Hacker News new | ask | show | jobs
by acadien 4109 days ago
"self-driving car" and "car capable driving itself on highway under optimal conditions" are 2 totally different things. The real excitement comes when I can tell my car to go pick up the kids from school or drop me off at the airport and then drive itself back home. Anyways this is a great step in the right direction but its not the game changing tech we're all waiting for.
4 comments

I bought a new car in Sept 2014. It uses radar to do adaptive cruise control all the way down to stop and go. It also uses camera's to sense the lanes and steer to keep lane. Also handles emergency braking and people cutting me off.

So basically, I choose the on-ramp and off-ramp and the car does the rest on the highway. I feel like it's the midpoint between a manual car and an autonomous car.

After a few months of using this every day, driving in a normal car feels different than it used to. It's like driving a manual instead of an automatic.

The car is a fully loaded 2015 Chrysler 200C.

Yep! I made the same call the 200C is the only car in that price range that offers full stop and go with active lane keep. I use it all the time and is easily the best feature on any car I've ever driven.

I believe some of the new Jeeps also have the new system but not sure if it does lane keep also.

Just curious, do you still keep your eyes on the road with adaptive cruise control? Or do you tend to do other things?
Mentally it's less stressful to get behind another car and just let the computer control the speed and distance the whole way. You stop caring if they are going 55 or 77. I just sit there and listen to podcasts. I stay alert though, and I've avoided the urge to read or something.
It would be an interesting experiment to measure actual alertness and ability to react.... I would hazard that while you think you might be alert you're probably less so as because you're not actively driving your brain is not ready to react in the same way as if it's constantly processing while driving.

Case in point - I've got two cars, our boring diesel daily driver and 380hp:600kg track day weapon which you really have to be mentally prepared to drive (no driver aids at all). Headed to the track one Saturday morning in said track car, a car in front of me pulled out of the lane to avoid a big a-- bird in the road. Partly because I have to be alert in that car and partly the car itself I managed to treat the bird more like an obstacle course and (narrowly) avoid it. The dumb thing started running off the road then back into my lane(!). Anyway the point is in my Corolla no way would I have reacted in time, while I think I'm conscientious I know my alertness level is way higher in the track day car - it's low, small, has a brutally fast steering rack - a single moment of inattention you'd get away with in Corolla would see you in the ditch in the other car. 3 hours drive in both cars and one of them I'm pretty happy to have a little break afterwards!

A byproduct of the environment, I think autonomous cars/controls are going to reduce the level of awareness and concentration and you'll have some pretty nasty "I don't know what happened" crashes when things happen that the systems can't respond to and human has to drive out of it.

Of course the flip side of the coin, and I can fully appreciate is the reduction of crashes where people doze off and plough right into the back of other cars. So I guess it's a risk reduction numbers game of which is the lesser of two evils?

I read something a while back where a rental company trialled renting cars with the emergency stop option - to stop, as you say 'people going to sleep and plowing into the back of cars'.

The results were surprising even to them I think - the crash rate went way, way down. Seems that a ton of accidents are of the 'I wasn't paying attention in traffic' type.

As the driver of a sports car and a luxo barge, I totally agree with you about the different levels of alertness required. It's not until you get out of a sports car do you realise how much you were concentrating. I think it's mainly the bumpier ride and the twitchy steering that forces focus. You're essentially never, ever driving with two fingers on the wheel.

I was thinking of a similar possible problem during my commute this morning, not with automated driving but with automatic equipment and systems in general. Does convenience breed mental laziness and/or decay? In our increasingly automated and comfortable existence, what experiences confront us with the need for high focus and alertness, or systematic problem solving? (Speaking of humans in general, not just engineers and other professionals who benefit from this kind of activity in their work.) So many things have been reduced (usually delightfully so) to point, click and enjoy.

Some physical environments for advanced age folks are designed with features to promote extra physical exertion, to help maintain the occupants' fitness, which in turn promotes safety. (Better balance, fewer falls.)

Similarly, we might consider building-in a baseline requirement of mental activity for the operation of certain systems. In the short term, this might increase alertness in the moment, as in the semi-autonomous driving example. (In that case, perhaps a UI that requires maintaining a certain amount of eye motion across the road, or particular posture.) In the long term, perhaps other (more challenging?) in-built exercises will promote mental and neurological health in people whose work and play would otherwise fail to be stimulating.

There was an article about crash statistics on roads that used to be surrounded by tree lanes [1]. There were as many accidents with or without the trees. Removing the trees avoids the opportunities to crash into them, so now people drive faster and carelessly.

I also think stress, in healthy dose, is necessary. I'd bet heavy on decay as a result of removing it. Until the potential self-driving car future where people won't even have to think about driving and will find other ways to be proactive.

[1] http://a395.idata.over-blog.com/1/74/49/41/SPECIALITES-GOURM...

I've been using radar cruise control for a few weeks now (waiting for lane keeping to come this summer as promised!) and I feel much more alert than without it. Obviously it could be an illusion, since the brain is really bad at evaluating itself, but it seems like I'm able to devote more attention to my surroundings and develop better situational awareness when I can stop worrying about the second-by-second job of keeping a safe distance from the car in front of me.
Personally I'll take the 5 edge case crashes a (insert time period) over the 10,000 any day. Not know anything about autonomous vehicles, though, I wonder if the edge case wrecks tend to be less serious simply because the safest choice will always be made in time of doubt I would think.
Do you find this experience pleasant? I mean being behind the wheel, not having anything to do technically, but still having the full responsibility and not being able to dose off or do other things with your eyes than looking ahead for a few hours? Even with podcasts I'd be bored to death I think. I'm glad to live in places where I can just take a train instead, read or even work on the way, and still arrive earlier than if I took the highway.
this is very subjective, but I do find it pleasant, given good car (ie bmw in my case). Even for a full day, but that's obviously tiring a bit. It might come down to personality type - i am not an easy to get bored one, which i consider very helpful in daily matters
I'm interested in the edge cases when the driver stops giving any input to the car (e.x. they fall asleep).

Does it follow the car in front through lane changes? Exits? Sharp turns? What if your car can't safely change lanes to follow?

Does the car attempt to alert the driver, or does it just slow down?

How far ahead does the radar look? What happens if you drive ~65mph straight at a brick wall?

I'm guessing he still keeps his eyes on the road, since poor merging by other drivers could plow into your side, something the active braking won't avoid.
I got a Mercedes E350 with the same thing last year. I got used to those features so fast, it feels like easy-mode driving when the car just wants to stay in the lane and keep the distance of the car in front of.

It does shut off the self-steering after a pretty short while if you don't move the steering wheel, and it doesn't work very well at high speeds, but a traffic jam on the highway is what it's made for.

I'm presuming that local constabularies and insurance companies adore adaptive cruise control.

A whole bunch of cars clocked at 12mph over the limit? That's a bonanza of low-mileage speeding tickets, which is good for police dept. revenue, and good for insurance company revenue, because it doesn't actually change the risk pool in exchange for a nice rate bump.

I had something similar on an infiniti a few years back. Literature on the 200C seems like it has the same "lane departure" protection. I don't recall that ever steering to follow the turns in freeways though.
It will to a certain extent. It's a powerful tug on the wheel and will follow the curve all the way through. You still need to keep your hands on the wheel or it yells at you.
Will one hand do?
Yes!
Didn't know stop and go was already deployed. I wonder how many people use such a system and thus how trusty it feels. I mean, it's kind of a large field-testing of self-driving car logic subset.
It took me a few days to "trust" the stop and go system. It is very disconcerting at first, especially since it tends to brake later than I normally would - I had to learn to "just let go" and now it feels very natural. With that being said, there are situations where I can clearly see traffic far ahead is stopped and I would need to slow down soon - in these cases the stop and go system doesn't react quite fast enough and ends up breaking way too late (and throws up an emergency brake signal). It doesn't ram into the car in front, but it can be scary.

So yeah, not perfect yet, but in 98% of the cases it's amazing.

I see the systems behavior relies on a little short-sighted data, and even safe, relying on direct neighbor for control isn't optimal and not broad enough compared to our way of thinking.

NVidia went hard on their mobile GPU computer vision system. They could recognize many cars, with this it can detect jams in time.

>in these cases the stop and go system doesn't react quite fast enough and

Wait... you, open an open road going full speed at a traffic jam, just sat back and tested if this was going to stop you in time? Without being sure it was going to work?

I had my foot hovering the brake in case it was going to not stop in time. When I say it "breaks way too late" I mean it comes to an abrupt halt, rather than a gradual slow down like I would normally do. There was no chance it was going to run into the car in front of me, but it wasn't too pleasant for drivers behind me
How about the 2015 Camry fully loaded?
"Mr. Musk said in a conference call that the self-driving technology was “technically capable of going from parking lot to parking lot,” meaning through cities as well. But, he said, Tesla will disable the autopilot when cars are not on highways or major roads, citing safety concerns."

I think the implication here is that the cars are capable of this but they are taking the smaller step out of legal and liability concerns.

With a secondary implication that, were the laws updated and the car's hardware "certified safe", they could theoretically just push a software update to enable last-mile autopilot on the car you already own.
My thoughts exactly. Hopefully it's true.
If that's the case I think it was a good decision. If anything went wrong I feel like the media would love to fixate on it as they did with the car batteries igniting previously.
It makes sense that exactly tesla might not want to be the first company to launch that technology. A major failure would be an existential risk to them (just from publicity and momentum).
He's already stated his desire for Tesla to be the first to market with autonomous vehicles.
With the capability to enable it by just a press of a button at Tesla HQ, he can wait for the last possible moment and still be first.
So what if they're capable? The tech is not available, end of story.
Judging by the article, "under optimal conditions" is an exaggeration. There was no citation of problems during traffic conditions.

My father sits through 2-4 hours of Los Angeles rush hour each day, and this is absolutely a game changer for someone like him. This is stellar.

This is solving the wrong problem. Car self-driving through 2-4 rush hours does not solve the problem. Not having 2-4 rush hours would.
This is the first step to that solution, though - if all cars are self-driving optimally, that 2-4 hours rush becomes much shorter.
Automated cars can in principle make subtle, shockwave-damping decisions in high density flow, zip together neatly at on ramps and make space for lane crossers at off ramps. And bonus, they can also in principle be doing useful things rather than occupying parking lots.

I expect a slow phase change to car-as-public-transport that never stops rolling when it isn't in maintenance.

Way more cars are needed during rush hour than other times, so I think even in the future cars will stop rolling.
I think it is likely that the rush becomes longer
the biggest change would come from carpool adoption
Perhaps make huge autonomous drones that can airlift your tesla from one spot to another, bypassing congested areas.
A drone with that kind of lift sounds hugely expensive to build and operate.
How is my use case a "wrong problem"? I'm providing a legitimate application of this limited technology and you're dismissing it with a distracting, unfounded argument.
I don't know how to solve the problem of having 2-4 rush hours. I do know how to solve the problem of having to manually manage my speed in traffic. It may be "the wrong problem" but it's a problem I face routinely, and a problem I'm delighted to solve for myself.
Get your job to let you work from the car on the way into the office for whatever face time you need, and you're on your way to cutting down the total time devoted to your job/commute.
But we are in a community currently in recoil from remote workers. On HN in the past week I have seen at least half a dozen front page articles about how remote doesn't work or how to do work life balance when remote. If people just cannot function without having their molecules in physical proximity to their bosses molecules you still get stuck with the dumb commutes.
Currently, most offices only need face time ~10-4 most days, but if your driving your out of touch with coworkers in normal business hours which is a hard sell. So often people only get to miss one of those segments either the show up ~10 most days or leave by 4pm most days but not both. Or, they might miss 1-3 days a week depending on meetings etc.

Now, suppose your self driving car let's you work while in the car. So, your still in the office from ~10-4 and get daily face time, but your only working 8-6. From a work life balance that's effectively zero commute time vs a killer 2h each way, with the added benefit that you get to avoid some of the peak traffic times.

The upside is it makes vary long commutes viable, the downside is it's it's vary energy expensive but with electric cars that's far less of an issue.

No to mention things like effective valet parking everywhere.

How much does a driver cost in the US? For well paid tech workers I'm wondering if it would be worth it to have a driver rather than spending X hours driving themselves.
You're seeing the recoil because it's expanding to place we haven't seen before. The outsourcing and remote trends gained traction in 2002, and despite a lot of failures and false starts, it's growing and reaching places and industries it hasn't been to before, hence the backlash by people and teams that aren't competent with them.
You wouldn't be a remote worker if you're still going in for daily face time.
Tesla already does this to some extent. TACC will go down to a full stop in traffic today, you just don't get the steering part(which isn't hard in stop & go).

It's incredible how less mentally wore down I feel after a long commute when using TACC. Still paying attention but at lot less effort required.

There are already cars that have this feature, he should look into them.

What Tesla is releasing at this stage brings absolutely nothing new to the table.

Absolutely. It's been around for at least a year or two.

My Landrover has this technology as an option as do most brands e.g. BMW, Mercedes.

Tesla is merely playing catch up this in regard.

Did it come as and over the air upgrade for all exisisting vehicles, with a promise of more upgrades along the way, and an even better upgrade of free long distance fueling?
That's probably why he is promising such a short timeframe to release.
And it's also why I'm baffled at all the people talking about how crazy Musk is and how they wish he'd stop promising the moon. This is cool stuff but it is not a moon shot.
The real excitement is when I can put those requests in to a network of public infrastructure and have them executed at roughly the same cost as driving for myself.
And we have our first day, week and month without fatal accidents.