Incident ?Have you seen this report: http://www.safetyresearch.net/Library/NHTSA_Autosteer_Safety_Claim.pdf - They attempted to replicate the results of the NHTSA study concerning the number of airbag deployments after the introduction of autosteering in their vehicles, and in so doing, came to the opposite conclusion the NHTSA originally did:
The model estimated from these specific data helps to answer the question concerning NHTSA’s safety claim about Autosteer, “Is the installation of Autosteer associated with a decreased risk of an airbag deployment crash, controlling for exposure mileage?” The answer is “No.” Table 1 demonstrates that Autosteer is actually associated with an increase in the odds ratio of airbag deployment by more than a factor of 2.4
Tesla has stated that they are opposed to the public release of their data because of the risk that the data may be interpreted negatively. Whether this research is an example of that, or whether it is another demonstration as to why the data should be made public, I can't say for sure.
That said, I don't see how anything other than transparency can be a long-term strategy, given the increased reliance on automation, and their desire to be approved for higher levels of driving automation. Their data will need to be examined by many regulatory approval bodies internationally, and one would hope that their conclusions would be based on sound a discovery, and not later brought into question.
Given the high profile nature of the Boeing incident and what we are coming to learn, I don't think anything other than transparency will work here -- even the regulatory bodies need oversight.
For now, all we have is our own personal subjective views, but so far it's gotten us here.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Another tragic fatality with a semi in Florida. This time a Model 3
- Thread starter Az_Rael
- Start date
- Status
Oldschool496
Member
The largest issue with the NHTSA study was already called out by Princeton U's head of vehicle engineering:
Alain Kornhauser, who heads Princeton University’s autonomous vehicle engineering program, has another problem with the NHTSA finding: The data show that, if determination of safety is the goal, NHTSA is asking the wrong questions, he said. He notes that the NHTSA study didn’t assess whether Autosteer was turned on or off when the air bags were triggered.
I don't have an issue with small groups being involved in an analysis, but the studies should be peer-reviewed, well designed, and based on open and reliable data.
Not surprisingly, however, Tesla got as much mileage out of it as they could:
Just after the report, Tesla Chief Executive Elon Musk tweeted, “The data show that the Tesla vehicle crash rate dropped by almost 40% after Autosteer installation.”
The company used the finding to support its contention that Autopilot made cars safer. In March 2018, after a Tesla driver was killed when Autopilot drove the car into a concrete barrier, Tesla posted a response on its website that included a reference to the NHTSA study: “Over a year ago, our first iteration of Autopilot was found by the U.S. government to reduce crash rates by as much as 40%. Internal data confirms that recent updates to Autopilot have improved system reliability.”
The company used the finding to support its contention that Autopilot made cars safer. In March 2018, after a Tesla driver was killed when Autopilot drove the car into a concrete barrier, Tesla posted a response on its website that included a reference to the NHTSA study: “Over a year ago, our first iteration of Autopilot was found by the U.S. government to reduce crash rates by as much as 40%. Internal data confirms that recent updates to Autopilot have improved system reliability.”
In the end, both the NHTSA and the team behind the report minimized the value of the data by stating:
NHTSA: Report was “cursory” and said it was not assessing Autosteer’s effectiveness.
Whitfield: “We did not produce this data. We don’t vouch for it. We don’t know if it’s true or not,” he said.
For me, this only raised more questions and I am looking forward to more careful study.
Oldschool496
Member
Or spilled coffee, stung by a bee, had in infarction, got something in his eye, sneezed repeatedly, a whole host of things may have happened. We all learned something here, pull over when it is safe to do so and correct the situation if possible.
AP is not a crutch or excuse to drive poorly anyway. After all this gentleman gave his life to show us exactly what not to do. My only wish was being able to show him to not do it before this incident.
We should be careful here when we draw comparisons between airbags and autonomous driving systems. Airbags were designed to keep people alive and minimize injury, and this is their only purpose. Currently, autonomous driving systems are designed for convenience with a goal of increasing safety.
There are early indications, and isolated reports emerging that safety under the right conditions has improved, but it has a long road ahead until we can say that it will have saved as many lives seatbelts and airbags have. Airbags and Seatbelts are designed to be used each time an adult passenger is seated, while the current autonomous systems are designed to be used in specific and narrow conditions.
In 2007, an estimated 15,147 lives were saved by seat belts, and 2,788 lives were saved by air bags. If seat belt use increased to 100 percent, then an additional 5,024 lives would have been saved (Starnes, 2008). When these seat belt numbers are added together to equal 20,171 (= 15,147 + 5,024), this estimate is often referred to as “potential lives saved” or “lives savable.”
See: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/811206
As an aside, I am very grateful to have autosteering as a feature. There are times when I feel that I should not be driving (after long trips when I am fatigued, or all-day hikes in the woods) and would like to rely on it more, but I know that it is not advisable given my own level of responsiveness. For the most part, I use it to reduce my fatigue while driving, but never a replacement for attentive driving.See: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/811206
Eno Deb
Active Member
Yeah, but that wasn't really the point. Whenever an accident like this one happens, the apologists pop up and claim "he shouldn't have used Autopilot on that road because the manual says so". But in this and several of the earlier cases, the manual says nothing of the sort. It just says it's OK to use Autopilot on "highways and limited access roads", which the US441 clearly is. Obviously the driver still has to be attentive, but that's a different question and we can only speculate why the driver apparently didn't take action.
Eno Deb
Active Member
Yes, and none of your quotes would exclude roads like the US441. The manual also says "Autosteer is intended for use only on highways and limited-access roads", which again the US441 is. If Autopilot could not be used on highways that have any unprotected entrances and roads that have "changing traffic conditions", it could be used pretty much nowhere. That's a nice CYA for Tesla, I guess.
I hear ya, but I would call the ability for a semi trailer to be across all lanes of traffic (without jack knifing) a changing traffic condition (right angle traffic is also a differentiator between limited access and city)
Oldschool496
Member
If you saw it you would not agree, it is not a limited access highway and by that fact this is how this whole think happened. You can't spin it any other way.
JetFalcon
Member
Sometimes, I can easily eat food while this car drives itself on the highway, usually when its empty and middle of the night.
Worst case, I just drop my food and grab the steering wheel if autopilot suddenly failed. But I only fully trust autopilot when its empty and there is enough room for error for the car to weave over 1-2 lanes without hitting anything.
Worst case, I just drop my food and grab the steering wheel if autopilot suddenly failed. But I only fully trust autopilot when its empty and there is enough room for error for the car to weave over 1-2 lanes without hitting anything.
Eno Deb
Active Member
Sorry, you're not going to convince me that US Highway 441 is not a highway. We could just as well discuss what the meaning of the word "is" is (in case anyone is old enough to get the reference).
Kant.Ing
Member
Tesla needs to define what changing traffic condition is. To me, in a one mile stretch of highway, that the speed drops from 80mph to 20mph then back to 80mph with other cars cutting and moving out of the carpool lane is.
Msjulie
Active Member
My car suggests autopilot on my neighborhood street, a road with cars parked on both sides and a 25mph speed limit. People aren't reading the manual for sure but the car is behaving too optimistic as well. Personally, I much prefer to drive my own car myself but the features are literally beckoning you to use them, I can see how people would mis-use them. There's blame to go around here I think...
Msjulie
Active Member
Agreed and it also can not anticipate another driver's upcoming behavior by noticing them looking at their phone etc.. I don't feel it looks near as far ahead down the road I do either.. when it can really see nearly everything I see and is taught how to behave, then it will be better and faster.
Knightshade
Well-Known Member
I'd say any road where cars might randomly appear going in a different direction than you would be part of changing traffic conditions.
Hence why the system isn't intended to be used anywhere there's cross-traffic (or oncoming, undivided, traffic, either)
AP today works on the fundamental assumption all cars it sees are going the same direction you are.
Is there any mission statement(from automaker and/or organization/body) that mentions convenience is the goal of autonomous driving? This is important as it frames the mindsets of the people involved with developing these systems.
Why is there such a limitation in terms of travel direction?
Maybe the real limitation is the speed delta(plus relative distance) of approaching object and car one is in. E.g. even if traveling at same direction, a car in front 200 yards that travels 30mph, versus your car travels at 90mph, can still overwhelm the system.
If one really really thinks through this, it makes no sense such a system(with primitive capability) can allow drivers to be hands off and not attentive.
Last edited:
Here is what I can share with you - and keep in mind, this is current (as of today) as reflected in the documentation we get with our vehicles. It comes up over and over again in the user manual.
The user manual, available is available online here: https://www.tesla.com/content/dam/tesla/Ownership/Own/Model 3 Owners Manual.pdf
I've listed some examples below:
On page 39:
Valet Mode When Model 3 is in Valet mode, the following restrictions apply: • Speed is limited to 70 mph (113 km/h). • Maximum acceleration and power are limited. • Front trunk and glovebox are locked. • Home and Work locations are not available in the navigation system. • Voice commands are disabled. • Autopilot convenience features are disabled. • The Allow Mobile Access setting is disabled.
On page 65:
Features These Autopilot safety features are available on all Model 3 vehicles: • Lane Assist (see Lane Assist on page 84). • Collision Avoidance Assist (see Collision Avoidance Assist on page 86). • Speed Assist (see Speed Assist on page 89). • Auto High Beam (see High Beam Headlights on page 48). These convenience features, designed to reduce driver workload, are available only if your Tesla vehicle is equipped with the optional Enhanced Autopilot or Full SelfDriving Capability packages: • Traffic-Aware Cruise Control (see TrafficAware Cruise Control on page 67). • Autosteer (see Autosteer on page 73). You can enable/disable Autopilot features and in some cases, control how they work. To access settings for Autopilot features, touch Controls > Autopilot.
Page 67:
Warning: Traffic-Aware Cruise Control is designed for your driving comfort and convenience and is not a collision warning or avoidance system. It is your responsibility to stay alert, drive safely, and be in control of the vehicle at all times. Never depend on Traffic-Aware Cruise Control to adequately slow down Model 3. Always watch the road in front of you and be prepared to take corrective action at all times. Failure to do so can result in serious injury or death.
Knightshade
Well-Known Member
Because the system has limited capabilities.
So AP is designed for "simpler" situations like "all traffic is going the same direction"
FSD, which will require much more powerful hardware and more advanced software/NNs will handle traffic going more than 1 direction- but nobody has that today.
Uh...the current system doesn't allow hands-off or a not attentive driver.
And never will.
Again that's an eventual goal of the more advanced FSD system, with more advanced HW, but it's not here yet (and won't be for a while).
- Status
Similar threads
