Self-driving vehicles.

[Funkworks]

Wondering if anyone here is interested in vehicular automation (I think it's the proper name for it).
I'm quite intrigued by it and might try to find out how far along they are in better detail.
Videos, news, articles, techs, etc.

 

[Hobie1dog]

I remember many years ago when MIT built a car that went from Mass. to California without a driver.

 

[dhbarr]

Just need them to be better than the average driver, not better than every driver in every circumstance. Here's hoping I can drink my coffee and read the news on the way to work before I retire.

 

[cls]

I hope to never share the road with anything running SAE level 4 or 5. Even level 3 is deceptively too much assist but at the same time not enough for most drivers.

 

[dhbarr]

I hope to never share the road with anything running SAE level 4 or 5. Even level 3 is deceptively too much assist but at the same time not enough for most drivers.

Can't be any worse than all the people fiddling with their phones while sleep deprived and late for work at their second job.

 

[ewomack]

I've seen a few decent documentaries on Self-Driving cars over the past few years. One was Frontline's "Look Who's Driving" from 2019. It may have aged a few years by now, but they mostly concluded that we're quite a ways off from a fully automated driving world. It seems to work great in small, heavily controlled situations, but a lot of kinks still need working out for weather, obstacles, etc. in the real world. It will likely happen someday, but probably not as quickly as many people initially wanted it to.

 

[Funkworks]

Ok well, I decided I wanted to know how far along Tesla was with their software (called Full Self Driving Beta), so I watched this.



I'm impressed. I understand they update the software every few weeks or months with gradual improvements.

Here's a limitation, called "Chuck's Unprotected Left Turn", showing the kind of problem software engineers want to solve:



So Tesla is my reference (the company more than the man), but I'm curious about other companies doing this and what more of the software, tech and issues can be understood or explained easily (without digging into actual code). Want to know more, have gaps to fill.

I remember many years ago when MIT built a car that went from Mass. to California without a driver.

Curious about this if you find a link.

 

[MClark]

I saw one of these trucks on Interstate 10. Truck had all the spinning sensors.
They have a driver riding shotgun.
https://kvia.com/news/2018/02/02/special-report-self-driving-semi-trucks-tested-on-i-10-in-el-paso/

 

[bjphoenix]

I've watched some documentaries on them. There was a story where a self-driving car collided with a person crossing the road and killed them, there was some question whether the driver riding along was paying attention or occupied with something else.

The gist of the self-driving car is complicated software connected to a lot of sensory input. How does the system "see"? During the day you can use cameras, but they don't work at night. At night you can use radar but it won't work in rain. So that part is complicated.

In the city the system has to track many moving objects, try to determine what they are, and try to predict what they are going to do. Is the person walking towards the road going to stop at the edge of the road? Is the approaching car going to follow its route along the curving road to go past you in the other lane? I think I've heard of situations where a driverless vehicle got stuck at an intersection because it didn't know what to do.

Only time will tell how the systems develop and how successful they are. I'm skeptical of such complicated software systems. I've done programming with my job occasionally and to me it seems that language developers have created languages for some other reason than to be used for real world applications. IOW it takes so much of the programming effort to deal with the complications of the language that there is not enough time to properly develop the software that is being developed. I learned C programming many years ago, then more recently I read a book detailing the full C++ language and to me it was a big step backwards.

 

[jderimig]

This guy is spending a ton of money bashing Tesla self driving software.

 

[cwbullet]

I am not sure how much reality there is in the claims that self-driving software is dangerous. You are still supposed to intervene when it does something unsafe. I am not sure it is meant to operate in high traffic despite the videos showing it do so.

I would only use it on open highways, and I would still feel the need to keep my foot on the brake in case.

 

[Funkworks]

This guy is spending a ton of money bashing Tesla self driving software.

I'd rather stand by the developpers and let Alex here handling his concerns.



Here's where Mercedes is at:





GM CEO Mary Barra commenting:

 

[neil_w]

This guy is spending a ton of money bashing Tesla self driving software.

That guy is suffering from the world's worst case of Musk Derangement Syndrome. He is irrationally obsessed with taking down Musk, and has already done some shady things in his campaign. Oh, and his company sells software that in a certain sense competes with Tesla, sort of.

I do not plan to pay any attention to him. And I say this as someone who is very critical of a lot of things Musk has done, especially lately, including the rollout of and promises around the FSD package.

 

[neil_w]

I am not sure how much reality there is in the claims that self-driving software is dangerous. You are still supposed to intervene when it does something unsafe. I am not sure it is meant to operate in high traffic despite the videos showing it do so.

I would only use it on open highways, and I would still feel the need to keep my foot on the brake in case.

The Tesla software (in beta!) is mean to handle all driving situations, although the operator is supposed to be alert and ready to take over. Highway autopilot has been working well for years now (again, with driver ready to intervene. These are level 2 systems).

 

[Woody's Workshop]

All this tech going into vehicles is just fine for those that want it, want to pay for and can afford it.
But what if you don't want it? Or want to pay for it? Or CAN'T afford it?
I wish I could get a simple pick-up truck, manual transmission, no power anything, cruise, tilt & air is about the only thing I would want.
Rubber floor mat, no fancy interior, something you can hose out when needed.
A basic work or farm truck. I know a few farmers in my area that have restored their old pickups for about 1/3 of the cost of a new one.
The don't want all the fancy interiors because it will just get messed up with cow poop, hay, stray, pig poop, etc.
If an auto maker would come out with an inexpensive work truck like this, I bet they would sell more than they would expect.
I know about 20 people right now that would buy one in a 5 mile radius from me, including me.
$30K and more for a work truck to just get all shitty inside and out is just plain ridiculous, without any alternatives from any manufacturer.

 

[jd2cylman]

Just need them to be better than the average driver, not better than every driver in every circumstance. Here's hoping I can drink my coffee and read the news on the way to work before I retire.

Isn't that called the commuter train?

 

[Funkworks]

... including the rollout of and promises around the FSD package.

I haven't been following closely, but they're releasing FSD beta version 10.69 this week, which is apparently a good step ahead, so I look forward to see what improvements it has (via news, I'm not a user). I think the amount of progress they make is an indication of just how hard the problem is.

There's also an AI day coming up, where they'll be explaining a lot of projects and trying to recruit the best of the best to work on them and boost progress.

Here's the 2021 AI day (summary), for anyone who missed it.

Next one is Aug 20 and I'll probably post it here (unless someone else does).

 

[Back_at_it]

For me this is one of those areas where I have to ask the question. Just because we can, does it mean we should?

Personally, I'm a car guy. I will never buy anything or do anything that takes me out of the drivers seat or takes the feel of driving away from me. For me this means a couple of things.

One, self driving cars, even partially automated will never be used by me. I have driven a Plaid with Auto Pilot and frankly it was entertaining for about 2 mins. After that I turned it off and never gave it another thought in any of the times I've driven that car.

Two, Electric vehicles. Not going to happen. Going back to the Plaid. I've put about 1000 miles on one total. My buddy owns a 2021 and we've taken it to various places. He likes to let others experience the car so I've been behind the wheel a fair bit and I think he is bored with it already.

Is it fast? Oh heck yeah it fast!! Does it do anything for me? Yeah the first few times I hit it I was impressed. After that wore off there was nothing left. The experience was boring. I literally felt like I was driving a microwave down the road as that was the sound it made. It literally hums and buzzes like my microwave does. There's no emotion, no connection.

For non-car people with electric cars, yeah I can see where they might go for self driving as there isn't any enjoyment in driving an electric anyway so why not let someone/something else do it for you. Call me an old school dinosaur, I want sound, gears, exhaust, tire noise and above all else I want to be in the drivers seat.

 

[neil_w]

People become very attached to what they are used to. This pattern is repeated ad nauseum in a million different areas of life. This may slow down the inevitable march of progress, but it can't stop it. My wife was for a while very adamant that we maintain a traditional copper POTS line because it would work in a power outage, even though we both had modern cellphones. That lasted for a while but eventually wore off.

"Sound, gears, and exhaust" are not inevitable components of a motorized vehicle, just artifacts of the technology that happened to be available at the time. And as it turns out, one of them is incredibly destructive to the environment and another is certainly not beneficial.

Fortunately, future generations will not have these same attachments, and will consider them to be quaint (and regrettable) relics of the past.

As for self-driving: it's going to happen eventually, the only question is when. I do not (and have never) shared Elon Musk's optimism in this regard; it is an extraordinarily difficult problem. But they are making progress, and will get there at some point. If not Tesla, then someone else.

 

[Funkworks]

... Just because we can, does it mean we should?

Yeah, because pay-offs include improved safety. I think one thing everyone agrees on is that new cars are generally safer than older ones and we always want cars to be even safer. Safety just keeps getting better overall. There's a lot of safety features between a horse carriage and a (as yet to exist) fully autonomous car. Cameras and warnings and automations can all be seen as safety features.

Other pay-offs are different kinds of economics ramifications (savings). I'd get into that later but I was stunned when I first realized the implications of a (as yet to exist) fully autonomous car. Some of these can sound so strange many would dismiss them as sci-fi, but there is no technical reason to dismiss them (no tech issue that seems impossible). There's one very simple short sentence that blows my mind each time.

Call me an old school dinosaur, I want sound, gears, exhaust, tire noise and above all else I want to be in the drivers seat.

Ok then. ( kidding). Driving is fun but I think it could become a weekend thing for many (like skiing? horse riding? sailing?). Practicality is often the main threshold for mass adoption.



Intro to the tech by a business grad. Something I watched a few days ago.

 

[dhbarr]

Isn't that called the commuter train?

If I worked in a metro area, sure. I don't think they're going to put a lightrail from my suburb to the industrial park

 

[boatgeek]

I am not sure how much reality there is in the claims that self-driving software is dangerous. You are still supposed to intervene when it does something unsafe. I am not sure it is meant to operate in high traffic despite the videos showing it do so.

I would only use it on open highways, and I would still feel the need to keep my foot on the brake in case.

IMHO, the real problem with the current self-driving tech is one of human nature. When the autopilot can handle 99.5% of the driving tasks, it’s hard to stay alert enough to notice the danger and handle the remaining 0.5% in time to prevent an accident.

Granted, the people who do stupid stuff with the car on autopilot are probably the same ones who eat lunch with both hands and steer with their knees so it may not be a significant difference is overall highway safety.

 

[Funkworks]

IMHO, the real problem with the current self-driving tech is one of human nature. When the autopilot can handle 99.5% of the driving tasks, it’s hard to stay alert enough to notice the danger and handle the remaining 0.5% in time to prevent an accident.

Granted, the people who do stupid stuff with the car on autopilot are probably the same ones who eat lunch with both hands and steer with their knees so it may not be a significant difference is overall highway safety.

I know there has been accidents, but Tesla drivers must be approved with a safety record before they're allowed to use the FSD software. This should weed out people who do stupid stuff. This group of ( > 100k ) customers are essentially part of the development team, sending data to the employees while using their car for their daily business.



Intro to the 5 levels of autonomy:

 

[kuririn]

Whenever a friend or co-worker has been involved in a car accident, invariably I hear "it happened so fast, I couldn't react in time". Technology can (and does) supercede human limitations. If the software were to be perfected I would expect a huge drop in the number of car crashes and fatalities should autonomous vehicles become the norm on the roads.
And I posted this on some other thread a long while back but.......
For your amusement:

 

[Peartree]

A few unconnected points that I think of as I read this thread:

1) It wasn't all that long ago that the military was the principal funder of driverless systems and hosted a competition for trucks to drive off-road from point a to point b. At the time, almost none of the competitors even finished. The military applications are still huge. Can you imagine how game changing it would be for resupply missions to not rely upon convoys with combined arms for security? For convoys of military vehicles to be replaced by driverless civilian vehicles that blend in? And for there to be zero risk of loss of life from IEDs or other enemy action?

2) I like driving, but for me the big attraction for a full auto driverless system is when I'm driving cross country for a thousand miles. At some point, it would be nice to just flip a switch and take a nap instead of paying for a hotel.

3) One of the long-term tech issues that has been considered (but not discussed much here) is that once a significant number of automated systems (not even full auto but even those that assist drivers) are on the road, there will come a day when those systems integrate and communicate with one another so that they absolutely know what the oncoming car is going to do and not merely guess. Once that happens, they take a quantum leap forward in safety.

 

[dhbarr]

Whenever a friend or co-worker has been involved in a car accident, invariably I hear "it happened so fast, I couldn't react in time". Technology can (and does) supercede human limitations. If the software were to be perfected I would expect a huge drop in the number of car crashes and fatalities should autonomous vehicles become the norm on the roads.

It's inevitable that at some point the actuarial tables will charge manual drivers extra for the privilege.

 

[Funkworks]

A few unconnected points that I think of as I read this thread:

3) One of the long-term tech issues that has been considered (but not discussed much here) is that once a significant number of automated systems (not even full auto but even those that assist drivers) are on the road, there will come a day when those systems integrate and communicate with one another so that they absolutely know what the oncoming car is going to do and not merely guess. Once that happens, they take a quantum leap forward in safety.

That's in the ball park of what I think is the most amazing prospect for level 5 autonomy, and I like to summarize it in one short and simple sentence:

A self-driving car never has to park.

Except for charging of course. But from that statement alone sprouts more possibilities or scenarios than I can imagine. Besides safety and comfort, that's the main driver: a car, or a fleet of cars, that can be used 90% of the time, instead of 5% or so.



On the GM side:

 

[boatgeek]

I didn't really make it clear above, but I'm more or less OK with 100% self driving even though I'm not comfortable enough with the technology to want to own one. I'm also pretty cool with 85% self-driving (lane changes, lane keeping, automatic braking, etc.). It's the 99.5% that really wigs me out. This is also less about the actual technology, but I think that the name "Full self-driving" does a disservice to a system that doesn't do 100% of the driving. And none of this is to say that the research on self-driving shouldn't be done--it's a me thing. Given how hard a time I have paying attention and staying on task in a Zoom meeting, I don't trust myself to be ready for the 0.5% situation if I'm not having to do the rest of the driving too.

I know there has been accidents, but Tesla drivers must be approved with a safety record before they're allowed to use the FSD software.

I didn't know that. It makes me feel slightly better about the technology, though I also wonder what happens when they release to a broader audience.

 

[cls]

The military applications are still huge. Can you imagine how game changing it would be for resupply missions to not rely upon convoys with combined arms for security?

It's called autonomous multi copters.

Can you imagine thinking beyond military applications, to the larger bulk of civilian applications, such as delivery of important things in places with poor or no roads?

 

[cls]

IMHO, the real problem with the current self-driving tech is one of human nature. When the autopilot can handle 99.5% of the driving tasks, it’s hard to stay alert enough to notice the danger and handle the remaining 0.5% in time to prevent an accident.

Exactly!

Intro to the 5 levels of autonomy:

Thanks!

Gives me hope to see folks here talking about the SAE levels of automation, because it lends nuance and detail to an otherwise black and white rat hole discussion.