jefnvk

Agreed. There is no reason that the internal workings of a system needs to be shared, it is extraneous information to the purpose of intracar communication. The car won't broadcast how it came to the decision it has made, only that it has made a decision and telling others what it is doing.

To break it down even further, even one company's system utilized between two different OEMs will likely operate in a different manners, as each OEM will have their own varying requirements on how their vehicle should react to a situation. All that needs to be the same is the communication format they are utilizing to talk to one another.

mr_bill

Forever amazed that someone could be reading/writing bikeforums on their smartphone and yet have absolutely no clue the thousands of software/hardware standards that make that possible.

Sausage, brought to you by sausage makers.

-mr. bill

squirtdad

people can and do handle the process the complexity of driving a car....... the brain is a wonderful thing which points out why this a hugely complex technical challenge.....

350htrr

Exactly, we are at the Betamax/VHS stage of AV. When the dust settles a standard system will come about that is truly workable and does the job better than anyone expected.

Ninety5rpm

I know some people with obviously pretty simple brains... and they still manage to drive. It's not that hard.

hotbike

Quote:
"Autonomous cars have a potentially fatal flaw: They struggle to detect and react to cyclists on the road. According to a January 2017 report by IEEE Spectrum, bicycles are generally considered 'the most difficult detection problem that autonomous vehicle systems face.' ..."

https://slate.com/comments/technolog...he-answer.html

squirtdad

that is not a surprise at all

Leisesturm

835 cyclists were killed by cars in 2016. Well over 5,000 pedestrians were killed by cars in 2016. Are bicycles really a huge problem? The author of that article seems to think so. In the comments that follow the article, Slate's side (the author?) doesn't even give cyclists the credit for self preservation that you give a small animal! Biased much?! The vast majority of cyclists that are killed and/or injured were/are behaving in accordance with DOT statutes. And to date all have been killed by human driven vehicles. It remains theoretical as to whether automated vehicles will exhibit the thirst for cyclist blood that humans do.

Daniel4

I'm glad these kinds of articles come out early so the engineers can work on them before rolling out the self-driving cars to market.

There's nothing like transparency and a healthy dose of skeptism.

Maelochs

We all have biases, and a thread like this brings them right out.

I am biased more like Daniel4. if this is an issue now, it is an issue engineers are working on now.

Whether the article is accurate or even actually based on real data ... Anyone who does the research can see lots of "pseudo-science" articles where a simple study is twisted into some headline which has nothing to do with the actual study ... but will get clicks.

But ... I am not here to teach people to think or to ask people to be honest.

And best of all ... no matter how accurate or wildly ridiculous our predictions, regardless how well- or unsupported our biases are ... None of it has anything to do with what will actually happen.

rseeker

(Replying to the topic, not anything anyone said) I think we can look forward to hearing the argument that not having a system for bicycle-to-vehicle communication on your ride is contributory negligence. Yay?

genec

Article is a year old, and does point out that AV companies like Waymo have a good success rate of detecting cyclists.

CliffordK

It isn't surprising. It is much easier to see a house rolling down the road on wheels than to see a small 2 wheeled vehicle. Plus, bicycles tend to move at odd speeds. Faster than pedestrians, but much slower than cars. And, everyone rides at different speeds from 5 MPH to 30 MPH.

If one can analyze frequencies, then IR should help some in colder temperatures, but it would tend to break down a bit when temperatures approach 90+ degrees.

Keep in mind that the problem of the cyclists is complex as is often noted in A&S.

The 800 deaths and 40,000 or so reportable cyclist/car accidents are not just hit from behind. They likely can be divided into two general categories:

• Predictable Actions by Cyclist, Stupid Actions by Automobile
• Predictable Actions by Automobile, Stupid Actions by Cyclist

Or, likely a combination of the two, so a stupid action by a cyclist PLUS a stupid action by the driver.

So, with a self driving car, we could potentially see fewer of the Stupid Auto type accidents, and as noted earlier, these can be remedied by software updates. So, for example passing with inadequate space, bumping with a mirror, or pulling out of a driveway in front of a cyclist could be minimized. Even put devices on doors to not open into traffic, or slowly and deliberately open plus flashing lights so they can be detected early.

To detect and prevent accidents with stupid cyclists is a big problem that plagues both drivers and will be a thorn for self-driving cars.

Some will be to implement algorithms to predict cyclist's actions, and that will take good programmers. For example, I had a near miss a couple of years ago... right turn from stop sign into a bike lane... should have been ok, except there was a bus stop with a shelter and pull-out just beyond the turn. So, while cars would stay to the left of the bike path, buses crossed it. So, in this case, a self driving bus would have to detect the potential for a conflict, increase predictive estimate that a cyclist will head for the bike path, and alter speed to avoid the possibility, whether or not the cyclist appears stopped or moving.

Of course, there will always be that one in a million situation unforeseen by the programmers, but trivial for a human to react to.

Part of the answer will be to program in basic defensive driving skills. So, for example, at a 4-way stop. Stop. Don't proceed until: 1) one can clear the intersection before all other traffic, or 2), one can ascertain that all other traffic is yielding (nobody is speeding into the stop at a speed too fast to stop).

That 4-way stop can be extended to right-of-way situations too. Analyze potential cross traffic, and if the cross traffic doesn't appear to be stopping, then alter course to avoid a collision. And, again, it is a bunch of calculations that a computer should be better than a car driver, if given the right instruction set.

One could see a new level of road courtesy. See a pedestrian peeking out between cars, and stop and give them the right-of-way (hopefully they won't get slammed by the human driven car in the next lane over). Do too much of this, and the self driving cars will be taken advantage of... especially by stupid cyclists trying to save those precious seconds on their commute.

CliffordK

Like riding your bike at night without lights. Or stripping off those silly reflectors from your bike, then riding it without lights... and with dark clothing.

Could one design batteryless tags that could be zapped with a signal from a distance to make them react, somewhat like those anti-theft devices stores like to use? Simply add them to bike reflectors, or similar mounts.

How do RFID and RF tags work? - Explain that Stuff

My guess, as above, is the issue is not as much detecting the cyclists. The article says the cars are supposed to be able to detect game animals. But, rather predicting what those crazy cyclists are going to do next which is a much harder problem.

scott967

I suspect V2V may be mandated in certain situations. Meanwhile I see San Diego is planning to install a sensor grid in its streetlights that will allow it to track everything moving.

scott s.
.

I-Like-To-Bike

Given how many vehicles driven by humans (many driven faster than 25mph) safely pass cyclists every day, day after day, everywhere, under all road conditions, (and conversely how many cyclists arrive at their destination safely everyday without being struck by motorists), a good case can be made that human drivers also have a good success rate of detecting cyclists.

CliffordK

800 deaths in 300 million people (about 2 to 3 deaths per million) people isn't bad. Perhaps cyclists are good at detecting cars. Or, as I like to say, riding where cars aren't driving.

But, 50x that many are involved in reportable accidents, and the statistics are a little less pleasing.

I-Like-To-Bike

Of course how many of those so-called 50x reportable accidents are the result of innocent lamb cyclists undetected by negligent/careless motorists is anybody's biased guess, eh?

Of course promoters can extrapolate their guesses about how good AV cars of the future will be from simulations and closely controlled "tests", and extrapolate that into whatever statistics fit their promotion/selling agenda, especially if/when the promoters and their press agents are in control of the release of the test data and results.

Maelochs

The cycling population is sufficiently small that even if AI cars killed More cyclists, most customers wouldn't care.

But that one article, already old, and based on who knows what data ... is really saying, "This is one of the problems engineers are solving now.

If people wanted to list the various issues which engineers had to solve to get AI cars on the road (which they have been for quite a few years now) we would need almost as many words as have been wasted in this thread to rage against the inevitable, or to supposedly help the inevitable occur.

AI cars are already a thing, folks. The question is how soon they will replace 90 percent of the current self-drive fleet. And in all likelihood All types road accidents will drop and farther the fewer self-drive cars are out there. Sorry yo doom-mongers can't use this topic effectively ... but don't worry, things are actually worse than you seem to realize. it just isn't AI cars that is the biggest threat to survival.

And really ... if the answer is manufacturers embedding an RFD chip somewhere in a bike where it cannot be removed without destroying the bike, that will just happen. Big Brother already knows whatever he wants about you---riding with Garmin on the bars and a cell phone in the back pocket, your car and house wired into the global circuitry ... satellites overhead ....

Bike manufacturers wouldn't think twice about putting a chip in your bike ... and every time a stolen bike was recovered because of the chip, people would post here praising it... "Why did anyone ever object?"

I-Like-To-Bike

Sure, why waste words on problems and issues? After all, if engineers are "working on a problem" the problems and issues are as good as solved; forget about it!

The driver-less vehicle transportation future solution/revolution as forecasted by the promoters and speculators and their press agents and fanbois is inevitable. If in doubt, just think and repeat: smartphones, moon rocks and buggy whips; all problems and issues are inevitably successfully solved, eh?

CliffordK

Part of the issue is that if the car is on full auto mode, then the liable party will be the software manufacturer, for example Google who may well have very deep pockets.

If it is something like a cyclist running a stop sign, then it may well be ignored... cyclist at fault.

On the other hand, if the car rear-ends a cyclist who isn't doing anything illegal, or perhaps the car pulls in front of a descending cyclist that should have been visible, then payouts could be HUGE.

And, with all the sensors on the car, they may have the data that will dig their own grave.

We've also seen huge recalls for extremely rare events. Firestone recalled a bunch of tires that should have been old enough to be worn out or replaced anyway.

Plus, it could be extremely bad publicity. Of course, there is the issue of population amnesia, and any news provides beneficial publicity. But, too much bad news could sink a company's efforts.

I-Like-To-Bike

Are autonomous cars really safer than human drivers?

https://theconversation.com/are-auto...-drivers-90202
Extract:

"Much of the push toward self-driving cars has been underwritten by the hope that they will save lives by getting involved in fewer crashes with fewer injuries and deaths than human-driven cars. But so far, most comparisons between human drivers and automated vehicles have been at best uneven, and at worst, unfair."

Feel free to ignore any thought that the future of self driving car may not be as predicted by the puppets of the promoters' PR machine who relentlessly repeat every hint of positive spin they hear, read or fabricate.

hotbike

More pedestrians than cyclists get killed, yes, But pedestrians don't take market share away from automobiles.
Bicycles are seen as drawing away customers, people who would otherwise be riding in a car.
$$$

I-Like-To-Bike

Unlikely that anyone but a handful of bicycling ideologues see bicycles as drawing away any significant, or even a minuscule, slice of market share away from automobiles.

$$$? More likely not even a tiny fraction of a ¢

Maelochs

Well, I tried. I stand aside and let the avalanche plummet downwards.