Firmware 8.0

[Mistersir]

Anyone? Anyone? VooDoo? VooDoo Software?

Is GlmnAlyAirCar a Rush reference?

 

[Flying Drummer]

Is GlmnAlyAirCar a Rush reference?

His avatar sure could be a Red Barchetta.

 

[GlmnAlyAirCar]

Is GlmnAlyAirCar a Rush reference?

My 1990 Cherry Red Cougar XR7 is my "Red Barchetta." The gleaming alloy air car is only fitting.

 

[lakerholic]

Is this sarcastic? It was definitely not. The first time I test drove 7.0 with ap1 we ended up nearly sideways on a freeway exit and the Tesla employee in the car screamed like a girl!

(Long story)

But it really wasn't until almost a year after autosteer was announced before it was robust, and that was almost 2 years after the hardware came out. Compared to that, I would say AP HW2 rollout is happening much faster....

I disagree. Took delivery of my 14 P85 in December 2014, AP1 HW rolled out in January/Feb 2015 if my memory serves correctly, and it just plain worked out of the box. I know there were many posts about people having issues with AP1 staying in lanes, etc... but for my use of it -- and how I use it today -- it worked out of the box: mostly divided highways, freeways, normal surface streets (meaning not a curved mountain road, or residential street with 15mph curves. All at speeds much greater than 35mph btw.

Admittedly, I have zero experience with AP2, but if the AP1 hardware roll-out was as slow and error filled as what I have been reading on these forums, I would not have been happy with what I thought I was receiving. I would think AP2 owners would be even more frustrated given the decent (I say excellent) response rate of AP1, vis-a-vis those who got AP1 initially, as we had nothing to compare against.

I said this once before and was quickly derided, but I am of this opinion until Tesla proves me otherwise: Tesla engaged MobileEye on AP1 software because this is what MobileEye does. It's the only thing they do. Then ego or threats of lawsuits or a power struggle occurred and now Musk is left reinventing an already pretty good wheel, instead of adding to said pretty good wheel. Which has resulted in the poor roll-out of AP2 we see today.

I have no doubt Tesla will get it right eventually, but at what cost? Pre September 16, you could've bought an AP1 car and used its functionality with minimal issue. Post September 16, your car has reverted to a pre September 14 Tesla in terms of functionality until such time as Tesla fixes all issues and rolls out a working AP2. How's that ok? Shouldn't this have been tested in the wild in various forms before moving from AP1 to AP2? I see how some people can say Tesla is all marketing -- they made a BIG SHOW about AP2 but have yet to deliver on it. Almost seems like the self driving car shown in their presser was THE ONLY working car...

My more than .02...

 

[Viking1]

I disagree. Took delivery of my 14 P85 in December 2014, AP1 HW rolled out in January/Feb 2015 if my memory serves correctly, and it just plain worked out of the box. I know there were many posts about people having issues with AP1 staying in lanes, etc... but for my use of it -- and how I use it today -- it worked out of the box: mostly divided highways, freeways, normal surface streets (meaning not a curved mountain road, or residential street with 15mph curves. All at speeds much greater than 35mph btw.

I don't have anywhere near your experience with AP1. However, I did some test driving using AP1 car in Sep. 2016. I recall it working fine on the freeway, but scared me a few times on the surface streets. I had to take control.

In Oct. 2016 I ordered my car. I received it in Dec. 2016 with HW2. It is not working as well as my test drives, but certainly my test drive car, from the viewpoint of a new Tesla driver, was scary at times. It would drift very into another lane, which was occupied by another car. Clear day, straight road, road markings clear, around 40 mph.

 

[lakerholic]

I don't have anywhere near your experience with AP1. However, I did some test driving using AP1 car in Sep. 2016. I recall it working fine on the freeway, but scared me a few times on the surface streets. I had to take control.

In Oct. 2016 I ordered my car. I received it in Dec. 2016 with HW2. It is not working as well as my test drives, but certainly my test drive car, from the viewpoint of a new Tesla driver, was scary at times.

It's definitely something to get used to -- and requires driver attention -- but I have loved every single minute of it.

 

[Saghost]

I disagree. Took delivery of my 14 P85 in December 2014, AP1 HW rolled out in January/Feb 2015 if my memory serves correctly, and it just plain worked out of the box. I know there were many posts about people having issues with AP1 staying in lanes, etc... but for my use of it -- and how I use it today -- it worked out of the box: mostly divided highways, freeways, normal surface streets (meaning not a curved mountain road, or residential street with 15mph curves. All at speeds much greater than 35mph btw.

Admittedly, I have zero experience with AP2, but if the AP1 hardware roll-out was as slow and error filled as what I have been reading on these forums, I would not have been happy with what I thought I was receiving. I would think AP2 owners would be even more frustrated given the decent (I say excellent) response rate of AP1, vis-a-vis those who got AP1 initially, as we had nothing to compare against.

I said this once before and was quickly derided, but I am of this opinion until Tesla proves me otherwise: Tesla engaged MobileEye on AP1 software because this is what MobileEye does. It's the only thing they do. Then ego or threats of lawsuits or a power struggle occurred and now Musk is left reinventing an already pretty good wheel, instead of adding to said pretty good wheel. Which has resulted in the poor roll-out of AP2 we see today.

I have no doubt Tesla will get it right eventually, but at what cost? Pre September 16, you could've bought an AP1 car and used its functionality with minimal issue. Post September 16, your car has reverted to a pre September 14 Tesla in terms of functionality until such time as Tesla fixes all issues and rolls out a working AP2. How's that ok? Shouldn't this have been tested in the wild in various forms before moving from AP1 to AP2? I see how some people can say Tesla is all marketing -- they made a BIG SHOW about AP2 but have yet to deliver on it. Almost seems like the self driving car shown in their presser was THE ONLY working car...

My more than .02...

I believe you're mistaken.

Autopilot was announced in October of 2014, and hardware was showing up on cars a week or two before the announcement (there are some interesting threads about the hardware from before the announcement. Cars in the fall of 14 were a very mixed bag - some with, some without.

However, at launch it only read speed limits and controlled the high beams. Other features arrived piecemeal, and it wasn't until September of 2015 that FW 7.0 turned on the Autosteer component for the first time ever. That's the "year later" folks are talking about. It was probably another 6 months before it became anything resembling safe and reliable.

 

[bmah]

I believe you're mistaken.

Autopilot was announced in October of 2014, and hardware was showing up on cars a week or two before the announcement (there are some interesting threads about the hardware from before the announcement. Cars in the fall of 14 were a very mixed bag - some with, some without.

However, at launch it only read speed limits and controlled the high beams. Other features arrived piecemeal, and it wasn't until September of 2015 that FW 7.0 turned on the Autosteer component for the first time ever. That's the "year later" folks are talking about. It was probably another 6 months before it became anything resembling safe and reliable.

That's closer to my recollection than what @lakerholic wrote. My car was an April 2015 build, with AP1 hardware included, but it wasn't until September or October 2015 that software 7.0 rolled out and activated autosteer. I think another feature that AP cars had at hardware launch was the lane departure warning. We could quibble about the "6 months...safe and reliable" (for me it was a shorter time, some people would say it's still not there now!), but otherwise I think you got the timeline right.

Bruce.

 

[Saghost]

That's closer to my recollection than what @lakerholic wrote. My car was an April 2015 build, with AP1 hardware included, but it wasn't until September or October 2015 that software 7.0 rolled out and activated autosteer. I think another feature that AP cars had at hardware launch was the lane departure warning. We could quibble about the "6 months...safe and reliable" (for me it was a shorter time, some people would say it's still not there now!), but otherwise I think you got the timeline right.

Bruce.

I'll happily defer to your personal experience on that point - as you can likely tell from the language, that's a vague point in time I picked without serious analysis based on my recollection of forum threads.

My only goal with that part of the post was to point out that even after the year of development, when Autosteer first launched it wasn't what it is today and likely wasn't much better than AP2 is now.

Aside from a couple short test drives, my first real experience with AP1 is from August, under 7.1, which was pretty good but easily flummoxed by unusual circumstances (showing it to people on the second day I had the car, I was in a freeway lane that opened into two lanes, and it happily stayed centered as the lane opened and straddled the line when the line started - and the driver's screen showed it on the line.)

It's gotten much better since then, of course.

 

[BertL]

IMHO, perhaps there is a little wiggle room to override Autosteer by applying just the right amount of pressure on the wheel. That's great, but I've never experienced it myself. I suppose I consider it too much trouble trying to maintain Autosteer like that for the sake of doing so vs. taking steering control back myself. The other thing is, when I'm on Autosteer and traveling fine down my lane, and have several times been drawn either very quickly (underneath) towards the semi immediately to my right, or towards a semi just ahead (well less than 3 car lengths) to my right at 65-75 MPH, I have NO TIME to fool around, attempt a subtle tweak of the wheel, and pray Autosteer would keep me from crashing. As I've done, it's far better for me to take more assertive immediate control turning the wheel and bumping Autosteer OFF in the process, with a higher probability of keeping me safe. I do the same with medians on my left that Autosteer begins to hug closer than I desire. It's not worth my having to work harder or increase my anxiety level, with something that is supposed to be a driver assistance tool. To each their own in that regard.

As to the Bernoulli Effect, I get it, but I disagree even with @Ingineer on this one. I'm not speculating, but talking from experience in the past week with 5-6 8.0 truck lust situations that occurred only with Autosteer ON, and never with TACC alone in broadly similar environments. When Autosteer and TACC are OFF, in almost 1-year having AP enabled on my MS, I have never felt any sort of pull in my MS towards even a few semis in any conditions like I've recently experienced. What happened to me a few days ago felt extremely controlled and deliberate -- not how external more natural and random forces like high wind gusts or even the Bernoulli Effect with a semi that passes you at higher speed can sometimes do. In both of those latter cases, the feel on the wheel and vehicle I'm in, is more like some external random force is at work -- sometimes a build-up/down of power, or short and longer bursts -- not something that is immediate and sustained, which felt exactly like HAL or Skynet has taken over with intent, as I've encountered with "8.0 truck lust" a few times this past week. Don't get me wrong -- I'm not saying Tesla did something malicious or deliberately bad with 8.0, only that the resultant way my MS has behaved a few times appeared much more precise than what Mother Nature can generally do on it's own. Despite other's well-meant supposition, I remain convinced there is an Autosteer programming bug related to "truck lust" that is more pervasive in 8.0 than it may have been previously.

BTW on the anecdotal front: As I said upstream, my MS has been in for annual service since Tuesday morning, and has been parked in the same spot outside the SvC since waiting to be worked on. Tuesday night I received a firmware update ready for installation notice, but it was not installed. Last night, another firmware update became ready for installation. IDK what either entail, but as I've said to a couple of folks privately, somethings afoot.

Truck Lust Revisited - 5 Months Later

I mentioned last Fall (see above post from this thread) when I was running 2.36.108 on my HW1 AP1 MS, several scary situations with Autosteer on a road trip. We now more generally describe those situations as "truck lust", riding too close to adjacent vehicles when it's wide-open on the other side, and what I witnessed as a situation with Autosteer out-of-the-blue heading towards a lone cactus/tree on the side of the road when nothing else was around. I documented the situations in detail for Tesla, and at my last SvC visit, they checked and recalibrated my radar, sensor, and camera as need be.

I just returned from my first road trip over the same course of several hundred miles one-way primarily on California/Nevada interstates, at the same times of day (pre-dawn to late morning), and torrential rain for the first part of both outbound legs. I'm now running 2.52.22 and decided to get a full nights sleep to be fresh, and try Autosteer for most of the trip as a comparison for myself of what's changed in 5 firmware releases in the same number of months on my MS. FWIW, my observations:

• I had none of the significant truck lust situations like last Fall, as my MS approached semis ahead on the right. That was great. But, there were other lesser, new-to-me, related anomalies:
  • I could not figure out an exact pattern to it, but my MS at times wanted to hug semis to my right even with wide-open spaces to the left. Perhaps half of the semis were open between the axels, but not all were. Autosteer was keeping my MS near the center of the lane, but under certain conditions would then purposely begin to hug the semi to my right, or when a cement construction barrier was on my left (with almost no clearance from the lane to the barrier.) I almost always pulled my MS back into it's lane manually when this occurred.
• Changing lanes with Autosteer seems better, but is still problematic. My MS started making the lane change, speeding up a bit and completing the action within a reasonable timeframe most of the time, but several times I physically took control to shift to the next lane myself because it seemed to stall or be extremely slow accomplishing the task waiting for something to happen. I didn't note any condition that should have prevented Lane Change from doing so in any of those situations. The problem is exacerbated (to the point IMHO it's not reasonable to use if there is any traffic around), when I tried to do a lane change on a sweeping curve (which are pretty normal on interstates).
• I still love Autosteer in interstate traffic. It does a good job and reduces some of the stress related to stop-and-go traffic. As discussed, it's likely just a limitation of how far the radar can see ahead, but I still find Autosteer a lot more jerky than if I were manually driving with conditions of large open areas quickly opening up, MS speeding up, then having to slow back down again as traffic is really stopped up ahead...
• Autosteer did a better job staying towards the center of it's lane most of the time, and shifting slightly one way or the other if a vehicle in an adjoining lane was too close to our joint lane divider. Much improved. Great job.
  • OTOH, I could not recreate on-demand or explain a pattern to it, but well more than half a dozen times, my MS all of a sudden started moving towards the left lane divider when the IC guidelines were still in place and there were no vehicles on my right. In some of those instances, MS actually went onto the road bumpers or onto the white line (and showed it as such on the IC) before I pulled MS back into my lane. I don't know if it would have kept going or not. Very odd.
  • Autosteer seems to do an even better job of not trying to take unintended off-ramps as others have reported. Great.
  • I still question the logic when Autosteer has a left lane marker (blue on the IC), but temporarily looses the lane markers for a period of time on the right (as in opening additional right lane, turnouts, etc), why Tesla does not have Autosteer hug more of the left lane marker. As it is, my impression is Autosteer tends to try and find the middle of the temporarily wider lane, which feels like the car is going where it shouldn't, before it self-corrects.
  • I had several instances where I was traveling in early morning with bright sun to my left, and Autosteer was keeping MS towards the center of the lane with clearly marked boundaries, but all of a sudden would tend to seek left-to-right/fight-to-left for a short time, then come back to normal. It seems there may be some anomalies in the logic perhaps as the camera is perhaps temporarily blinded. IDK.
• HW1 sensor limitations for traffic that likes to ride on your left or right side IMHO will just remain problematic. IMHO it's something drivers must just be aware of. My aftermarket wide-angle stick-ons for the side mirrors helped some. I expect HW2 should eventually see great improvement with this.
• Autosteer in construction zones remains problematic. I generally turn it off when entering such an area especially if there are concrete barriers (especially to my left) or when I hit the first patch of uneven pavement. Autosteer is just too squirly in those conditions and increases my anxiety level.

My net: Overall Autosteer has improved over the last 5 months, but I still don't consider it anything that reduces my need for being extremely attentive and keeping at least one hand on the wheel at all times. I'll never fully trust it. It is beta. It is NOT fully autonomous ...and Tesla doesn't give it's owners much of a clue what things may have changed from release-to-release, so it's hard for me to be comfortable with what it may or may not do in the heat of the moment.

 

[WannabeOwner]

several times I physically took control to shift to the next lane myself because it seemed to stall or be extremely slow accomplishing the task waiting for something to happen. I didn't note any condition that should have prevented Lane Change from doing so in any of those situations

I have that too. Subjective-guess but I would say I intervene [on lane change] 30-50% of the time. When there is no other traffic about I leave it a decent while (probably 3-5 seconds), to see if it will move, but if it doesn't move in 3 seconds I'm pretty sure it doesn't do so before I decide to take over.

I've been meaning to try it on an empty road - indicate, rinse&repeat until it takes more than 3 seconds to decide to move, then just leave the turn-signal on and time how long it actually takes. I'd be interested to know if it actually changes lanes after 10, 20, ... seconds - obviously that would be useless in Real World but it would tell me if the software is actually trying to do it, or not (if "not" a "Can't do it" chime etc. might be helpful, pro temp ...)

 

[Bet TSLA]

Hey, @HankLloydRight, if you get a chance to mess with ev-fw between scuba dives, I'd like to claim my first reporter's right to name 17.8.16 "Fanboy's Delight". Thanks!

 

[phaduman]

It's definitely something to get used to -- and requires driver attention -- but I have loved every single minute of it.

adding my bit:
I know exactly where to use the AP1 autopilot. Get into the freeway, have a long 16mile commute, pick the carpool lane, double-pull the autopilot stick. Car takes control of steering, brakes and accelerator: and I am watching, quite relaxed but attentive. Best use case of AP I can think of. Any other road - I can probably have TACC on, but never AP. I prefer to drive fun roads on my own. If AP2 cars are delivering this much (freeways only), then they have met the same level of AP as AP1 cars - and I would be quite impressed.

My car was built Sept 16 (AP1). Can't be happier - AP1 really helps me where I need it the most.

 

[lakerholic]

My apologies. My recollection was definitely off.. did not get working AP1 til 10/15 (had to check my phone for a video I made of the experience). But my sentiment remains the same: I believe AP2 was rushed out simply because AP1 cars were out there and Tesla couldn't take a step backwards with waiting for hardware/software bugs to be worked out... but that has ended up happening anyway.

Along the lines of my main point: I do wonder if Tesla/MobileEye could have salvaged their relationship, if AP2 could've rolled out more smoothly with AP2 hardware cars running AP1 software until the bugs were worked out and a software update made those cars with AP2 even more amazing... oh well, I guess we'll never know.

Thanks for not bashing me for my poor memory!

 

[Saghost]

Along the lines of my main point: I do wonder if Tesla/MobileEye could have salvaged their relationship, if AP2 could've rolled out more smoothly with AP2 hardware cars running AP1 software until the bugs were worked out and a software update made those cars with AP2 even more amazing... oh well, I guess we'll never know.

Woulda coulda shoulda. Things definitely would have been different if that relationship hadn't died. There's clear physical evidence that the X had an upgrade path planned into it with dual cameras that never happened, and I've read that Tesla had plans to include both the new vision hardware and the Mobileye chips in the first generation of AP2 cars, which would have let them make a seamless transition.

But in another six months or a year it'll all be irrelevant. eAP on AP2 hardware will equal and then surpass AP on first generation hardware, and then Tesla will start reaching for FSDC. Then I'll have to think about just how much it'll cost me to upgrade my X (which will probably mean trading in, but they took away my favorite colors and interiors. )

 

[phaduman]

It wasn't too long ago, so people might remember....

Apple was sick of google maps popularity on apple phones (knowledge google had of apple users, money that was going to google on an apple hardware). So apple said, hell with you google maps, I will make apple map - it would be easy - no big deal. Got a team together, and released apple maps into all apple phones. Result: gross underestimation of the power and complexity of google maps software and typical apple arrogance. The maps lacked accuracy, depth, eco-system etc etc. People got stranded with apple maps direction...Senior executive fired, apple released statement suggesting people to use google maps, and now, even 2-3yrs later, apple users still do not trust apple maps. It should be a Harward business review topic if it isn't yet.

Replace Apple with Tesla, Steve Jobs with Musk, and google with MobileEye. They are eerily similar. We are all Tesla fanboys, so we want Tesla to succeed in making Tesla AP2 H/W & S/W better than MobileEye/AP1 h/w and s/w. But let's not underestimate the singular focus MobileEye had in developing AP1.

 

[DJ 240V]

All - Would love to get your opinion on this safety poll : Do you feel safe? [ Calling all HW2 EAP pilots.]

 

[Bet TSLA]

It wasn't too long ago, so people might remember....

Yeah, I remember. Most of what you write about Apple is incorrect. I have little doubt that your opinions on Tesla are just as misguided.

 

[DJ 240V]

Yeah, I remember. Most of what you write about Apple is incorrect. I have little doubt that your opinions on Tesla are just as misguided.

What about it is misguided ?

 

[WannabeOwner]

I've been meaning to try it [experiment with lane change delays] on an empty road

Hahaha ... 200 mile highway journey last night, on a regular route of mine, and the car didn't miss a single lane change, neither my deliberate empty-road experimental ones nor normal in-traffic driving. One of the lane changes took maybe 5 seconds, but all the rest were sub 3-second. No recent software updates, So bang goes my theory ...

I'll make some more, considered, experiments at different times of the day.