Level 3 self-driving cars

[kevm14]

http://www.motortrend.com/news/ford-eng ... 4287E26019

So I guess Level 3 is really where the conversation is.

On the flip side, research from Swedish automaker Volvo shows that drivers at rest or distracted by electronics need two minutes before they can properly react and take over the vehicle. “And that’s absolutely impossible. That really rules out Level 3,” says Volvo Cars CEO. By aligning with Google/Alphabet’s self-driving research unit, Waymo, who made similar discoveries, Ford plans to skip Level 3 altogether. According to Waymo CEO John Krafcik, “Level 3 may turn out to be a myth,” and “Perhaps it’s just not worth doing.”

Most automakers believe that humans should assist the self-driving system in order to avoid a collision in complicated situations or emergencies. Others, like Ford and Volvo, deem that an inattentive and distracted human would only make the situation worse if asked to respond in seconds to a life-or-death situation.

What do you think? Should Level 3 self-driving vehicles allow distracted human drivers to respond, or should automakers skip that level completely and wait until more advanced systems that don’t need as much human intervention are ready? Tell us in the comments below.

[kevm14]

The biggest problem with the SAE-defined Level 3 is that drivers no longer maintain whatever driving skills they once learned. I am old enough to have grown up with cars where the most advanced electronics was a cassette player. The traction control I learned consisted of a clutch pedal, a parking brake and the skill to do a 180 on a snowy road if something bad was in your way. But just daily driving in cities and on highways keeps you alert and maintains your basic driving skills. Once cars are autonomous except for "complicated situations", drivers who are used to slouching in the driver's seat won't have a clue what to do.

I fully agree. You either do full trustworthy automation, or you stick to the Level 1 and 2 stuff where the car can intervene sometimes when the situation calls for it. This whole thing of "oh you might need to wake up and take control of the car to prevent an accident" is just a bad path.

[kevm14]

I'm not surprised. Many drivers have a terrible time staying awake when involved now on long trips. I understand on one end the desire for it, but I think full automation should be strictly on a separate set of roads. Mixing automation with existing drivers makes for a highly unpredictable set of circumstances.

Isolating big rigs and other fully automated equipment to certain roads would certainly make programming simpler and travel safer for either.

I don't want to give up the thrill of driving personally.

This, too, for all the points he makes. Mixing automation with non-automation is going to get really bad really fast if the numbers add up. Keep the self-driving cars to that group. Not that we can afford to do this across the US but we could roll it out over time, starting with Jersey-barrier separated automation lanes, right? I mean we do have HOV lane infrastructure all over the place already. That could just convert to self-driving only. Maybe the car has an RFID thing that opens a gate at each entry point?

[kevm14]

Agree with this, too.

Anyone notice how today's televised car ads show how much you DON'T have to pay attention?

@PrydeWater Yes, it is horrifying.

I'm not going to say I have extreme defensive driving skills, but I will say I (like other car enthusiasts probably) do pride myself on vehicle control skills. I may not be practicing the most safe driving maneuvers all the time, but expect to be in control of the vehicle, even in these situations. Which is why self-driving is actually offensive to me. And let's get past this "we just fill the road with self-driving idiots who can't hurt us so we can have the roads to ourselves." That is not how it will work out.

[kevm14]

And let's get past this "we just fill the road with self-driving idiots who can't hurt us so we can have the roads to ourselves." That is not how it will work out.

Unfortunately, my state (Nevada) is spending a lot of tax money catering to the self-driving fascination, but like EVs, it's only a matter of time before they're here. I'm fine with them as long as I'm not forced to give up driving. If that day comes, I hope I'm not around. If I am I guess I could always take the bus or some other mindless vehicle. I guess the one good thing is the people that now text, eat, nap, put on makeup while they are driving can do it without driving so that would be good.

Yes it would be, except the gov't will want you, too, in a self-driving car.

[kevm14]

I guess the distinction is that while getting stupid people (poor drivers) from controlling their own cars poorly is good, I think this argument makes more sense from a pure safety standpoint. And the thing is, I am not concerned so much for my safety (I am a little) from these people, but because they generally prevent me from driving the way I want to (less safe now, but more safe when these people are in some special lane somewhere else).

[kevm14]

This guy makes a decent argument.

I think we will eventually have driverless cars. However, I am very concerned about the safety of these vehicles, especially in the snow belt areas of the country. My first concern is that there isn't enough redundancy in the control systems. A single point of failure could result in an accident or death. I work with computers everyday. They do make mistakes. In an environment with extreme heat, cold, road salt, and moisture, I would be concerned that corrosion or just oxidation on a connector could cause a sensor to provide a wrong, but "in range" value to the control system, which would cause the system to make an incorrect decision.

Other studies have shown that as we add more electronics to vehicles, the long term reliability decreases. I suspect it is partially from the corrosion I mentioned above, but also from the "low price imported parts". If any circuit uses capacitors manufactured in China, you can expect those circuits to fail. I see these failures daily in consumer electronics, and auto makers are always looking to save money.

Can a photo sensor be disabled by road spray? Can a radar based collision detection system fail if salt water penetrates the sensors and effectively shorts them (which would erase the signal?

Until the reliability issues are addressed, I still want a person behind the wheel.

Making safe and reliable driverless cars is possible, but could we afford to buy such a vehicle?

Unfortunately, by many standards, reliability past the original owner is largely irrelevant. No one but me and a few people on this forum care about the reliability of a 10-20 year old car with 200k or more.

[kevm14]

I honestly don't think any car maker is ever going to get close to level 3. Some may try for a limited period of time, but with the first accident that is potentially caused by the software, the automaker's attorney's will pull that program instantly. The risk/reward ratio just isn't there, in my opinion. Just look at the fiasco at Takata as a good example. That is just one component that has several automakers on the wrong side of the headlines. Now add to that a situation where they're taking on the risk of virtually every single action the entire car takes? No way. Besides, do the majority of drivers really want to be a passenger in something that's essentially driven by the risk mitigators at a company? I'd rather be invited to a colonoscopy.

That is why it will take the gov't to legislate Level 3 and beyond, rather than the automaker doing it voluntarily. Which has been my conspiracy theory the entire time. I like driving. If that makes me a horse enthusiast by analogy, I am fine with that.

[kevm14]

A little tinfoil hat at the end but mostly I agree.

I wish we could get to self driving cars. And we will, in three situations. The car moving without a driver, movement of goods (Semi's) on point to point predetermined routes, and in localized high density areas where the speed will not get above 25 mph. (this one is questionable with the accuracy of current GPS technology in cities.)

The reason is simple. The AI that powers the vehicles will have to make a determination in an emergency situation to do act in a way that causes the least harm to humans, governments will never allow it to be built any other way. If the only way to harm the least amount of humans is to kill you, the driver, then that is what it will do. And for insurance reasons you will have to sign a form somewhere that says you are ok with that.

I do not think many people would be willing to give any AI the ability to kill them if necessary.

The first paragraph in particular makes a whole lot of sense.

[frankbreau]

I feel like a lot of this is chasing shiny new technology towards the goal of increased safety when simpler solutions exist. The next time you're a passenger in a car on any highway, look around at all the other drivers. Only about 75% of the people are actually looking at the road at any given time. Maybe I notice it more living near a college campus, but to me it's terrifying.

If there were a way to place a 1995 driver (i.e. more engaged, not distracted) in a 2016 car (with air bags, ESC), the roads would be a waaaaaay safer place to be.

If we can build a self driving car, or come reasonably close to one, why can't we build a car that knows when the driver's eyes are pointing out the windshield and not down into his/her lap?