Slate: Self-driving cars aren’t good at detecting cyclists... proposed fix a cop-out

[ORMojo]

https://amp.slate.com/technology/2018/02/self-driving-cars-struggle-to-detect-cyclists-bicycle-to-vehicle-communications-arent-the-answer.html

[jtakeda]

Oh great. so now we could be forced to wear sensors and all of the sudden if a cyclist gets hit by a self driving car blame will automatically be shifted on US.

Sounds like a good deal for the auto manufacturers

[tuscanyswe]

I think we are a long way away from implementing chips on all cyclist kid jogger baby dog etc out there.

Cant see it for a really long time. No one will accept to wear one so others dont need to pay attention while getting to work. No one will want to pay for it. + what if you leave it behind. it gets stolen. It malfunctions etc.

Nah long way away from those beeing accepted (imo).

[Ken Robb]

Heck, driver-driven cars aren't very good at sensing cyclists either. I have the scars and dents to prove it. :)

[93KgBike]

heck, driver-driven cars aren't very good at sensing cyclists either. I have the scars and dents to prove it. :)

+1 x4.

The only thing that could be more galling than being hit by someone not paying any attention at all, which really sucks, would be getting hit by a car with no driver at all. :confused:

[zap]

This is relatively old news.

But really, we won't have to worry about this for many years as those in the business admit that there are many hurdles to overcome.

[David Tollefson]

I've long said that the end-game of self-driving motor vehicles will be all human-controlled vehicles banned from the roadways. Including bicycles. And when it's put to a vote, what do you think the numbers will be?

I have enough trouble with auto lane correction - it's steering the car right at us.

[unterhausen]

the tech is unlikely to ever be there, especially since most of the geniuses involved want to have very cheap systems. From where I sit, pedestrians are even worse than cyclists because discerning them from a stationary object can be really difficult. And you have to ignore stationary objects to the side that you are not on a path to hit. That's why the first death was a pedestrian. I can see an attempt to sterilize the roadways to avoid any evidence of humans. Jaywalking laws on steroids

[Spaghetti Legs]

Shouldn’t be too hard. The Google-Borg knows where most of us are anyway.

[martl]

This is why i see an opportunity for autonomous cars outside of cities, but not inside them. Having, for example, lorries "talking" to each other wehen on the highway, forming autonomously steered "trains", for example, would help both accident numbers and fuel consumption. The space for an extra alne for such "trains" is easier to find than inside cities. Lorries are owned by companies and get replaced frequently, so the timespan for a transition until the majority of the vehicles are equipped with such technology is bearable. Private cars would be a very different cup of tea, doubt any politician would be happy to kill his career by telling people they have to buy new cars or walk.

Autonomous driving inside cities is two-edged at best. This is why:
If the massive hurdles toward an autonomous car which can deal with inner city traffic as we have today (with cyclists, pedestrians etc.) are solved, it would be entirely possible to envision a car which drops you at the doorstep of wherever you wanted to go and then buggers of on its own to find a parking spot (or, in lack of one, just circles the block until you come back).
This may sound nice at first glance, but would multiply the congestion (+ dust/emissions/noise) issues already at hand. I'm not so sure about the US, but we see in mid/western europe a growing number of people who prefer a redistribution of available city traffic space, away from the car. So the most realistic we could hope for in terms of automated driving would be something like an automated speed-governor not allowing the car to go faster than city regulations allow.

[oldpotatoe]

https://amp.slate.com/technology/2018/02/self-driving-cars-struggle-to-detect-cyclists-bicycle-to-vehicle-communications-arent-the-answer.html

Of course, that's dumm, but how long before you-all think actual self driving cars and trucks, w/o a 'monitor' onboard, will be commonplace? I think at least a decade. Expensive/scarce fuel, something will drive this, not just the 'gee-whiz' factor.

[paredown]

I agree with Martl--the payback seems to be on restricted access roadways where the problem of what hazards exist for the autonomous vehicle are reduced. I personally would love to be able to get on a freeway, switch to 'auto' and crack a book (especially for long trips).

The line in the article that cracked me up though (seeing that I drive and ride in New York)--restricting jay-walking. Good luck with that....

[unterhausen]

there is no economic case for autonomous cars on limited access highways. Waymo fired one of their employees for taking one of their cars on one and causing a crash. Apparently it's easier to drive on a city street.

There was a big push for "smart highways" back in the mid-'90s. That's just totally gone now, because it doesn't make sense. And you can watch the videos of Teslas heading straight for concrete walls if you want to know how good of an idea highway super cruise control is in a "value-engineered" car.

The truth is the tech is barely on the edge of always crashing, not the edge of working perfectly. It's just the old libertarian dream of personal rail transit (pods) put on rubber wheels. Transit is much better idea, and we could have a lot more of it for the money we have wasted on autonomous cars.

[oldpotatoe]

there is no economic case for autonomous cars on limited access highways. waymo fired one of their employees for taking one of their cars on one and causing a crash. Apparently it's easier to drive on a city street.

There was a big push for "smart highways" back in the mid-'90s. That's just totally gone now, because it doesn't make sense. And you can watch the videos of teslas heading straight for concrete walls if you want to know how good of an idea highway super cruise control is in a "value-engineered" car.

the truth is the tech is barely on the edge of always crashing, not the edge of working perfectly. It's just the old libertarian dream of personal rail transit (pods) put on rubber wheels. transit is much better idea, and we could have a lot more of it for the money we have wasted on autonomous cars.

+1...

[Mr. Pink]

Yup. If you want to save the enviornment, ride the bus/train and live in a multi use apartment building.

[martl]

there is no economic case for autonomous cars on limited access highways. Waymo fired one of their employees for taking one of their cars on one and causing a crash. Apparently it's easier to drive on a city street.
We have no background knowledge of what the car was designed to do, maybe there was a reason that employee was told not to use the highway.

there is no economic case for autonomous cars on limited access highways.

Why do you think so? I think it *is* a lot easier to design a KI for highways; there are few intersections and -in a normal traffic situation- no objects like cyclists/pedestrians to deal with - the initial article even says so, as does Carlos Ghosn (of recent fame) -> https://www.cyclingweekly.com/news/latest-news/renault-ceo-says-cyclists-are-the-biggest-problem-for-driverless-cars-208159 "Renault CEO says cyclists are the biggest problem for driverless cars"

Of course a fully autonomous KI would still have a function to deal with pedestrians/wildlife/other out-of-the-ordinary events, but the difference would be: An a highway, such an occurrence is always an emergency situation justifying an emergency brake maneuver, whereas in a city it might also just be traffic, and designing a KI that can decide which is which, that is the bit that's hard.

Also, highways could easier be equipped with outside support infrastructure, suchg as cables or antennas alongside the road, opening paths to a different approach - no more need for a fully autonomous KI, as the vehicles could "talk" to each other or even be managed from some entity that has the full information of all cars in the vicinity. As said before, this is not so realistic for privately run vehicles for privacy concerns, but feasible for commercial vehicles.

I don't know if it is harder to design that KI for higher speeds. I drive a car with lane assist and adaptive cruise, which makes it sort of half-autonomous already, and that works well (German Autobahn tested :D), even though it uses a stereo camera and nothing else, i imagine a full-scale system using radar as well could do an even better job.

I disagree that there would be "no benefits". Crash rates would definitely be improved, as would traffic flow. If, say, lorries would be run in virtual "chains" they could move up closer in theory, reducing drag -> fuel consumption. This option might be ruled out for safety concerns, though.

[merckxman]

Don't look like a robot!!!:
https://www.gpsworld.com/autonomous-car-hits-autonomous-robot-in-bizarre-collision/