unterhausen

the head of development of that autonomous car happily let it pass a cyclist far too closely. This is on a low traffic two lane road where another lane was available. The head of the entire effort. So yeah, u gonna die.

Ninety5rpm

Nissan is barely on the radar of companies at the forefront of AV technology. It's all in the R&D stage. Of the myriad issues facing each AV company, getting bicyclist passing and treatment in general is probably not at the top of the list, but it will be handled properly long before production. They can't afford the marketing nightmare of hitting a pedestrian or bicyclist.

unterhausen

the R&D phase is probably the most dangerous, since they are out on the roads with unproven software. Like this video https://www.youtube.com/watch?v=zGoE...ature=youtu.be

Ninety5rpm

Phantom AI ?? Who is that?

I see they're very new, near the bottom of the list of companies with permission to test in California.

https://www.dmv.ca.gov/portal/dmv/de...nomous/testing

But even this most dangerous period is proving to be safer than human driving, mile for mile.

Chris0516

Simple and succinct.

Rollfast

It's seems that many of these thread morph into a similar version of a similar argument, which in turn suggests that it's


a) misunderstood
b) not comprehended
c) annoying
d) all that the your little dog too!


I don't even understand how this got to be a thing here and I try to see what the thread are about but I often bail shrugging my shoulders.


Why not start a site about driverless cars and let's all get a new bike?


Bikes are damned neat-o!

Ninety5rpm

Driverless cars will affect every aspect of our lives perhaps more than any other technology has. Two aspects that will be affected greatly are cycling safety and getting more people on bikes (advocacy). That's why these discussions are here.

I-Like-To-Bike

Perhaps you should consider prefacing your assumptions with an occasional "IF" or "in my opinion", so as not to seem so star-struck by the PR hype.

Chris0516

I knew this was going to happen.

When I originally said it, back when autonomous vehicles were brought to market. People said the autonomous vehicles would be safer than human-driven vehicles.

Now they have to eat crow.

Ninety5rpm

You're conflating these driver-assist systems with autonomous vehicles. Not the same!!!

genec

A lot of folks are doing that... including the drivers of two crashed Teslas... who apparently didn't bother to read the manual.

The semismart cars are causing all sorts of grief, as their required human drivers fail to maintain control, and blame the car.

There needs to be a huge sign in these things, right on the dashboard... "Warning, human driver required to maintain control."

Rollfast

I don't know about that.

Ninety5rpm

Well, I don't know that either, of course. But I'm quite confident that's where it's going. However, I was also confident Trump wouldn't win the primary, let alone the general...

Rollfast

Semi-autonomous still means the operator wasn't doing their job.

Ninety5rpm

Exactly. I'm surprised there isn't even more carnage from semi-autonomous.

Years ago Google observed human drivers getting so lackadaisical in their semi-autonomous cars that they abandoned that approach and decided to go straight to Level 4/5.

whitecat

Stupid designs and intrusions into our way of driving are getting to the point enough is enough in new cars. When I am driving, I am going to steer the car and I am going to choose where it goes. That is precisely the reason that I will not buy a new car these days, it really irks me if some computer chip started tugging at my steering wheel because of some paint on the road. There must be a way to turn that damn thing permanently off, even if its done by removing components.

My opinion is the same for that "lane keeping system" as with all other similar systems - if you cannot keep the car in a lane by yourself, or cant park by yourself, or cant start on a hill by yourself, chances are that your skill isnt enough to be a proficient driver, better turn in the licence and sell the car.

CliffordK

If one dozes off behind the wheel... it might be nice if the car could safely self-park, although stopping in the middle of the road would not be appropriate.

So far I haven't driven a car with any steering assist. Mom likes the adaptive cruise control on her car. I suppose I gave up on cruise controls years ago because I like to adapt my speed to road and traffic conditions.

Mom has lane beeps on her car. An interesting concept. Perhaps good for some things, but it beeps for things like avoiding potholes (sinkholes) on not very busy rural roads. Thus, one also quickly learns to ignore many of the beeps.

I certainly wouldn't want a car that deliberately steered me into potholes. At the same time, I suppose I wouldn't want it to steer me away from them either.

noimagination

IANAL, but I'm pretty sure that Tesla can write in all the liability disclaimers it wants, but until the liability issue (for the particular case of an accident occurring while a car is in Level 2 driver assist mode) is tested in the courts the level of protection to Tesla provided by those disclaimers is unknown.

-----------

Road Fan

@unterhausen, I don't really see anything that indicates it can or cannot detect cycles. I assume that when the BMW is switched (commanded is a better word to follow the lane boundaries, in that algorithm it does not consider the effect of vehicles or objects that are not in its path, OR it is not capable of reliably detecting and tracking (necessary to steer around) a cycle in the BOL (for example). In Level 0 and Level 1 systems, often OEMs will limit the performance envelope (i.e. turn it off when the speed is less than 35 mph) when the system known performance limitations cannot be resolved with the technology as it is.

If the bike is in-path I assume the ACC would have had to function, but reliable detection and tracking can still be difficult. An automatic steering function may be developed to the point where the car can stay within its lane, but not to the point where it can reliable establish clearance to cross the lane boundary. This capability would be needed for the car to violate the lane safely in to achieve a 3 (or 5 or whatever) cycle passing gap when practicable. Essentially, is the car capable of determining when executing the cycle passing gap is practicable?

Another possibility is that the BMW does detect bikes, but the required (was it required in the testing locale?) passing gap was not programmed. It might or might not be a big software change, but as an update it would be a software change. Hence "a miss is as good as a mile" would be what's built-in to the steering algorithm.

So, I dunno the exact situation for that BMW development car, but these are some of the system design complexities as I see them.

Road Fan

I agree with your first. I think the second one is grey. If what you mean by "collision avoidance" is forward vehicle collision mitigation ISO 22839, it sets up requirements for when an object is in path, at least partially.

Road Fan

Yes! An R&D vehicle does not always have all the "driver expectations" built in.

The software has to have some qualification. Jumping the program pointer to a random location would be absolutely unpredictable in its outcomes.