Auto Safety Features: Good, Bad, and Indifferent

Transport
21

The figure only shows road fatalities and does not indicate the impacts of injury/trauma on road users. The later can affect one for the rest of their lives.

Research has shown that in Australia, distraction is the main contributing factor in approximately 16% of serious casualty road crashes and also suggests that distracted driving is as dangerous, if not more dangerous than drink driving. It is a serious problem on Australian roads (source).

LKA are adopted by manufacturers to ameliorate the actions of a distracted driver.

Edit: It is worth noting that:

In the United States, Police-reported data estimates that between 18 and 22 % of all crashes involve some form of driver distraction as a contributing factor (2010).

However, information from the largest analysis of driver behaviour conducted, the 100-Car Naturalistic Study in the USA, found that 78% of the crashes and 65% of near crashes had one form of distraction or inattention as a contributing factor. Of these 30% were derived from driver engagement with sources of distraction outside the vehicle (crash risk increased by 3.7 times when looking at external objects).(source)

While similar investigations are yet to be carried out in similar detail in Australia, they are likely to be somewhat representative of that in Australia.

1 Like
22

Agree, but is it being appropriately valued?
Assuming the road environment/conditions at the time enable LKA to assess and provide an effective response. For those stuck in the daily commute on one of the “M” roads around our larger cities.

What percentage of the 16% of accidents and for which road environments would it have proven most effective. There is certainly a need to break down accident statistics further. 2020-21 are not typical years.

In 2019 the trend was towards safer roads in urban environments. Nationally 431 fatalities in urban areas and 753 fatalities outside of urban areas. No surprise as to where most of the road money is being spent.

1 Like
23

I can imagine fighting with a safety system as seems may often be required on our roads would keep the driver focused on the road?

What is good for the American system of super highways and main roads might be applicable to our East Coast and a few other odd routes, but I digress since there is nothing of importance beyond the general rectangle bordered by the Pacific ocean, Paramatta, Gosford, or Sutherland NSW, right?

Down my way (or up from Tassie) what makes sense on the M’s makes none at all once off it on the C’s and lesser roads.

1 Like
24

GoogyEgg,

I hate to be the bearer of bad news, and I hope I don’t simply confuse you with this. It’s an area that I have some knowledge of, although my knowledge is not necessarily current
 I absolutely understand software engineering though, and safety critical system engineering process needs. Bear with me, if you can. I drive a vehicle with Forward collision avoidance (FCA); adaptive cruise control; blind spot warning (BSW).

It doesn’t have LKA or LKS. The BSW can get confused. e.g. two left or right turn lanes, where you’re in the longest radius turn lane, with turning vehicles in the lane adjacent the side you’re turning towards. The adaptive cruise control (and presumably the FCA) does not always detect a car entering your lane as soon as it should - that’s because there’s a single radar / camera, and there’s a geometric blind spot. The technology is great if you understand its limits, and don’t rely on it as if it’s perfect. We’re 50 years away from truly reliable self-driving vehicle technology. The question is precisely what does your vehicle cope with, and what can it not cope with - especially the issues you’ve described with LKA.

Australian Design Rules changed in maybe 2018, and announced that FCA is mandatory for all new vehicles

ANCAP keeps track of vehicle sales in Australia to some extent, and provides testing of the efficacy of various systems, and obtains results of equivalent test reults for equivalent vehicles in other jurisdictions. They provide interesting details. You can assume the claims for crashes are based on actual crash statistics. ANCAP
The crash statistics are the real reason there’s been so much development of these technologies, and the proliferation of them. Despite any shortcomings, they save lives, or mitigate the outcome of actual crash situations.

In the case of the system on the Hyundai i30 Elite 2022, and the problems you describe, I don’t know. One thing I am confident of, is that whatever it does, in each vehicle, the software will be absolutely consistent - even if it appears to behave randomly. The randomness is the conditions required to trigger the specific action. If it operates via camera (highly likely), the camera must be clean, or it won’t pick things up. If you drive through rain, and get oily road mist on the camera, it won’t work properly. The same as if it gets covered in dust, etc. It won’t have a backup system, or the capacity to verify the amount of light detected etc


I would be surprised if you can navigate the manufacturer’s systems to get a sound answer to these questions. The dealer network won’t know. The questions you’re asking relate to software functionality. That might have one (or two) support people for Australia and New Zealand. There job would be to liaise between the dealer support network and the factory software team.
This software is extremely complex. It is proprietary. There are standards, but they don’t impose functional equality between manufacturers, as far as I know.

I regard self-driving car technologies (all driver assist technologies fall under this banner, of level 1-5 driver assists, with advisory assists at the lowest level to fully automated, self-driving at the highest) as the bleeding edge of design. I reviewed some early prototype technologies for things not yet used, and although enthusiastic about the ultimate potential, I was uncomfortable with way too many aspects to detail here.

I can’t speak specifically for Hyundai, nor any other manufacturer, but there is no way they can test these in every possible driving environment. For electronic financial transactions, most back-end systems operate on fault-tolerant storage, with real-time, off-site mirroring. It is highly unlikely that a transaction will be lost. The computers are housed in temperature controlled data centres, with redundancy built in.

Cars are designed to operate on a wide range of road conditions, in a wide range of temperatures; electrical interference; vibration and so on. Yet the systems have no backup. You buy the car, maybe you read the manual (you can’t possibly remember every detail); maybe you don’t.

When something unexpected occurs, you have no idea if the system is working as it should (regardless of whether it works as designed - it’s not necessarily the same thing), and there really isn’t anyone knowledgeable that you can ask.

None of the technologies are bulletproof.

In your specific case, any question you ask needs to have the conditions that trigger the behaviour you wish to understand very carefully described. It will probably require translation into another language - maybe Korean, maybe not. It depends whether it’s developed in-house or outsourced. The LKA software may be licenced from a third party, and Hyundai might not have any direct access to the code


It’s very difficult to find answers to the questions you raise, for a raft of reasons. I hope this helps.

2 Likes
25

Does this mean that if the camera was to get too dirty, that the LKA would stop working?

1 Like
26

@CaptJack

even though

duct tape can solve the operational annoyance problem? :laughing:

3 Likes
27

I have a 2021 Hyundai Kona Elite fitted with LKA.
I found it just as you described. More likely to cause an accident rather than prevent one.
The LKA also gets confused when you are driving alongside cars in two lanes that converge to one lane.
The steering wheel is quite stiff when LKA kicks in, and as you say, requires a small amount of extra force to correct it back on track. A force that you would not normally need to apply to correct slight deviations.
While it unfortunately won’t help you, I am able to completely turn off LKA, including warnings, as I don’t like it, and it has been that way for the past year since the car was new.
I have been driving cars for many years now, and have always managed to keep between the white lines, even on twisting winding roads. If and when my driving skills deteriorate to the extent I need Lane Guidance Assistance, I will give up driving the car altogether

6 Likes
28

Worse. It won’t work properly.

In the distant future, there will be multiple transducer/detectors on multiple independent circuits
i.e. redundancy.

Until then, you can’t rely on the electronic automotive goodies without a few daily checks.

1 Like
29

I don’t believe I’m going off-topic with this post, but please let me know if I am.

An annoying “safety” feature which is now probably compulsory on all new models is the annoying pop-up window which tells you in a nanny-state fashion to drive safely and to observe the road rules. The exact wording seems to depend on the vehicle model, but it’s certainly present on my VW T-Cross and on the Hyundai i30.

You see it each time you start driving and are required to hit the touch screen in order to clear it. Surely that act in itself is unsafe, especially if you don’t happen to notice its presence until you’re already driving in heavy traffic. As with the LKA, it doesn’t seem possible to deactivate it. Ironically, the distraction is more likely to cause accidents than to prevent them.

Surely that’s taking things too far.

3 Likes
30

Our Renaults (2013, 2017) and a Nissan (2015) have the inane popup windows but on all 3 they will disappear in a few seconds of driving if ignored. Have you tried ignoring them for 10 seconds or so, or have always just acknowledged them?

That suggests it does not clear itself and if that is the case I’ll agree that

2 Likes
31

No, sadly it persists until one of two things happens. Either you click OK or hit a power pole while attempting to do so


2 Likes
32

It is.

My favourite is the warning about low washer bottle water. I understand that being able to wash the windscreen is a safety feature but when the water gets low to constantly flash up a reminder that cannot be turned off is worse than the petty problem that it is intended to solve.

4 Likes
33

There is also no accounting for the human factor. One who shall remain nameless and is of the old world has a modern 2021 vehicle that replaced one from last century. When a particularly annoying symbol lit up the dash, the immediate thought was there was a fault with the instrument panel. Having driven home from the shops with the indicator still flashing the car was garaged and a call made to ask. Do you think something is wrong and should I book it into the dealer for a look? Fortunately wise advise prevailed and the breakdown service that came with the warranty turned up the next day and changed the flat tyre for the spare.

It’s an interesting assumption about human behaviour. While the vehicle will happily offer numerous voice responses to many things, a simple low tyre pressure warning requires only a vague symbol on the dash. Many of us may not have owned or driven or received proper training on a vehicle with the latest of so called safety enhancements? Should it require more than a short chat from the dealer and please read the manual advice before one is legally permitted to take control?

2 Likes
34

Thank you everyone for your responses. I will answer the questions about how I have found the LKA and other AI safety features in my car when driving in different conditions in another reply.

As for your question Syncretic

How can I be legally responsible for the actions of my car when I am behind the wheel if I have to counteract the efforts of the system to drive badly?

I strongly suspect that the car manufacturers will happily, (hopefully not literally), throw drivers under the bus if the AI/safety features cause a crash. Hyundai, and possibly other manufacturers, have introduced a screen notification as soon as the ignition is switched on, requesting that drivers click on one of two choices before the journey. The choices are

  1. I agree to drive safely OR
  2. I do not agree to drive safely.

A ludicrous attempt at getting us to waive our rights? Machiavelli would find it too far beneath him to even laugh.

Sometimes I just mutter to the car grumpily, NO, I WANT TO DRIVE UNSAFELY TODAY. SO THERE!

Curiously, I have on two occasions accidentally forgotten to click on either choice. The nagging reminder on the screen simply disappears and lets me drive. Maybe my Artificially Intelligent car just knows that my real intelligence has a strong leaning toward Oppositional Defiance Disorder and gives up.

If a driver were to crash and had not selected the safe driving option before an accident, and that accident was caused by a safety feature and not driver error, I’m sure the manufacture would attempt to use it to minimise liability.

I would hope that argument would be as compelling as “It’s the vibe” in Australia. I’ve never known anybody here to leave the Winnebago to drive itself while they go and make a coffee out the back, but
 who needs the hassle of trying to deal with this kind of legal gotcha when injured or trying to get the car fixed.

3 Likes
35

In response to the Hyundai issue I looked at a Hyundai i30 owners manual and the amount of safety noise and admonishments was impressive. It anecdotally seemed to almost overweight against information on how the vehicle works. Impressive enough to take ‘safety information’ from useful guidance into noise, at least for myself.

Their ‘technical writing’ department may be completely populated by lawyers whose only KPI is zero successful customer lawsuits, as defended by their writing?

Yet with the related posts on the Community it seems we are on a one-way roller coaster adopting international harmonisation led by the US and EU interests, with our own being sacrificed because we are a comparatively minor market and mandating anything special would raise our prices and/or reduce our choice (no pun intended) so we get to suck it up.

3 Likes
36

The same is true of most electrical appliances, and many other things - e.g. kids toys. They have a whole range of largely unhelpful butt-covering disclaimers (as safety warnings) preceding the operating instructions.

PhilT, you’ve described the problem well. The question is whether (a) such instructions can be construed as part of your purchase agreement (supplied after the fact); (b) whether or not they can be construed as ambiguous in the context they are presented; (if so, who can say what they really mean)
and so on.

I’m not a lawyer, so hopefully someone well versed in consumer law and case law can offer an opinion as to how much weight these disclaimer gotchas actually carry. I’ve been told it often comes down to how previous cases have been decided by the judge(s); i.e. case law.

I wonder what happens if the driver answers “I do not agree
” but then proceeds to drive safely and still has an accident caused by vehicle malfunction? Was the Q&A relevant? Is it material? Who or what is the “agreement” with? What is the term of the agreement? How are you supposed to know?

It’s nonsensical. The questions are insulting. The sold the car before they asked the question, now they seek to obtain a “get out of liability free” card from their customer.

1 Like
37

Yes, an interesting observation.
Is it also fact that many of the added features intended to enhance safety are now required by design rules and compliance for sale/registration. There are various design and or testing requirements that are beyond any one manufacturer to control.

A thought is the manufacturers and authorities responsible for testing and approval would present a common position in any formal process. One of support for what had been achieved to date and that all parties recognised the need for continual improvement.

When mandatory wearing of seat belts was introduced a supporting argument has been ‘seat belts save lives’. It was never promised seat belts will save every life or prevent every accident. The suggestion is all subsequent safety enhancements to motor vehicles are considered on a similar basis. On measured overall improvements, rather than individual outcomes.

Operating instructions for nearly every purchase of a device, tool, powered item are supplied these days - ‘after the fact’.

I could add that for construction machinery, high risk plant etc both the owner operators and suppliers manufacturers do have their responsibilities for ensuring safe use and operation more clearly defined.

1 Like
38

I wonder if many of these warnings are because enough people have failed in the past to do their regular checks of the vehicle to ensure that the components they can check, are within safe operating conditions. I think (my opinion is) things like window washer contents should be checked regularly enough that a warning shouldn’t generally occur unless there has been a catastrophic failure. Fluid levels (oil (engine and auto gear box), brake fluid, radiator fluid, window washer fluid) should be a very regular check, weekly if not a daily check. Tyre pressures (including for the spare) and wear (bulging, cuts, penetration by nails etc) should be at least a weekly check. Lights should be a daily check, they can go at any time The number of cars that I notice who have at least one blown light is quite a large number, some have multiple blown lights. Wiper blades should be at least a monthly check, rubber eventually becomes damaged, stiff, brittle, hard such that they, in operation don’t work properly. Seat belt webbing should be a regular check to ensure no fraying or other damage. I am impressed by and I like the cars whose lights when failed generate a warning, some warnings so clear that even which light which has failed is displayed. Seat belt warnings when one is not being used when it should be are often a feature of most new vehicles.

Now that many of a vehicle’s systems are able to be monitored by the vehicle’s computer systems, manufacturers have decided it is easier to be a nanny (and probably save lives) than allow motorists to just drive until some larger failure or an accident occurs. I also think from a manufacturer’s perspective that if a warning is generated and the owner is responsive enough, then this will reduce repair bills (in and out of warranty), and I’m sure it does save lives.

I do see where a feature like LKA may be inappropriate in certain road conditions, and in those cases is a retrograde (and dangerous) setting to be enabled. I think the difficulty is, how is it able to be monitored such that when the conditions are appropriate that a disabled feature becomes active again (either through automatic enablement or by driver intervention)? Currently it is probably a nag warning, one that ends up annoying drivers more than it assists them.

1 Like
39

Would you consider that seat belts are unlikely to contribute to accidents, although there will always be an exception; but a lane guidance ‘safety feature’ that [tries to] take control of the car requiring driver correction can directly contribute to an accident?

Perhaps you support drivers needing to get ‘checked’ in particular makes and models of car to manage their eccentricities akin to how airline pilots need to get type checked prior to being able to captain a type airplane?

I disagree there is any similarity between seat belts, intrusive features that require any amount of defensive driving to protect oneself from them, and a greater good. A seat belt protects in a crash regardless of where. A feature like lane guidance with steering inputs may make sense on an M motorway in Australia or the UK, an autobahn in Germany, or an interstate in the US, among others. On our local A, B, C, and worse roads? An ability to turn it off seems common sense - thus might never happen in the ADR?

1 Like
40

The difference with operating instructions is that there is no recorded question and answer demanding that you effectively give an undertaking as to how you will apply the content of these. We can assume that the vehicle records the Q&A session, with other logged diagnostics - the extent of which the manufacturer controls.

I’d argue that seat belt operation is more straight forward than driver assistance aids. In most complex software projects - and these are at the pointy end of the scale - few of the developers have a solid knowledge of how the entire system works. Most would only understand a piece of it. The test team might have a better understanding, and they should understand how it is supposed to function, which they will challenge. In a complex operating environment, it takes many iterations of development to gain a long list of potential failure circumstances to test for, and even more to accrue co-incident circumstances to test.

There’s two competing factors: Regulators (representing consumers) versus manufacturers (proprietary motives).

The regulators lack a deep technical understanding of these technologies. They understand the opportunity they offer, and the high level theory, but don’t have visibility as to how each product works in detail. Safety agencies, such as ANCAP and its international counterparts, provide standardised testing of these technologies. They gives us consistent rating of a very small range of their operating capability for new vehicles in controlled tests (e.g. destructive tests) as well as non-destructive controlled lane-keeping and forward collision avoidance tests. It would be prohibitively expensive for such agencies to conduct truly extensive testing.

The manufacturers don’t place equal value on safety. Some (e.g. Volvo) have made it a key differentiator, and (I assume) they generally put a lot of effort into developing their systems. Some manufacturers adopt a "let others invest and we’ll licence what we can or otherwise build enough to tick the necessary boxes. They have a commercial incentive to keep the detail of their technology secret. It is all proprietary - even though they might all contribute to some industry funded development and share the results. After that, they go their own way for commercial reasons.

With safety critical software, most of the safety built in is in normally invisible functionality. The user (driver) is aware of less than 5% of the system’s logic - maybe less than 1%. Most of the code in an advanced and reliable system deals with all manner of operational integrity checking, in support of performing the same basic functions. Forward collision avoidance is generally good technology, albeit with some challenges that aren’t worth exploring here. In general, if it fails, then it fails to save or reduce the severity of a collision that the driver has already failed to avoid. In other words - it doesn’t make anything worse, and crash statistics since 2015 generally bear this out the number and severity of rear end crashes has reduced. The US NTSB issued a report on these technologies which had a long term study of these technologies in trucks. The greatest gain in incident reductions per million miles logged compared to the unfitted trucks tracked was in forward collision avoidance (emergency braking) It still has challenges with wildlife; pedestrian detection and prediction; cyclists; motorbikes and so on.

Lane keeping is slightly different. As PhilT points out, it works well on freeways and national highways with lanes accurately marked. It gets tricky where there are temporary lane markings; lane markings worn away; lanes merging; degraded roads; dirt roads and so on. We are back in the bad old days of MS-DOS 2.1 with these technologies (for those that even know what that means) but instead of word processing and spreadsheets, which are not safety critical, we’re using these in cars, which clearly are.
The software should be able to detect circumstances where the lanes are not marked, and fall back into a fail-safe state. I’m not a fan of “driver alerts”; I call them “driver distractions”. The first version of forward collision avoidance was only a warning - i.e. a noise or vibration intended to alert the driver to the urgent need to reduce speed - which is highly likely to confuse a distracted driver rather than cause them to brake. The Hyundai lane keeping assist may have this issue to some degree. I don’t know. Such driver aids should not confuse nor distract the driver - that’s counterproductive.

Future systems will know whether the camera is clean or dirty, and probably wash it. They’ll have multiple cameras in each direction anyway. There will be in-built system redundancy, so that even if the central computer fails, it will automatically fail over to its backup - which will know the system state of the primary system. There are other technologies in the pipe, such as connected vehicles or co-operative intelligent transport systems (basically the same thing) which will likely come in the next decade.

The other challenge is the level of electronics in the vehicles. Most highly reliable computer systems operate in climate controlled, voltage regulated data centres, with all manner of systems to ensure optimal performance is supported. These rooms don’t vibrate, or suffer temperatures from -20C to 50C, because all of the tiny, delicate components in such electronics are soldered with minimal solder joins. Over time, these joins can fail - or worse - develop intermittent faults. That adds a whole additional layer of challenge to the diagnostic software. It is well and truly beyond the capability of the mechanic in the dealership to diagnose other than at a large, expensive subsystem replacement level. Manufacturers love selling you things that you cannot possibly understand. You can’t fix it yourself; and you have to buy a new one
maybe future systems will have more fibre optics than electronics - I don’t know.

On a positive note, the diagnostics built in can identify failed components that the driver can’t see - as grahroll notes. It’s not all bad news, and cars are undoubtedly much safer now than they were twenty years ago, but much less easy to fully understand.

2 Likes