Hi, all.

Been a P3 package owner for about 5 months now and really liking the car. Having gone from a '14 CLS550 to this, I thought I would miss the power but I've been pretty pleased with the pep this 4-banger puts out. Yes - the power is quite different, but the gas mileage upside is nice.

Ok - I've had this incident happen twice now. First time I wrote off but since it's happened again I am going to take into dealer. Has anyone happened where the auto brake kicks in as if there is imminent crash detected but there is plenty of space in front? It kicked in for 1/2 second, violently, then let up. Luckily no one was tailing close but this is concerning.

Any other experiences like this? Thanks!

 

​​​​​​
You can adjust the distance and the severity of the braking. Curious what you meant by "plenty of space"
​​​

 

Thanks, HB. Should have read 'plenty *of* space'. I am aware of the distance adjustment but there was no one in front of me close enough to trigger the braking. I have the dynamic display on the left instrument panel that also shows the driving aid / distance graphic and no car was registering within 100ft and the 'safe breaking' line was not yellow.

Was really strange and braking so violent that it felt like it MUST have detected something or software glitch. Either way, going to have MB look at it. Just disconcerting knowing car may do this at any time - hopefully not at any speed!

 

The same thing has happened twice already in my E300. Car violently stopped for no reason, no car in front of me on a two lane highway. I've had Distronic Plus in my '12 E350 and have never had this happen. I'm curious to know what the dealership says.

 

Most likely a software issue that needs to be updated. Or a defect with the camera/sensor.

 

I thought the software option for early/late auto brake was completely separate to the drive pilot adaptive cruise/vehicle distance setting. I've set mine to early and I've once had the car panic brake for no reason. It had nothing to do with adaptive cruise, when that brakes it aparantly can only do so with Max 50 percent braking force. When it happened, it was full brakes along with seatbelt pretension pulling me back. I had that happen a couple times in my old w212 on the motorway when braking hard, those times were entirely appropriate as both were near accidents.

On the w213, it wasn't appropriate at all. When it happened I was on the motorway approaching a gantry that stretched over the carriageway. No traffic in front, one car broadly alongside. The gantry threw a really sharp shadow on the road (strong sunny day) and I wondered if the system takes radar info and camera info and put the gantry gantry radar reflection with the sharp black "object"in the road and triggered braking because of it. It if only relies on radar then I have zero clue why it happened. Only time in 4,000 miles so far but it was unnerving and I was glad I had no tailgater behind me!

​​​​my wife's old boss drives a cla and that has braked hard on its own a few times. The system is clearly a little twitchy. I'm sure I've read of false triggers in the w213 in a review as well. I guess no system is infallible.

 

False positives are a reality. The solution is meta-data tagging, very much a part of autonomous vehicle technology.

The concept is meta-data information attached to the navigation real time position logging. When cars have nav and their own wifi connection, it is possible to "tag" that 30-40 meter stretch of road for sudden shadows. The rule would be something like distance control not tracking vehicles (clear road) and wide view radar not triggered (no animals running across the road). A harsh shadow false positive at this position on the road would then be bypassed. It could also be set for specific time periods during the day.

This meta-data tagging is essential for certain building and bill-board placements and a host of other "gotcha's".

In Germany, a smaller sample space, a lot of this meta-data tagging has been accomplished. In the US, routes traveled by Tesla vehicles have been heavily meta-tagged as well. It is simply a matter of time for heavily traveled US routes to attain an improved level of meta-data detail.

 

Do you mean it would be OK then to hit an object within this tagged stretch of the road, if a person, animal or a broken vehicle appeared there next time you pass that area?

I'd rather have the SW bug fixed.

 

Understanding the equipment, concepts and steps involved in object identification is not a trivial subject, neither is it overly complex. It works very intuitively once you have been introduced to the parts required. Meta-data tagging is a given for all proposed autonomous driving systems. Lack of it in this example is precisely the SW bug, as you have called it.

The logic testing works very quickly, in nano-seconds. The the Merc system, which has three radars, the long range radar would pick up anything actually "standing in the road". What is happening is the short range radar is triggering on the shadow, while the long range radar had nothing to contribute. The system it is being 'panicked' for lack of the complete picture, as it were. Meta-data is pretty much a way of saying, if you read this condition in this exact way, ignore, but it is quite precise, and not quite as simplistic as total shut down for that small stretch of road. Instead, the logic parameters for that small stretch of road must be more accurately defined. I merely suggested how this would be accomplished, you guessed it, through additional SW programming refinement.

The meta-data concept in this example is what must be implemented into for autonomous driving levels three and four to happens. Bridges, overpasses, bill boards and buildings are all false triggers in this way.

We are at level two and without this type of extra logic and testing provided by meta-data use, there will be no level three.

Got it?

 

I'd much rather endure a false positive than suffer a false negative such as with the fatal Tesla crash in Florida.

 

True, I was going to say something about false negatives vs. false positives, but thought to point out what needs to happen avoid both in the future. That is the constructive, if less sensational approach.

It has been pretty conclusively established that this guy was watching a movie in a mode two autonomous level car. Candidate for the Darwin Award. Done and done.

Here is the deal. There are lots of accidents today. Not too many are fatal, but there are plenty of serious injuries. If mode two and especially mode three is implemented in the name of active safety in every car, many accidents will be avoided, but most important is multiple thousands of accidents, thousands of injuries will be reduced in severity, and hundreds of deaths, perhaps eventually thousands of deaths will be prevented.

We are hoping to develop an interactive protection system and it will occasionally be, not so much as proven to have not caused, but to have contributed to, a dozen or or several dozen accidents. But, this is getting it wrong. It is the net reduction of injury that is the key. The situation very much reminds me of my aunt Elsie, who would not wear a safety belt. She was sure that she would drive off the road and into the river, and drown in the safety belt. Yes, this could have happened. But, it ignores the 99th percentile probability of being involved in an accident that does not involve a river! Here again, in a different, much more complex technology is a risk of getting it wrong.

This Tesla guy was and had been repeatedly doing something he could do, but had been instructed he should not do. How is this different than the person who loses control of their speeding car and for the very same reason? It is an outlier sample that contributes little to understanding the promise of active safety. This is tough love, lawyers will also love it and victims will not, but implementing serious stages of active safety is about acceptable risk, not about absolute, zero defect safety.

Active safety, up to and including autonomous vehicles will prove to be by far a net positive for each and every driver who sits behind the wheel. Consumers, regulators and manufactures only need to be reasonable in their stepwise approach to lowering the frequency and severity of traffic accidents. Let's hope we achieve this goal.

 

A radar based system (or a laser system) cannot be triggered by shadows since a shadow has no mass to reflect emmited energy. This is why I wonder if on later mercs the core system is radar based and if the drive assist option is fitted, hence cameras are fitted, the optical feed then augments the detection system. I'd be surprised if that was the case due to system engineering/verification effort involved but I've not seen anything to confirm it or rule out out. If merc is only radar then my false positive wasn't because of the shadow despite that being the only obvious cause.

I agree entirely though, better to have occasional false positives than ever missing a genuine need for braking assistance and roll on the day all cars have this and true car to x communication to cascade that warning back along the road.

 

Sure if the Tesla driver was watching the road he might still be alive. But from a technology point of view it matters not because the system did not react to a large truck in his path, a classic false negative.

The bottom line is that we are not discussing human safety but whether the technology reliably works as designed. Each failure has a different cost both economically and in human terms. Reading various blogs and forum posts one can surmise that many of the new technology features in vehicles today are unreliable including remote start, parking pilot, lane keeping, . . .

 

My comment is this is early times for much of this active safety technology.

1) Parking pilot comes with detailed instructions in the manual. All of the complaints I have read are from people who did not read the manual. Its intention is to find and navigate in potentially difficult parking situations, i.e. between two other cars. If there is enough space for a idiot to park, well...there you have it.

2) As a person with two more take-offs than landings, it is difficult for me to rotate my body and neck to quickly double check my mirrors anymore. Blind spot assistance is just the thing. It is working better than I do.

3) Lane keeping tech will bring more frequent paint jobs to public roads in the interest of safety, and certification that stretches of road meet those standards will be recorded and monitored by in-car GPS. Lane keeping's most significant hope is avoiding bridge abutment or other barrier collisions in the future. It does not presently assure this at level two assisted driving technology, but it is very close to being able to do so.

4) Remote start is being improved by a built-on-crankshaft, 48v low power KERS in the next generation of MB engines, specifically so that the car will not restart, but simply pull away from a stop with the electric motor allowing the engine to seamlessly restart once rolling. This KERS will also effect a 15% fuel savings in urban settings.

I think we are discussing human safety, and not just technology, per se. Active safety is a complex issue, but here is the gist of its origin:

It is sometimes difficult for conscientious drivers to imagine to what degree their competencies are not universal. But this truth wills out: For every hyper alert, highly skilled and well experienced driver, there are two who are on a given day not alert, unsure, indecisive, or inattentive and at best, simply waiting to meet a kindred spirit.

We ought to keep this reason for active safety developments foremost in our mind's eye. It is the variation in driver attentiveness and downright quality of judgement that is the core issue. Some of us, at least some of the time, need all the help they can get.

What is true is that in developing these systems we indeed are creating a new moral dilemma. If we develop a way to significantly decrease the number of bug splats on the windshield with technology, will we accept that the cause for some small percentage of the far fewer bug splats may have changed, that a collective notion of error must be accepted in place of individual notion of error? It is a good question. Some cultures would not have much problem with this question, but do trust American lawyers to challenge anything that might reduce their business.

 

The core system of MB W213 optional driver assist, the full Monte of you will, is radar based and camera assisted. The camera is for a digital form of visual object identification. There are three forward facing radars for object detection at much greater distances and angles.

A false positive such as you described would have been a pure camera i.d. error, and meta-data is key to reduction of such occurrences. If you go on-line and search videos, there are a couple of keynote addresses by Amnon Shashua, who is particularly articulate on the meta-data issue and the progression thorough levels two, three, four and five. I think your questions will be well answered by spending a few minutes listening to him.

 

My comments are above in red.

 

The gantry is your answer. Overhead objects can in some instances trigger these systems, though usually just for an instant. I have had it happen on both my acuras and my C class, however, at most Ive had the seatbelt retractor and dash warnings activate, Ive never had any of my venicles come to a full stop. All my false activations have been in chicago where there a lots of low train overpasses. Another thing that can cause activation are the large steel sheets that are used on the roadways to cover excavations. No system is perfect!

 

So, I started this thread and wanted to post a couple of updates:

1) Immediately on posting this - a Mercedes engineer reached out to me looking for more information. He helps lead the NA development of the autonomous systems and wanted more info. I thought this was pretty impressive that Mercedes is monitoring this forum. He also gave me some talking points to recommend to the dealer.

2) The dealer visit took 2+ days. I passed along comments from the engineer about software updates and camera calibrations.

3) When I picked up the car, the following had been completed:

- Check fault codes of Distronic system (they indicated the event was recorded)

- Replace left radar sensor due to fault codes indicating improper installation

- Performed reset of adapatations and SNC coding per guided test (not sure what this is - dealer indicated software updates were completed so maybe this was that item)

- No further faults after road test

All in all, seems some comprehensive activity so we'll see what's happening and go from here.

Thanks, all.

 

Must have been SCN coding (software calibration number)? Almost all control units are supposed to be SCN-coded after a SW update. It is an automated process to version code the control unit correctly to match the individual car to which it has been installed.

Hopefully the part swap/SW update fixed it.

 

Hope everything goes well from now on

 

Wait - they are watching us?!

 

Big brother is watching YOU!

Hey, big brother! Is it possible to put an 80 liter E43 fuel tank in the E300? The 66 liter E300 tank doesn't last a full day of driving.

 

Happened to me last night, car kept on braking and beeping intermittently as if i was tailgating someone. I suspected a dirty sensor so i pulled over. The culprit was the temporary cardboard license plate which was swaying to and fro. Must have gone loose, as soon as i smoothed it the problem disappeared

 

Anything new on this errant braking issue? Pretty worrisome.

 

Thanks for the update!!!