1st drive of 2017 MBZ E-class: lane keeping is STILL inferior Tesla's

[ebwb]

Thank you Spidy for posting the video, its very long and I don't speak German so I scanned through most of it.

I like the crossing traffic test at 1:11:00

Can you tell what speed they're traveling during the test at 1:13:00?

At 1:14:30, what are they demonstrating? The vehicle detects the crossing traffic and determines it won't be a factor or what?

Also, would you mind translating what they're saying during the test at 1:16:30?

Danke!

1:13:00 - I heard the driver say "neunzig", which is ninety. The speedometer also shows 90. So they were likely approaching stationary traffic at 90 km/h (56 mph). The car first chimed the alarm, started "soft" braking, then engaged maximum braking as it approached the faux vehicle in front.

 

[SR22pilot]

For the record, I own TSLA stock and appreciate KZKZ's contributions to this forum. It is troubling how the standard response to criticism in these circles is suspicion, defamation of character, etc. Here's my two cents.

I'll bite on this. My question is whether the person joined TMC out of a sincere interest in the car vs. an agenda to slam the car and manipulate the stock. Most new users go through a process. Some say "hello" while others lurk for a bit. You may see questions such as whether a particular option is worth the money or how much difference upgrading the sound system makes. Others ask questions involving charging on long distance trips. However, on the thread about the broken ball join there was a group of users that shared a common profile. All had joined recently and all had only posted heavily critical posts. They were quick to call for all Teslas to be recalled or to have government action taken against Tesla. It became clear they had an agenda and it wasn't about honest open debate of the issues.

You mention the E class. That is Mercedes' latest system and it isn't even deployed on their other cars yet. The regular Distronic Plus manual lists inability to detect stationary objects as a limitation. I think we all want better lateral intrusion detection and better emergency braking but Tesla is not an outlier in this area. Heck the majority of cars lack any AEB system. It is better to have a limited AEB which mitigated rear end collisions rather than nothing at all but that's not what some people are calling for. Some people rail against Tesla when other cars are much worse. Personally I want to see a move to phased array scanning radar or to scanning LIDAR.

 

[SR22pilot]

Here is a report from the early days of trying to get AEB adopted.
https://circabc.europa.eu/sd/a/3ab87fdc-5715-4733-af50-c3608034ca56/report_aebs_en.pdf
Notice the there were numerous types of accidents not being used since they were deemed difficult to handle:

Rule out additional crash types: Exclude crossing no turns, crossing vehicle turning, merging, right turn against, Manoeuvring. Reason for ruling out other crash types is that they would either leave no time to react and therefore not be able to helped by a current system or the system sensors would not be able to perform in that scenario.​

Now we are much further along and it is great that lateral intrusions are being attacked. I expect Tesla AP 2 to also be much better with a vastly improved set of sensors. Competition in this area will make all vehicles safer. Right now I just wish my car had any form of AEB.

 

[MarkS22]

For the record, I own TSLA stock and appreciate KZKZ's contributions to this forum. It is troubling how the standard response to criticism in these circles is suspicion, defamation of character, etc. Here's my two cents.

To me, it's obvious that the optional system on the new E-Class has more safety capabilities than current Autopilot (I'm sure future iterations of Autopilot will have more hardware and capabilities):

- E-Class can detect lateral, cross path traffic
- E-Class can can detect stationary vehicles from a distance
- E-Class can detect pedestrians

That said, Autopilot is tuned to mimic autonomous driving, as xG35 said, with much better lane-centering abilities than DRIVE PILOT. This creates a sense of user confidence, right up until the moment there's a situation that Autopilot was not designed to address. Tesla can do a better job educating Autopilot users with what the system can and cannot do.

It's really a difference in approach. Tesla's system builds confidence with impeccable lane centering behavior, allowing users to not touch the steering wheel for miles and miles. DRIVE PILOT (no idea why Mercedes capitalizes non-acronyms...) still requires constant human input and nags the driver to hold the steering wheel, which is the more conservative and, in my opinion, the more appropriate approach given current limitations. It certainly doesn't have the same cool factor though, which is what makes Tesla so unique and talked about.

I'm not sure where you're seeing the "standard response," but I own both a Tesla and Mercedes. I've been very clear that Mercedes has better isolated safety features. For example, Pre-Safe is fantastic. It uses rear radar to flash the brake lights and warn a fast approaching car. If they don't stop, it locks the brakes and tightens the seat belts. Tesla doesn't have anything like that. The blind spot detection is vastly superior on the Mercedes. It has more intuitive lights on the mirrors, better detection of fast accelerating cars coming up in the lane next to you, and beeps if you put on your turn signal when there's danger. It's very well implemented. Pedestrian detection and cross traffic (at least for 2016, which they claim to be on) is mixed... not nearly as "wow" as the previous two examples. A large deer darted out across the road this week and nothing went off. (To be fair, it's "pedestrian" and "cross traffic" not "animal.) But, I haven't ever seen it detect or warn about pedestrian in months, including some times where I feel it might have been useful?

What Tesla does much better is TACC/lane keeping with integrated dash UI. There's no comparison there. There's also no argument Mercedes has excellent safety features taking advantage of more sensors. Most of us here can't wait for the next generation of sensors. But for now, and it seems for at least the next year, in terms of TACC/lane keeping (Autopilot vs Drive Pilot) has Tesla coming out in front.

I feel Tesla has explained the capabilities quite well. A single accident (still under investigation) is hardly an indication of a need for mass education after 130 million miles of use. Just my opinion, but I'm very upfront where Mercedes is superior and I'm happy to discuss it here. I have two young children and my wife drives them often in that Mercedes, which is why I appreciate all those safety features. I don't know how much more of an endorsement you can get for their safety technology.

 

[luca]

For the record, I own TSLA stock and appreciate KZKZ's contributions to this forum. It is troubling how the standard response to criticism in these circles is suspicion, defamation of character, etc. Here's my two cents.

To me, it's obvious that the optional system on the new E-Class has more safety capabilities than current Autopilot (I'm sure future iterations of Autopilot will have more hardware and capabilities):

- E-Class can detect lateral, cross path traffic
- E-Class can can detect stationary vehicles from a distance
- E-Class can detect pedestrians

That said, Autopilot is tuned to mimic autonomous driving, as xG35 said, with much better lane-centering abilities than DRIVE PILOT. This creates a sense of user confidence, right up until the moment there's a situation that Autopilot was not designed to address. Tesla can do a better job educating Autopilot users with what the system can and cannot do.

It's really a difference in approach. Tesla's system builds confidence with impeccable lane centering behavior, allowing users to not touch the steering wheel for miles and miles. DRIVE PILOT (no idea why Mercedes capitalizes non-acronyms...) still requires constant human input and nags the driver to hold the steering wheel, which is the more conservative and, in my opinion, the more appropriate approach given current limitations. It certainly doesn't have the same cool factor though, which is what makes Tesla so unique and talked about.

autopilot can also detect lateral, cross path traffic lol

 

[luca]

For the record, I own TSLA stock and appreciate KZKZ's contributions to this forum. It is troubling how the standard response to criticism in these circles is suspicion, defamation of character, etc. Here's my two cents.

To me, it's obvious that the optional system on the new E-Class has more safety capabilities than current Autopilot (I'm sure future iterations of Autopilot will have more hardware and capabilities):

- E-Class can detect lateral, cross path traffic
- E-Class can can detect stationary vehicles from a distance
- E-Class can detect pedestrians

That said, Autopilot is tuned to mimic autonomous driving, as xG35 said, with much better lane-centering abilities than DRIVE PILOT. This creates a sense of user confidence, right up until the moment there's a situation that Autopilot was not designed to address. Tesla can do a better job educating Autopilot users with what the system can and cannot do.

It's really a difference in approach. Tesla's system builds confidence with impeccable lane centering behavior, allowing users to not touch the steering wheel for miles and miles. DRIVE PILOT (no idea why Mercedes capitalizes non-acronyms...) still requires constant human input and nags the driver to hold the steering wheel, which is the more conservative and, in my opinion, the more appropriate approach given current limitations. It certainly doesn't have the same cool factor though, which is what makes Tesla so unique and talked about.

And autopilot can also detect stationary vehicles from a distance and pedestrians

 

[SR22pilot]

autopilot can also detect lateral, cross path traffic lol

This really isn't lateral cross traffic. The vehicle is only totally lateral at the end where the Tesla has stopped already. It is, however, a great example of how useful the present system is.

 

[techmaven]

KZKZ has thrown some marketing literature our way, but is actually very misleading on the facts. He keeps making arguments that this is a sensor or hardware missing problem. It is not. It is a software issue in handling false positives. The hardware, at a sensory level was sufficient. Instead, recognizing the lack of an obstruction at ground level and sorting through the height clearance issues without causing too many false positives is the issue at hand. If the truck has side skirts, the AEB would have altered the circumstances greatly. If the intersection point was the cab or the wheels, even without side view cameras or software that recognizes the size of a tractor trailer, the AEB would have stepped in.

As shipped, the 2017 E-class semi-autonomous system is far more dangerous than the Tesla one simply due to UI design at the moment. As for the performance of the rest of the systems, we would have to do much more testing to know... There is a lot of software tuning that makes most of the difference. Additional hardware makes no difference if the system can't use it in a better way.

 

[KZKZ]

Second, it is not permitted for autonomous driving. It is actually only permitted to test exactly what everyone's tesla is already doing.

"Self-driving tests are permitted on all interstates and state highways in Nevada, human drivers being required only for turning, merging and departing. The autonomous test drives in everyday traffic will be carried out by specially trained test drivers. Nevada Department of Motor Vehicles (NDMV) rules also stipulate that there must be one passenger behind the wheel and a second passenger in the vehicle on test drives."

That's an awesome point that you make, that the E-Class that the State of Nevada certified is testing "exactly what everyone's tesla is already doing."

That's a great perspective of what's going on.

I think this will be just one of the issues NHTSA will be looking into after the fatal AutoPilot accident.

Should average consumers be "testing" beta semi-autonomous features on public roads?

Why does one company go through a certification process using trained test drivers and another company uses the public to test beta software.

 

[KZKZ]

1:13:00 - I heard the driver say "neunzig", which is ninety. The speedometer also shows 90. So they were likely approaching stationary traffic at 90 km/h (56 mph). The car first chimed the alarm, started "soft" braking, then engaged maximum braking as it approached the faux vehicle in front.

Thanks for the translation! Its good to see the vehicle went all the way to 0 autonomously.

 

[KZKZ]

And autopilot can also detect stationary vehicles from a distance and pedestrians

The video you posted states it failed both pedestrian tests by not automatically braking.

Are you saying it does or does not automatically brake for pedestrians?

 

[KZKZ]

fyi

Mobileye released a statement that confirms the limitation of their current system in relation to crossing traffic.

“We have read the account of what happened in this case. Today’s collision avoidance technology, or Automatic Emergency Braking (AEB) is defined as rear-end collision avoidance, and is designed specifically for that. This incident involved a laterally crossing vehicle, which current-generation AEB systems are not designed to actuate upon. Mobileye systems will include Lateral Turn Across Path (LTAP) detection capabilities beginning in 2018, and the Euro NCAP safety ratings will include this beginning in 2020.”

Tesla Autopilot partner Mobileye comments on fatal crash, says tech isn’t meant to avoid this type of accident [Updated]

 

[KZKZ]

I agree on the software and big data part. I disagree about MB et. al. going for the home run. On paper the latest MB system does everything Tesla's AP does except lane change. Actually it does more "on paper." It just doesn't stay in the lane as well. I think blind spot detection may be better than Tesla and I suspect auto braking is equal.

The MB does lane change too.

It's interesting to see MB's approach to the issue. Mercedes approach is heavy with additional sensors. It includes a rear and side facing multi-mode radars in addition to the normal ultra-sonic sensors. Mercedes also calls the system Lane Change ASSIST.

 

[trils0n]

Tesla will certainly add more sensors for their next gen autopilot. If that is important to you, you should wait for the next generation, as it will undoubtedly be more capable. This is Tesla's first gen autopilot setup. As I understand it, the new 2017 E-Class hardware is a step up from the recently redesigned 2016 S-Class, which had more sensors/capability than MB's previous version. So at least the 3rd iteration, right?

I think what is impressive is that Tesla has done so much with such a limited suite of sensors. The performance exceeds cars with better sensor suites in a few areas. Can't wait to see what they bring out in their next gen version with more sensors.

Will MB upgrade the software so actual capabilities can more closely match the potential of the extensive sensors? If MB isn't going to get the performance up to the promise of it's sensors, it seems like quite the wasted opportunity.

 

[SR22pilot]

I say kudos to MB for their sensor suite. Competition is great. This enhanced sensor suite was released after Tesla's gen 1 AP system. For the most part manufacturers have been ignoring Tesla. As consumers we are better off when competition is taken seriously.

 

[Drivin]

an agenda to slam the car and manipulate the stock.

That is the most hysterical comment I have read here! Thanks for that!

Sure, someone can post an unsubstantiated anonymous comment in a forum (ranked all the way in the top 35,000 websites!) and it will result in a sufficiently large decline of a $20B highly traded stock so the person can profit from it enough while escaping the scrutiny of the SEC.

Sure. That makes total sense. LOL.

Now of course that is totally different from the people who would post how much they love the company/stock even though their X has been back in the shop 3 times - they would never do that because they wanted to keep the stock up. LOL.

There are also people who whine about negative posters because they feel that they will drive the stock down, or profit from it, so they play kill the messenger because they think that will actually help prop up the stock. Those people are far more damaging to the brand.

 

[xav-]

fyi

Mobileye released a statement that confirms the limitation of their current system in relation to crossing traffic.

“We have read the account of what happened in this case. Today’s collision avoidance technology, or Automatic Emergency Braking (AEB) is defined as rear-end collision avoidance, and is designed specifically for that. This incident involved a laterally crossing vehicle, which current-generation AEB systems are not designed to actuate upon. Mobileye systems will include Lateral Turn Across Path (LTAP) detection capabilities beginning in 2018, and the Euro NCAP safety ratings will include this beginning in 2020.”

Tesla Autopilot partner Mobileye comments on fatal crash, says tech isn’t meant to avoid this type of accident [Updated]

Mobileye is so advanced it can detect a turd 200 feet ahead.. But a massive truck crossing 50 feet ahead? Nope

 

[Saghost]

And autopilot can also detect stationary vehicles from a distance and pedestrians

Am I wrong in thinking that the car sounds the alert for a little while before the emergency braking kicks in?

You got the alert, which means the car saw him and the danger - my guess is the car would have hit the brakes pretty late, after the point you decided to intervene this time.

Having said that, I'm certainly not suggesting rerunning the test and waiting - you're already a lot closer to hitting someone than I'm comfortable with, and I could be wrong.

 

[cronosx]

Mobileye is so advanced it can detect a turd 200 feet ahead.. But a massive truck crossing 50 feet ahead? Nope

detect is one thing, act.. is another thing.
You can detect one thing, but not being sure it need to be addressed, and when you aren't sure, it's best not to act. it's counter intuictive, but acting in a false-positive case is much wrost than not doing something.
So, yes, surely they detected it, but at the moment they weren't sure it needs to be addressed.
They will address more case every new iteration and in due time they will address 99.9% of the cases, but just not yet

 

[SR22pilot]

That is the most hysterical comment I have read here! Thanks for that!

Sure, someone can post an unsubstantiated anonymous comment in a forum (ranked all the way in the top 35,000 websites!) and it will result in a sufficiently large decline of a $20B highly traded stock so the person can profit from it enough while escaping the scrutiny of the SEC.

Sure. That makes total sense. LOL.

Now of course that is totally different from the people who would post how much they love the company/stock even though their X has been back in the shop 3 times - they would never do that because they wanted to keep the stock up. LOL.

There are also people who whine about negative posters because they feel that they will drive the stock down, or profit from it, so they play kill the messenger because they think that will actually help prop up the stock. Those people are far more damaging to the brand.

So why then would someone join a forum solely to put down Tedla? I don't see the same thing happening on BMW or MB forums. I understand owners with issues looking for answers or simply complaining about the car they own. Why would someone who doesn't own a Resla join TMC just to bash Tesla's? If it was an agenda with a particular technology then I would expect to see them on other forums. However, if a single ball joint fails on a Tesla then it is something horrible but ball joint failures on other cars are fine. An accident in a Tesla is horrible but one in a Mercedes is ok.

BTW the suspension issue reporting did affect the stock price and the source for the reports was mostly TMC.