Firmware 8.0

[sandpiper]

Elons statements were a bit contradictory on this point. He then goes on to say:

They get very comfortable with it and repeatedly ignore the car’s warnings. It’s like a reflex. The car will beep at them, they tug the wheel, the car will beep at them, they tug the wheel, and it becomes an unconscious reflex action. So we will see half a dozen or more, sometimes as many as 10 warning in one hour continuously ignored by the driver. We really want to avoid that situation.

Which implies "ignoring" might mean having the audio cue come up three times (I.e. "Ignoring" the visual cue). It's hard to say. How else can you "ignore" 10 warnings now and not have autopilot go into emergency stop mode?

No... by "ignoring" he means that they tugged the wheel but didn't continue to hold onto it. The warning says "Hold the wheel" - not "give the wheel a quick tug".

This is Tesla's very carefully orchestrated way of backing away from the current technique where it will only ask you to hold-on when it's getting confused. We're going to be getting timed nags.

 

[Az_Rael]

No... by "ignoring" he means that they tugged the wheel but didn't continue to hold onto it. The warning says "Hold the wheel" - not "give the wheel a quick tug".

This is Tesla's very carefully orchestrated way of backing away from the current technique where it will only ask you to hold-on when it's getting confused. We're going to be getting timed nags.

I agree, I think the system is going to require hands on wheel to use for any length of time. We will see when the rest of the transcript releases, but it sounds like Tesla is going to fairly short timed nags based on the reporting.

 

[Panu]

If they will require torque on the wheel almost all the time unfortunately it encourages to solutions like one in this Mercedes:

 

[sandpiper]

If they will require torque on the wheel almost all the time unfortunately it encourages to solutions like one in this Mercedes:

It will. But at that point any legal argument that somebody has against Tesla goes out the window. This is equivalent to the old "brick on the accelerator" cruise control. And lest you think that you could pitch the weight out of the window if something happened... I assure you that Tesla logs the torque on the wheel. They will be able to tell the difference between a hand and and constant hanging weight.

 

[technomation]

In a different blog, tweet, or interview (I've lost track at this point), Elon mentioned working on enhanced autopilot hardware that was "months" away. B/c of Elon time, I would imagine that is over a year away (from then). I'm not surprised they are working on "the next thing" b/c it takes times to properly vet and test these things out. In that process, they likely wrote some new code on handing the information and that led to thinking on smoothing out the data over time, which then could be back-ported to the current hardware.

Now this is just theorizing on my part, but totally could be how we got to the new Autopilot software for 8.0

Might LiDAR be added to the radar and camera for the AP 2.0 hardware after all? Elon’s well publicized opinion against LiDAR is that it’s not necessary, radar with a camera is adequate, and that LiDAR doesn’t work well in rain, snow, or fog. LiDAR is also prohibitively expensive (check out Velodyne’s website, $8,000 a puck!). Nonetheless, Teslerati just issued a Sep 9 report with spy shots of a Model S with a couple overhead LiDAR pucks, which they assumed were from Velodyne (they’d already reported on July 9 a Model S with a huge LiDAR tophat, Stanford insignia on the back windshield).

A privately held SV startup, Quanergy, has a svelte LiDAR sensor solution, M8-1, which they claim “resists false returns by dust, mist, rain, or snow,” which was one of Elon’s objections to LiDAR. The M8-1 is unique because it’s solid state, and Quanergy claims it’s more performant and cost effective than existing LiDAR solutions. Possible Elon mindchangers, but Quanergy and Velodyne will have to get the price down into the $100’s, which they say should happen by 2018. That puts AP 2.0 a ways off. Just speculating…

 

[Stoneymonster]

Might LiDAR be added to the radar and camera for the AP 2.0 hardware after all? Elon’s well publicized opinion against LiDAR is that it’s not necessary, radar with a camera is adequate, and that LiDAR doesn’t work well in rain, snow, or fog. LiDAR is also prohibitively expensive (check out Velodyne’s website, $8,000 a puck!). Nonetheless, Teslerati just issued a Sep 9 report with spy shots of a Model S with a couple overhead LiDAR pucks, which they assumed were from Velodyne (they’d already reported on July 9 a Model S with a huge LiDAR tophat, Stanford insignia on the back windshield).

A privately held SV startup, Quanergy, has a svelte LiDAR sensor solution, M8-1, which they claim “resists false returns by dust, mist, rain, or snow,” which was one of Elon’s objections to LiDAR. The M8-1 is unique because it’s solid state, and Quanergy claims it’s more performant and cost effective than existing LiDAR solutions. Possible Elon mindchangers, but Quanergy and Velodyne will have to get the price down into the $100’s, which they say should happen by 2018. That puts AP 2.0 a ways off. Just speculating…

I think they'll go with whatever is technically the best solution that is reasonable to price into the car, regardless of what Elon is saying right now. Having said that, seeing test Model Ss with all manner of various sensors is indicative of nothing. The LIDAR on those cars could be providing ground truth for testing the new radar solution, for instance.

 

[Saghost]

Might LiDAR be added to the radar and camera for the AP 2.0 hardware after all? Elon’s well publicized opinion against LiDAR is that it’s not necessary, radar with a camera is adequate, and that LiDAR doesn’t work well in rain, snow, or fog. LiDAR is also prohibitively expensive (check out Velodyne’s website, $8,000 a puck!). Nonetheless, Teslerati just issued a Sep 9 report with spy shots of a Model S with a couple overhead LiDAR pucks, which they assumed were from Velodyne (they’d already reported on July 9 a Model S with a huge LiDAR tophat, Stanford insignia on the back windshield).

A privately held SV startup, Quanergy, has a svelte LiDAR sensor solution, M8-1, which they claim “resists false returns by dust, mist, rain, or snow,” which was one of Elon’s objections to LiDAR. The M8-1 is unique because it’s solid state, and Quanergy claims it’s more performant and cost effective than existing LiDAR solutions. Possible Elon mindchangers, but Quanergy and Velodyne will have to get the price down into the $100’s, which they say should happen by 2018. That puts AP 2.0 a ways off. Just speculating…

Anything is possible. That said, Elon reiterated his stance against LIDAR in the just posted part 5 of the transcript from yesterday...

Sorry you blacked out for a second during your question, but we do not anticipate using lidar. Just to make it clear, lidar essentially is active photon generator in the visible spectrum – radar is active photon generation in essentially the radio spectrum. But lidar doesn’t penetrate intrusions so it does not penetrate rain, fog, dust and snow, whereas a radar does. Radar also bounces and lidar doesn’t bounce very well.

You can’t do the “look in front of the car in front of you” thing. So I think obvious thing is to use radar and not use lidar.

Transcript: Elon Musk’s press conference about Tesla Autopilot under v8.0 update [Part 5]

 

[Model S M.D.]

Part 6 of the transcript isn't up yet, but I really do think EM does not want us to do the typical reflex action of hearing an audible nag and tugging the wheel. I can accept that as long as the visual only nag (no noise) and tugging the wheel is deemed okay and won't count against the 3 per hour limit. Additionally, this visual nag (white border, white light, etc.) better be obvious in my peripheral vision. Otherwise, constantly looking down at the IC for a visible nag does not sound very safe....

 

[technomation]

Maybe they are using the Camera HW to do that processing

A post in another thread said that they are, with neural net algorithms.

 

[rog]

Additionally, this visual nag (white border, white light, etc.) better be obvious in my peripheral vision. Otherwise, constantly looking down at the IC for a visible nag does not sound very safe....

You don't have to look down. You just have to keep your hands on the wheel. That's what Tesla (?) wants you to do. They won't implement something that will allow you to not constantly have your hands on the wheel.

 

[rog]

A post in another thread said that they are, with neural net algorithms.

Which would mean they continue to source HW from MobilEye.

 

[ohmman]

I'm loving people being upset at what they speculate. Does that mean you're mad at your imagination?

On disabling Autosteer, my read is:

if the user ignores repeated warning to keep their hands on the wheel, if it ignores the audible alarm more than 3 times in an hour then the driver will have to park and restart in order to enable Autosteer.

So you'll get visual(s) and then audible. It's been said that they're much more prominent. If you, the user, ignore the audible alarm more than 3 times, you have to park and restart. That means if you react to the audible alarm 3 times in an hour, you may be just fine.

 

[BertL]

IMHO, futures are all speculation with little absolutes especially as timeframes increase. Prices, technology, requirements and understanding all evolve over time. Elon says no lidar is in the cards for tomorrow, but does that mean for just the next AP iteration or forever? Let's remember, cameras were primary in the initial AP implementation, but Tesla has changed priorities with 8.x as requirements and tech became better understood. I'm personally very happy to see that especially Elon appears in this instance to have maintained an open aperture with such things, as it's a win for the consumer and moving the tech forward for everyone.

I think all the speculation going on is interesting banter, but since most of us have no inside knowledge, and there will likely be many future AP iterations with hardware and firmware, it's just that -- interesting speculation that may or may not become reality. I'll cast more of a personal opinion or concern as I know what has actually been delivered.

 

[technomation]

Anything is possible. That said, Elon reiterated his stance against LIDAR in the just posted part 5 of the transcript from yesterday...

"...Just to make it clear, lidar essentially is active photon generator in the visible spectrum – radar is active photon generation in essentially the radio spectrum. But lidar doesn’t penetrate intrusions so it does not penetrate rain, fog, dust and snow, whereas a radar does."

Interesting...Quanergy is saying that they have solved the rain/fog/dust/snow problem with LiDAR. They have strong interest by MB, Ford, and others.

 

[Saghost]

Interesting...Quanergy is saying that they have solved the rain/fog/dust/snow problem with LiDAR. They have strong interest by MB, Ford, and others.

I'll be interested to see how they "solved" it, since AFAIK it's basic physics (though there are some near IR bands that go right through water which might work for everything but dust I believe...)

I've been wrong before, and surely will be again.

 

[RubberToe]

I'm loving people being upset at what they speculate. Does that mean you're mad at your imagination?

On disabling Autosteer, my read is:

So you'll get visual(s) and then audible. It's been said that they're much more prominent. If you, the user, ignore the audible alarm more than 3 times, you have to park and restart. That means if you react to the audible alarm 3 times in an hour, you may be just fine.

Sounds like if you ignore it 4 or more times in an hour, you win the prize. So if there are 8 audible alarms in an hour, you would have to respond to 5 to keep the ignored count to the 3 allowable.

A bit different than just responding to 3 an hour. Right?

RT

 

[JohnnyG]

... Nonetheless, Teslerati just issued a Sep 9 report with spy shots of a Model S with a couple overhead LiDAR pucks, which they assumed were from Velodyne (they’d already reported on July 9 a Model S with a huge LiDAR tophat, Stanford insignia on the back windshield)....

There's been a lot of talk over the spy photos of the Model S with the LIDAR pucks. I work in engineering (and closely with R&D) for an automotive manufacturer. While it is very possible that Tesla could start including, or completely move to, LIDAR; the spy photos really don't imply that. It is at least equally, or even more likely, that they are comparing the effects of LIDAR on their vehicles with what they are currently achieving with RADAR. This data can be used to further refine their RADAR approach. It can also be used to compare / contrast the results with RADAR, to be presented to investors and safety boards.

Now, if we started to see a bunch of test mules fitted with LIDAR, that would sway me to believe they are at least considering using LIDAR in production. But it would still be a guess as to whether they would switch from RADAR, or use them in conjunction. Redundancy and complimentary systems could be key to Level 4 autonomy.

 

[technomation]

I'll be interested to see how they "solved" it, since AFAIK it's basic physics (though there are some near IR bands that go right through water which might work for everything but dust I believe...)

I've been wrong before, and surely will be again.

I'd like to know more about how they do it too. Quanergy claims they're on to a whole new generation of LiDAR. Maybe this will shed a little more light (ptp). Solid state S3 enables faster, more accurate laser firing, greater range than radar. Small footprint, NVIDIA Tegra integration, cheaper cost, going into production next quarter. The WSJ article below quotes the founder: "The Quanergy sensor won’t have spinning parts and Mr. Eldada says it will function in rain or snow." I don't know how Tesla could be ignoring this, considering that they can keep their radars and have a vastly improved 360 point cloud at greater range than the radar coarse point cloud, and Quanergy claims that the weather problem is solved. The pdf prezo is especially intriguing .

Quanergy Says It Will Offer Low Cost Sensors for Autonomous Cars Next Year

http://on-demand.gputechconf.com/gtc/2016/presentation/s6726-louay-eldada-quanergy-systems.pdf

From Robotics Trends re Quanergy on NVIDIA deep learning platform (CES 2016):
"One of the Nvidia demos in pedestrian detection combined a Quanergy 8 plane LIDAR and their camera systems. In the demo, they had water jets able to simulate rain, in which case it was the vision that failed and the LIDAR which kept detecting the pedestrians. Quanergy’s LIDAR looks at the returns from objects and is able to tell returns from raindrops (which are more dispersed) from returns off of solid objects."

 

[davidc18]

I keep it at 6.
I don't sweat the bozos who cut in. They usually leave quicky.
Much less stressful.

+1 ^^^^^^^^^ Mine has been set on '7' for over a year.

 

[ohmman]

Sounds like if you ignore it 4 or more times in an hour, you win the prize. So if there are 8 audible alarms in an hour, you would have to respond to 5 to keep the ignored count to the 3 allowable.

A bit different than just responding to 3 an hour. Right?

RT

Sorry, that was terrible phrasing on my part. When I said "respond to 3 an hour" I was imagining only getting 3 in my example.

Currently, ignoring the audible alert disengages Autopilot, right? Is there more than one audible alert? I'm just curious if the 3 number means allowing Autopilot to disengage that many times in an hour. If so, I'd call that too lenient. There's also the matter of how long it takes to respond before it's an "ignored" event.

As I mentioned before, I get the alert with my hands both planted firmly on the wheel and providing at least some level of resistance to the trajectory. My Service Center is investigating, and will hopefully find some way to tweak it. Otherwise, depending on the actual implementation, I could get disabled for receiving 3 audible alerts. I don't ever ignore them, though, so if it's just ignoring, that wouldn't change anything for me, even in my "torque-insensitive" X.