Tesla's autopilot, on the road

[GRA]

https://electrek.co/2018/04/24/tesla-mo ... eleration/


Let the FUD begin. I'm going to bet that the logs again show this person hit the wrong pedal as we see repeatedly in these cases but I'm sure it will be smeared about to try and drive the stock down. My biggest concern with EVs is the people driving them not the systems. Add this to the 16,000 instances yearly of sudden acceleration due to the wrong pedal being depressed. This will be tagged into AP stories to create more confusion with the masses.

I wouldn't take that bet, as you're almost certainly correct. OTOH, as with any such claim it does need to be investigated and the data made available to make sure that it was in fact driver error and not a car malfunction.

[GRA]

Via IEVS:

18-Month Ban For Leaving Driver Seat Vacant While On Tesla Autopilot

https://insideevs.com/driver-in-uk-was- ... autopilot/

There apparently are some individuals in this world who don’t care much for their life or the lives of others and, unfortunately, some of those individuals are driving on roads.

Bhavesh Patel, 39, was caught on video when sitting in the passenger seat in a Tesla Model S on the highway, enjoying Autopilot driving. But Tesla Autopilot is not autonomous and even if it was, there’s no law in the UK allowing one to drive autonomous cars on public road. For the next one and a half years, he will not be allowed to drive any car at all. . . .

“As well as the 18-month driving ban he was ordered to carry out 100 hours of unpaid work. He was also put on a 10-day rehabilitation programme and will have to pay £1,800 in costs.”

There are definitely times when I wish the U.S. had the same attitude towards dangerous driving and DUIs as the U.K. does.

[RegGuheert]

In another case of Tesla not meeting their promised delivery schedule, they are now paying Autopilot 2.0 customers cash as a settlement for failing to provide OTA updates which provide the promised features:

The lawsuit itself emerged due to some Tesla owners growing frustrated at the automaker missing deadlines for the rollout of Autopilot 2.0 features.

In October 2016, Tesla introduced a new Autopilot hardware suite, dubbed Autopilot 2.0, and promised that it would enable a series of new features to be released to owners via over-the-air software updates – eventually leading up to “fully self-driving capability”.

The cars equipped with this hardware suite were available with two packages: a $5,000 “Enhanced Autopilot” package, which promised more advanced features building on the first generation Autopilot’s features, and a $3,000 “Fully Self-Driving” package, which was sold on top of the “Enhanced Autopilot” package.

The suite was also supposed to enable several active safety features, which are standard regardless of if any Autopilot packages are added to the vehicle.

After bringing the new hardware to production, Tesla’s goal was first to release software upgrades to reach feature parity with the first generation Autopilot by December 2016 – just a few months after releasing the hardware – and to continue releasing new software updates to eventually lead to fully self-driving capability.

[RegGuheert]

One question: Since the Tegra (X1?) processor doesn't not have a "deep-learning accelerator" like is found on the upcoming Xavier processor, are the neural networks in the Tesla Autopilot all implemented in the GPU?

The answer is yes given the extensive processor demands, i.e. the enormity of data, of the neural network system being implemented
by Tesla for AP without using LIDAR. Tesla most likely will switch to the later Nvidia processor when available as system demands approach
AP 5. The Nvidia stock was a great buy at about $80 -$90 (now $245) when Tesla switched from Mobileye.

It looks like you made a good call on this lorenfb: Tesla could have to offer computer retrofits to all Autopilot 2.0 and 2.5 cars by the end of next year.

Several comments made by CEO Elon Musk since the launch of its Autopilot 2.0 hardware suite in all Tesla vehicles made since October 2016 indicate that the company might have to update its onboard computer in order to achieve the fully self-driving capability that it has been promising to customers.

Now it looks like Tesla might have to also offer computer retrofits for Autopilot 2.5 cars.

Now Musk said during Tesla’s Q1 conference call last week that even this computer, which is the one being installed in vehicles currently in production, might also need to be replaced to achieve fully self-driving capability:

In order for that to be in place, we have to obviously sell full autonomy and we’re making really good progress on that front. I believe that the vehicles that we are currently producing are capable of full autonomy with the only thing that maybe is probably needed is a computer upgrade to have more processing power for the vision neural net. But that’s a plug-in replacement, a thing that can be done quite easily.

Tesla indeed made the onboard Autopilot computer easily swappable if they need to upgrade, but even if the task to replace it is somewhat easy, it can still add up to quite a significant retrofit program for the automaker.

[edatoakrun]

Not only is TSLA increasingly being portrayed as a loser in the race to autonomy, its continued insistence that lidar is unnecessary puts it trailing behind the leaders in the Unusual Cases and Dark Horses category near the end of this report:

Who’s Winning the Self-Driving Car Race?

A scorecard breaking down everyone from Alphabet’s Waymo to Zoox....

Unusual Cases and Dark Horses...Where things get murky is that Musk eschews the Lidar systems that most carmakers and tech companies are using. He says he wants to develop more advanced imaging to give his cars a much better pair of eyes.

Musk wants to use cameras and develop image-recognition capabilities so cars can read signs and truly see the road ahead. He has said Tesla is taking the more difficult path, but if he can come up with a better system, he will have mastered true autonomy without the bulky and expensive hardware that sits on top of rival self-driving cars.

“They’re going to have a whole bunch of expensive equipment, most of which makes the car expensive, ugly and unnecessary,” Musk told analysts in February. “And I think they will find themselves at a competitive disadvantage.”

Analysts from BNEF project that Tesla will be able to field Level 4 cars in 2020, although that timetable could be subject to change now that the company entered into a public spat with federal safety investigators over the fatal crash involving Autopilot...

https://www.bloomberg.com/news/features ... g-car-race

[GRA]

^^^Seeing as how the vast majority of the Bloomberg article is about non-Tesla autonomy, shouldn't the preceding post be in the "Autonomous Vehicles" topic rather than this one?

[lorenfb]

One question: Since the Tegra (X1?) processor doesn't not have a "deep-learning accelerator" like is found on the upcoming Xavier processor, are the neural networks in the Tesla Autopilot all implemented in the GPU?

The answer is yes given the extensive processor demands, i.e. the enormity of data, of the neural network system being implemented
by Tesla for AP without using LIDAR. Tesla most likely will switch to the later Nvidia processor when available as system demands approach
AP 5. The Nvidia stock was a great buy at about $80 -$90 (now $245) when Tesla switched from Mobileye.

It looks like you made a good call on this lorenfb: Tesla could have to offer computer retrofits to all Autopilot 2.0 and 2.5 cars by the end of next year.

Several comments made by CEO Elon Musk since the launch of its Autopilot 2.0 hardware suite in all Tesla vehicles made since October 2016 indicate that the company might have to update its onboard computer in order to achieve the fully self-driving capability that it has been promising to customers.

Now it looks like Tesla might have to also offer computer retrofits for Autopilot 2.5 cars.

Now Musk said during Tesla’s Q1 conference call last week that even this computer, which is the one being installed in vehicles currently in production, might also need to be replaced to achieve fully self-driving capability:

In order for that to be in place, we have to obviously sell full autonomy and we’re making really good progress on that front. I believe that the vehicles that we are currently producing are capable of full autonomy with the only thing that maybe is probably needed is a computer upgrade to have more processing power for the vision neural net. But that’s a plug-in replacement, a thing that can be done quite easily.

Tesla indeed made the onboard Autopilot computer easily swappable if they need to upgrade, but even if the task to replace it is somewhat easy, it can still add up to quite a significant retrofit program for the automaker.

Thanks for the feedback and update. It''s not easy to forecast present processor requirements based on future software developments.

[GRA]

Yet more evidence that AP still has no business being allowed to be used by customers on two-way roads, via IEVS:

Watch Tesla Autopilot Cross Double Yellow, Drive Wrong Way

https://insideevs.com/watch-tesla-autop ... wrong-way/

[jlv]

Yet more evidence that AP still has no business being allowed to be used by customers on two-way roads

Yet more evidence that you need to use AP the way it is designed to be used, with both hands on the wheel and paying attention. There's a reason it tells you this every time you activate it.

[GRA]

Yet more evidence that AP still has no business being allowed to be used by customers on two-way roads

Yet more evidence that you need to use AP the way it is designed to be used, with both hands on the wheel and paying attention. There's a reason it tells you this every time you activate it.

Yet the public will continue to use it improperly, and Tesla has no business allowing it to be used in situations where it is manifestly inadequate as has been shown numerous times ever since it was introduced. It continues to cross over centerlines into oncoming traffic despite 2.5 years of customer experience and development. Regulators need to step up, since Tesla obviously isn't going to do so. To date Tesla's been fortunate in that only its own customers have been injured or killed while on A/P (the 101 crash could easily have been the first case, if either of the two cars which hit the Model X or its debris had resulted in serious injury or death to their occupants), but the large increase in numbers of A/P-equipped cars as the Model 3 enters the market will mean far more opportunities for A/P to hurt someone who isn't in the Tesla, and major lawsuits and/or regulation will inevitably follow. If the intention is to seriously delay the deployment of all AVs, Tesla is taking the right approach.