise he is restricting it to 35 mph .
My confidence level will increase if they start showing that the sensors can see on the adjacent lanes as well and not just the current lane. I feel the rate of display of a car that shifts into your lane is a second too slow for my confidence.

 

[oktane]

Same problem with the side-collision monitor and display. The obstacle pops up about 2 seconds after you've already passed it.