Featured CR on Driver Assist Systems

Why are you rambling on about disengagement on an advanced cruise control? Do you have a problem? It's almost like you are running out of ways to provoke people.

Oh, I see. Based on the interactions in a sample of 100 of the 19,382 messages that are attributed to Trollbait, an oddly large number are contradicting each other.

Let me make it better for you, Troll. You are right. Even though Openpilot is not doing the same things in the same way, I should be checking to see if it's acting suspiciously about it's usability. Since trolls know what I others are thinking, how many messages will it take before you start criticizing my posts on that subject?

 

if you - as one person are provoked by discussion/facts - then who's the person having a problem .... other than countering your view(s)
Someone who assumes others are trying to provoke them - often presume that motive - because it's what THEY do

.

 

Autopilot got a 9 out of 10 in performance. But a 10/10 Level 2 system that lets a the driver take their hands off the wheel for nap has no business on public roads. The other categories are measures of how well the system keeps the driver out of trouble, and others safe.

Does the latest update make the system easier to use for newer drivers? Better at informing when it is safe to use or can no longer function? Does the LCA still turn off everytime the driver steers the wheel? Is Tesla turning the driver monitoring camera on with Autopilot now, or is still just with FSD?

 

We do have multiple crash instances where people without L1 or L2 ADAS systems took a nap. Should all cars be equipped with ADAS systems that prevent distracted or impaired drivers? Tesla made a mistake in 2015 by trusting drivers. They have improved and a driver needs to take active measures to fool the system. People are idiots and will continue dangerous behavior, but we should not make car makers make sure drivers aren't using novelty glasses to fool a system into thinking a driver that is looking at their phone or sleeping. There needs to be personal responsibility.

I don't really trust Consumer reports when it comes to analyzing new tech, they have an old tech bias which we saw against the prius. AAA did tests and found visual systems were much faster to detect inattentive drivers, which means warnings can be given earlier. If we can figure out the privacy issues, visual systems plus a touch sensate steering wheel may be the way regulators should go.

Tesla's self driving v 11.3 is still in pre beta form. Some of these drivers (small number of tesla employees) have shared the data. This appears to be a better version of the level 2 system, but much better mapping and unified "stack" so the same software is running on highways as non highways. It is already easy to use. No date of when it will be out to non employees. The biggest feature upgrade appears to be the ability of drivers to record voice messages that gets sent to tesla when they think the system is not operating properly and what they think it should do.

It already is very easy to use, so I doubt they would be able to make it easier. Visualization have improved so a driver can tell more easily what the system is thinking and take over if they don't like it, and braking for potential hazards starts earlier which may make it smoother too.

Lane keeping will still turn off when the driver takes over steering as it should. The driver can then easily reengage by clicking the stalk. I have lane warnings enabled on my system, which vibrates my wheel as a warning if it thinks I am crossing lanes unintentially (which sometimes just means slowly or to get around a car that is over in my lane).

You need to enable data to be sent to tesla to use the self driving beta about the driving facing camera. Autopilot does enable the camera but only locally (data not shared with tesla) unless the driver sets it to. Either way tesla can use the data to monitor drivers. I am not sure how well it works today with autopilot but it likely is running all the time, but likely is not active based on reports. Tesla has reported all accidents involving autopilot and full self driving beta to the government since june 2021, including those going back to years before. NHTSA should be issuing a report this summer. Actual wide release of FSD beta 11.3 is likely to have to wait until new features the NHTSA has requested are put in and tested.

 

To be clear, my post at #64 was in relation to CR's test results. That test was the use of DCC and LCA at highway speeds. They did not test FSD, and maybe not even Enhanced Autopilot. The article only states Autopilot.

The article is lacking on the actual testing details. A complaint I've had about other test results they've published. Maybe the earlier results I linked to has the info.

The Outback's ADAS and driver monitoring settings are separate. I'm pretty sure it will watch me for attentiveness with all the Eyesight features disabled.

It's the person's responsibility to fasten their seatbelt. Should we get rid of chime? When people prefasten a seatbelt, and sit on top of it, are we holding the car maker responsible if they get hurt in a crash? People are idiots, but a lot of that idiocy is passive. They'll forget to put on a seatbelt, or their focus drifts from the task of driving for whatever reason vs. actively defeating safety features. That's what the belt reminder and driver monitoring is for.

Since distracted driving has caused crashes before ADAS, driver monitoring might be a good idea for all cars. CR stressing it with ADAS is cause research is showing people can become too reliant on the ADAS, which leads to less focus on the driving task.

As for active idiots, I'm all for letting the rule of 'play stupid games, win stupid prizes' play out.

Relooking, Tesla and most were rated the same in ease of use. Where Tesla got dinged seems to be in not having the direct driver monitoring and too long for alerts to happen. I've seen the driver camera only being on for FSD reported elsewhere. Since this was suppose to be of newer cars to see any changes, I'll lean towards it being turned off without the testers knowing vs. any alerts from the system taking longer than the test condition. On but not doing anything in Autopilot isn't effectively different than being off. I have the facial recognization settings on in the Outback, so I know if it is working when the I sit in it.

Seems the movement of the technology is to the cooperative driving for the eventual step to Level 3. Otherwise, the continue operation of LCA after manually turning the wheel can be debated for either side. Cruise control disengages after touching the brake pedal, shouldn't it do the same after touching the accelerator? Planes have crashed after a pilot turned the yoke because they didn't release that had turned off that part of the autopilot. Maybe it behaved differently in other aircraft.

I have my doubts about openpilot use on the road, but it seems to do one thing right. You have to go through a tutorial or training session before it lets you use it.

 

When I have looked at CR tests I always seem to find problems in methodology. We can look to better testing organizations, so I have no reason to dig into this one. In there previous negatives about autopilot it did not like what it did when other systems automatically would be off. How is that a huge negative that it could deal with more difficult situations but they are difficult so it wasn't as well as they and others dealt with easy situations. They really need to make clear what version they are testing and against what. The drivers judgement on driving in bad weather, while impaired, or not properly using a system is an important component in having accidents.

Without an ADAS active all a monitor can do is make noise or light or vibration. It needs to be active to safely slow down and pull over.

Click full report to get all the details.
https://info.oregon.aaa.com/aaa-driver-monitoring-systems-need-improvements/

The tesla they tested in 2022 did not have the camera monitoring system active only the steering wheel torque sensor. The GM and Subaru camera system provided faster response and less time not focused. There are privacy issues about turning the cameras on, but with cameras the car could detect drunk driving more easily, and NHTSA may require such a system in the future.

AAA's data appears to be sparse so conclusions may not really be true in the real world, still it is a good valid test and data point.

 

See the CR MPG discussions for my opinion on their methodology.

Autopilot got a 9/10 in both CR tests for how well the system worked. They don't seem to have a problem in that regard. Tesla dropping in the overall rankings has to do with the other categories, and the fact that others have improved more.

"Systems from Audi, BMW, Tesla, and Volvo were capable of being operated during test drives through residential areas when the road had only a single, center lane line. These systems, that lack effective direct driver monitoring, should restrict use in less-than-ideal situations."
https://data.consumerreports.org/wp-content/uploads/2020/11/consumer-reports-active-driving-assistance-systems-november-16-2020.pdf

It appears CR issues isn't in the system operating in riskier situations, but that there should be better driver monitoring if it does. That's the industry report from 2020. The public article may have made different statements. In addition to the above, there were two other critiques of Tesla. One was the long time it took for the hands off alarm to go off; repeated in the 2022 OP article. The other was in not having a display in front of the driver.

The PDF does go into the test methodology for those interested. It also has test results broken down within the categories(ex. separate results for DCC and LKA). They also have more data behind a paywall. CR likely knows the software versions being tested. They should report it. If this is still something few car buyers ask about, I can understand why CR left it out. Did anyone here asking for it send CR an email?

I don't expect the seatbelt minder to keep me from shifting out of park. People will do stupid things. Those things fall into two groups; intentional and unintentional. Both can lead to bad outcomes. Alarms are for the unintentional group to let them realize they are being stupid, and correct that behavior before the bad outcome.

Their conclusions seem to be in alignment with CR's.

 

I just don't like CR's methodology that seems to try to slow the progress of technology. I would not trust it. I did post an aligning test that had proper methodology from AAA. Small sample size and limited scope means any results should be taken with some

That is why I referred to it. It is a very small set of drivers and it should be a blind test with a large number of drivers in their own cars that they have spent some time getting used to the L2 ADAS system.

A previous study predicted based on limited use that a L2 ADAS system could prevent 47% of accidents. The key tech though is adaptive cruise control, Automatic emergency braking, and blind spot monitoring. Lane departure warning and lane keeping were thought to prevent a minimal percentage of accidents. Tesla recently greatly improved blind spot monitoring, it not only shows other vehicles and people on the screen, but when the turn signal is active it shows a picture from the side of the car. I check that before my rearview. It is the first 3 systems that may account for teslas low accident rate on autopilot. A Carnegie professor theorized that when systems get good people trust them too much, and they cause more accidents. It would be good to see a study with a large number of drivers (insurance companies have monitoring devices) and see if there are less accidents with adaptive cruise control than with this in conjunction with auto steering.

An optical monitoring system looks like it will improve driver monitoring, but according to AAA not enough and drivers could get around the optical system. Drivers need to stay alert in situations. L2 is not safe by itself in bad weather, where pedestrians cross not at crossings especially children darting out between parked cars or in a construction zone. None of the systems is fool proof, and there are a lot of fools out there. More people are getting toyotas with L2 ADAS systems. I think that will make them safer.

 

The rating system does obfuscate data, and they need to drop qualitative tests for quantitative when they can. Why not simply measure DCC following distance instead of asking if it was adequate?

CR did give the drivers 'hundreds' of miles of driving to get familiar with the systems. Not sure how some of these tests can be done blind. How do you time hands off the wheel without an observer and a stopwatch?

The AAA acknowledges that cheating the camera is possible, and that there is room for improvement, but it is the systems they want to see in cars.

"AAA recommends that automakers opt for camera-based driver monitoring systems over steering wheel monitoring; however, more refinement is required to prevent driver distraction and misuse.
...
Vehicles equipped with camera-based driver monitoring systems were significantly better at preventing each type of tested distraction scenario by issuing alerts faster and more persistently than a steering wheel system, no matter the external lighting conditions. On average, the percent of time test drivers were forced to focus on driving was five times greater when facing a camera than with steering wheel input.

Both driver monitoring types were prone to being intentionally fooled, although those using a camera were harder to trick."
- AAA press release

"Direct systems were also adversely affected by active circumvention attempts by all test drivers. However, test drivers generally found it more challenging to evade detection on a prolonged and consistent basis. With additional effort to evade detection relative to indirect systems, direct systems may discourage continual and intentional misuse of ADA systems to a greater extent."
- AAA report

 

As far as understand, Tesla lacks a blind-spot monitor. Things like that would be a deal breaker for me. Model 3 is a nice car, yes, but misses safety essentials I need.

 

But did it alert you before you've steered towards other cars not visible in the outside mirror?

 

Pretty sure Tesla's had BSM standard longer than it was on the Prius.

 

What other car can you drive off a cliff and have everyone survive?

Man who drove Tesla off cliff with family inside charged with attempted murder - CBS News

Mike

 

What other car can you drive off a cliff and have everyone survive?

That's not as rare as it appears. 64 years ago my mom hit a patch of black ice while driving a Hudson sedan through Wyoming. We went into a skid that ended when an oncoming car lost control and pushed us off the road. We finally stopped at the bottom of a 75 foot cliff. Mom had a broken back. Dad had a concussion. Both had been thrown from the car. All three boys were sleeping in the back seat and were uninjured. All of us recovered within the year.

That was in the era of big heavy cars and metal dashboards. No seatbelts. All of us survived.

 

As @dbstoo mentioned, that was pure luck. Accidents are governed by laws of physics, and these laws do not change with the make of the car. If that Tesla hit the wrong thing when it was rolling down the cliff or free-falled and hit a rock, it would be a very different accident.

Tesla has never had a BSM as far as I know.

However, watching the video, there seems to be an ad hoc BSM if Autopilot is turned off and you change the settings. It is an annoying beep that most drivers wouldn't stand. A real BSM has warning lights on mirrors.

The whole point of a BSM is it warns you before you change lanes. That is not a BSM if it warns you after you steer into a car in your blind spot. That defeats the whole purpose of a BSM. I am not interested in postmortem alerts and debugging.

That is why I do not like Tesla. Trying to reinvent the wheel always fails because you cannot fight common sense and common knowledge.

 

Yes I can tell you it has a much better blind spot monitor than any other vehicle I've driven. They recently added functionality that when you hit the turn signal to change lanes or turn, the camera in that direction is shown on the screen. It also has a graphic on the left side showing cars, bikes, trucks motorcycles, construction signs, etc around you.

It doesn't do anything on the mirror, but it is there in the center so that you know if there is something in the way even before you check the mirror.

You can set a blind spot warning chime this only goes off if turn signal is on, or steering toward a lane change or a turn.
Lane Assist.