1.7M Hondas are being investigated for phantom braking
arstechnica.com
arstechnica.com
The 04 R1 I used to ride had a goofy double throttle system where the cable directly drove the first set of intake butterflies, but a duplicate set downstream was computer controlled via servo. This was the generation that introduced catalytic converters to bikes, and I presumed the intention was to limit throttle opening rates via computer overriding what the rider had done when opened too quickly. Conventionally that transient pressure drop situation is handled by throwing a bunch of excess fuel into the air stream to not hesitate and stumble from most the fuel falling out of suspension and clinging to the walls, i.e. they were cleaning up emissions.
Even this seemingly benign and dead-stupid bit of a computer override almost caused me to crash on multiple occasions. If the rider opened the throttle aggressively far beyond where the servo operated throttle was at, and immediately began dialing it back to reduce the acceleration, the servo operated throttle would continue opening to meet the still more open cable operated throttle's position. So rider is telling bike to accelerate less, and bike is accelerating more in a game of computer catch up. This is not only very unexpected on a naturally aspirated ICE, but outright dangerous on a nearly 200HP sportbike.
At least it was relatively trivial to remove the plates from the servo actuated throttle shaft and let the computer keep pointlessly twisting a bare shaft.
https://www.motorcycledaily.com/2002/02/27february02yamahamo...
Wait, when you say "downstream" do you mean the butterflies were in series on the same intake path? I'm trying to visualize what a small time delay and rate mismatch would look like here.
Yes, in series, literally redundant throttle plates on a common set of ITB runners:
\ / \ / \ / \ /
| | | | | | | |
|/|-|/|-|/|-|/|-cable/rider-operated
| | | | | | | |
|/|-|/|-|/|-|/|-servo/ecu-operated
| | | | | | | |
* * * *It's worth noting later versions of YCC-T are servo-actuated using a single throttle shaft. This early version was probably just something slapped on at the last minute using off-the-shelf components, and they didn't refine the control software enough to do things like at least resync the logical servo position with the cable throttle position when the rider has started closing the throttle from an overshoot position. It was just a very naive rate-limiting filter of sorts where the servo is continuously chasing the actual throttle's instantaneous position.
The experience occasionally resembled a laggy high boost turbo car where tq keeps increasing despite a shrinking throttle open %. Kind of amusing in a car, Tesla FSD levels of idiotic on a ~1000hp/ton bike.
The R1 didn't so much as accelerate as it teleported at wide throttle angles in 1st gear, combined with no squishy rubber isolation bits adding mechanical latency to the system. The phenomenon was very apparent and nerve-wracking in the right conditions.
But things like this are certainly just one more reason that I do not ride liter bikes.
I've never heard of a car with a second throttle for traction control. You might be confusing variable intake geometry vanes.
See: http://www.eatsleeptinker.com/2014/06/02/asc-throttle-body-d...
But I should be careful what I wish for. Current BEVs have many user-hostile clown-car features.
The excessively (and negatively) complicated addons described in the thread are not inherent to ICE cars. Take any car from 80s or 90s and the behavior is predictable and simple. It's only into the 00s that manufacturers started making things too complex and as a consequence, fragile and unpredictable.
They should result in corporate manslaughter charges ahainst the manufacturer.
If we let this slide, a JavaScript library with 100 known vulnerabilities will be running on 10 year old kernel controlling your breaks and then you will get 'oops, something went wrong!' Message and die because you car started mining buttcoin
And that ignores the even more complex problem of self-driving which is something even aviation, which has a lot less opportunities for external errors, doesn't even try to tackle.
I'm still on team "put computers into cars" for the long term benefit of ending routine automotive collisions but the medium term is frightening.
Second time was in a bend, with a big sign on the far side of the bend (me on the inner lane). Same stuff. If there had been oncoming traffic I would have certainly hit it because the car became uncontrollable instantly.
Vision based computer systems really need the persistence and object recognition factors. A strong specialized faked intelligence might be sufficient, more so if it's less distracted than humans.
There isn't a whole lot of room on those bridges for evasive maneuvers either.
In both cases - aircraft and cars - the primary goal was reducing weight.
But this term is not intended to refer to driver-assistance features.
The part you're thinking of is the throttle (which is why the accelerator pedal is sometimes called the throttle pedal), which controls the flow of air into the engine.
The flow of gasoline has always been automated as a derived value from this (and often, other factors) -- even when engines were entirely mechanical; this was the function of the carburetor.
Some diesels use so called J-brakes which effectively function as throttle plates to induce engine braking, but they are exclusively for slowing down the vehicle.
That wikipedia article is a lot of unsourced nonsense written by complete idiots who clearly lack subject-area knowledge
The goal for drive-by-wire in cars was primarily emissions reduction. Many cars "smooth" throttle inputs to improve mileage and emissions. In some cars "eco" mode exaggerates that smoothing, or changes the mapping of the throttle to allow for much more control over a lower portion of the throttle range. In others, "drive" is smoothed, and "sport" mode is not (unfortunately "sport" mode often makes the transmission far too aggressive.)
It brings mechanical simplicity, because cruise control can be integrated in.
Drive by wire throttles were notoriously unreliable for almost a decade after their introduction. They get gummed up, the actuators fail, the position sensors fail, etc. It's a laugh that the article claims they're "easier to service" - they're usually buried in difficult-to-access places and there's nothing "serviceable" about them compared to a true mechanical throttle that can usually be completely rebuilt.
https://en.wikipedia.org/wiki/Stoichiometry
If you don't do that then the excess fuel goes back out unburned. Multi point injection has changed this process a bit in that compared to the first single point injection systems the fuel is mixed with the air on a per-cylinder basis leading to better fuel economy and even more efficient combustion. This is also why these systems have air mass sensors, the amount of air is an important thing to measure in order to determine the right amount of fuel. Essentially fuel injection systems they have all of the complexity of a servo system, and the number of failure points compared to a direct linkage is huge, especially when you have a block that is split into two cylinder banks, V or Boxer.
I know this is very off-topic but don't beat yourself up about making quick decisions like this that cut that kind of persistent, simmering rage out of your life.
We have enough of this stuff with people in our lives, often unavoidably, so letting appliances/devices/products cause it is not something we have to tolerate. If a _thing_ makes you feel unsatisfiable anger, get rid of it.
Isn't the correct term "vote with your wallet"? "voting with your feet" implies switching jobs or emigrating.
I couldn't get on with the thing, but I felt bad about selling it to anyone because the resonator was more trouble than it was worth. Some frustrations should not be sold on.
So I kept the very nice guitar case, and gave the guitar away to a talented, broke musician, who used its rather scratchy, broken sound for effect on a couple of recordings.
I could have sold it for a few quid but I have never regretted this; I turned my frustration at this thing into a gift and a bit of joy for someone else and then moved on.
Not something I'd recommend everyone does regularly, because if we all do it, it just prolongs poor product design by failing to reject it, but it helped this one time.
(It also taught me an important lesson about musical instruments; the next guitar I bought was better, but more basic in appearance and already scratched from the shop.)
I had a similar experience and called an attorney, who explained that, in the U.S., you can really only sue for any actual loss (whether to the vehicle or to you). If you weren't injured or killed, and didn't suffer extreme anxiety or something like that, then you suffered no real loss and really have nothing to sue over (unfortunately). You can't sue for what might have happened, only for what actually did. (You probably could have sued, at least in small claims, for the loss of use, cost of repairs, decrease in value, etc, if you wanted to take the time.)
With that said, you might have been able to get them to take it back under your state's lemon law, if you hadn't already sold it, but the lemon laws generally only cover multiple defects over a period of time anyway. :(
Now I'm going to assume that adding the 4WD option to your sedan isn't that good at increasing safety. Yeah, more driving wheels does improve control and will probably reduce fatalities. But from 28 dpm to 0 is going to also be attributable to statistical noise. (Case in point, IIHS' 2014 statistics have the 4WD version at 25, and the 2WD at 11).
Also note that the death rate is in terms of deaths per million registration years and only includes the driver of the car. So for the C-Class, which sold about 80k units in 2014, one driver death in a 3-year span would be a '4' on the rate. The intervals on this data are very wide, and every year relatively low-volume cars will score a '0' just by statistical chance.
[1] Or sedans and station wagons if you insist.
Passive systems like ABS are pretty much the only thing I have ever found value in, and even then it's been a mixed bag when operating near limits. Never heard of a car locking up on the freeway because an ABS fuse went out...
Because human beings have such a terrible track record that even "dumb" computers running code written by shabby programmers can do better. Even with these errors the computers are still doing a better job. The NIH studied Automatic Emergency Braking systems and found that, "With a 100% market penetration rate, fatalities could be reduced by 13.2%, and injuries could be reduced by 9.1%."
We only have 1 side of this story - that a few people have been surprised and had accidents. We don’t have the stats on how many people the active braking has saved from a collision. (BTW I have a CRV, and have experienced sudden braking myself, both with and without something in front of me, but it hasn’t caused an accident yet.)
ABS is a computer between the brake and the physics, and there’s widespread agreement about the benefits of this arrangement. So maybe the problem isn’t the computer, or the idea to use them in the loop. Maybe the problem is an incomplete understanding of people, or maybe it’s just a sensor defect or manufacturing issue. (Which, I would agree, absolutely should be considered when plugging things into a computer controller.)
We have lots of computers in the physics loop, and very few of them give any of us pause. ECUs are standard these days and are critical to fuel efficiency and reducing environmental impact. Self driving cars are on the way and are a whole new level of computer control. I think I’m still in favor of self-driving cars and the potential they have to reduce traffic, collisions, and environmental impact...
Assume you are traveling down the 405 at 80mph during rush hour.
What is the worst case scenario for an ABS failure mode?
What is the worst case scenario for an emergency braking failure mode?
Okay, bad driving choice there. You are 1.5x more likely to die in a collision driving 80mph than if you were driving the speed limit (65mph), since collision energy is proportional to velocity squared. ;)
Worst case for ABS failure vs emergency braking failure? Seems like both can easily result in death, and I don’t see reasons yet to believe one is more likely than the other. We probably have a lot more data on the number of deaths lack of ABS has caused over the years than computer braking systems, and the good news is that we can fix computer braking systems right now and forever.
The big difference here is control, not outcome. Having brakes seize suddenly is surprising, whereas going from static to dynamic friction when slamming on the brakes isn’t very surprising. Of course we strive to minimize surprise, and of course it’s preferable if cars don’t slam on the brakes without cause.
In the article, there were several links at the bottom with stats on the impact on Volvo cars in the US: https://www.iihs.org/topics/bibliography/ref/2111. The stats aren't directly equivalent, but I think it helps to support your argument.
That said... I had the rental because of an accident that would have been avoided if my own car had frontal collision detection and automatic braking. Assuming it had worked.
It either didn't have lane-centering or it was disabled, but I think in general that's a way more noticeable input than braking. If you have a weird input in steering (ie, a strong gust of wind) your instinct is to steer the opposite way, but in general I think most people that have weird input with acceleration/braking tend to respond with just braking universally.
Unexpected surprises can be dangerous. I feel like safety related systems really should be as simple as possible.
I like ABS but in one of my vehicles I had an issue with it randomly engaging when it shouldn't due to rusted and cracked tone rings. It's actually pretty scary losing the majority of your braking power when you desperately need it.
Agreed. It's bad both ways. I would say the more dangerous failure mode is the one which results in the most unexpected amount of delta between input and output.
If you are already in the process of trying to stop, there is going to be some finite difference between expectation and reality, even if the braking system does not perform perfectly (or at all). Your brain is already thinking "ok let's find a way to slow this thing down". You might have a few precious milliseconds to go for a shoulder or other external soft-stop opportunity.
If you are inputting zero braking signal and receive a 100% stop force arbitrarily, that is pretty much an infinite level of surprise. Incidentally, this may also provoke any adjacent ABS failure modes.
I think it's a great idea for the driver behind me to have such a system. It makes them less likely to rear-end me.
> I think it's a great idea for the driver behind me to have such a system. It makes them less likely to rear-end me.
Yes. But it's a scary idea for the driver in front of you to have such a system. If they (their car) suddenly goes into hard braking for no visible reason on the highway, you might be rear-ending them.
Not really a world I'd prefer to live in.
I would guess the car detected the diagonal line across my lane and thought I was going to steer into the truck. It's probably a bug that it didn't give priority to the lane markings it was already following.
A lady from Honda called me a few days later to ask how I liked my new car. I talked to her about this and she told me I should contact the city to tell them to clean up the construction markings. I suppose she was trying to shift blame and liability off of Honda.
I do hope they solve the phantom breaking issue, but I hope it doesn’t neuter the collision avoidance system.
Had a 2015 Accord that I had to sell due to... Reasons and boy di I miss that car...
Never had any issues with that one fortunately
It's a huge PITA that these systems work most of the time but have glitches occasionally which are completely unpredictable and distracting. That's unacceptable imo. I'd rather not have it at all
I think it still makes sense for the average US driver, who is quite bad and not attentive even when not on the phone. For the rest of us, this is a misfeature.
Yeah, right. Blame it on "old people". Certainly not the code base with 10,000 global variables:
The design review found things like: Other egregious deviations from standard practice were the number of global variables in the system. (A variable is a location in memory that has a number in it. A global variable is any piece of software anywhere in the system can get to that number and read it or write it.) The academic standard is zero. Toyota had more than 10,000 global variables.
[1] https://news.ycombinator.com/item?id=19613055 That original link is now broken. I don't think Toyota paid to scrub it, probably just decay.
To all appearances, the way it is set up on my car is designed to prevent phantom braking and does so.
The big orange warning flashes occasionally when there is a large speed difference compared to the car ahead. This can happen on an off ramp, or when a car ahead is turning right.
In these cases, it's never braked on its own, and it seems likely that it would only brake to mitigate a collision rather than avoiding one. This, to me, is a reasonable tradeoff that surely is by design.
When using the adaptive cruise control, it once or twice (in well over a year) has reacted to a car in the wrong lane, for instance, on a three lane highway it detected the vehicle in the left lane momentarily and slowed slightly. But it didn't panic brake, and it's very rare that it has an issue with lanes, even on sweeping curves.
The only time it has actually panic braked was when someone ahead of me slowed to turn right, and I was startled and hit the brake and it assisted so as to brake much harder. It seems like it was activated only when I hit the brakes and only because of the velocity/abruptness of the pedal movement, which is exactly what it's intended to do (may be classified separately as brake assist). I didn't need to brake that hard, but I wasn't rear ended so...
The final glitch I've had is that when driving in heavy snow/sleet/rain the radar stops working when gunk covers the sensor. This has only happened once and when it happened, it only shut off the adaptive cruise control without doing anything else undesirable. It's not clear to me if there is some reason there isn't a smooth cover to prevent accumulation.
Sometimes incredibly weird electrical glitches can arise from somewhere that the average service place will not find or even really look for. I had a different car (not a Honda) that was about 6 years old, no visible rust, and seemed afflicted by electrical gremlins. It turned out that due to poor design of the taillights, water had collected in them, and the exposed wiring had corroded to where it looked like it had been at the bottom of the ocean. I only found out because one taillight finally stopped working at all. But it was a lesson on how simple damage to wiring can be hard to diagnose, because inconsistent voltage or whatever scrambles the computers. In theory, that car had a built in diagnostic function to warn a taillight was out. But it never registered at all.
I expect a Japanese car to not have problems like this until 200,000 miles in the rust belt. But everybody tends to rest on their laurels.