Hacker News Reader

Comment by umvi | Hacker News Reader

Seems like electromagnets could help compensate for distance by drawing more current when far apart, and less when close together. A microcontroller could essentially make the force linear like a spring. Of course, then you need wires hooked up to your pogo, and that might be scary if the microcontroller has a bug (or somehow fails) since you are now essentially hopping onto a railgun.

View on HN

About as scary as my Tesla on autopilot going around a curve at 75mph. Every time I ask myself, what if there is a bug or if it turns off...

I know exactly what you mean. I will admit at 20mph over the speed limit going over a poorly maintained curve on the Mass Pike I usually just disengage for a moment and then switch it back on. It can take the curve, but it doesn’t feel good to let it.

I’ve had it lose tracking on the highway. It just leaves the wheel in the last position and beeps very urgently to let you know to take over.

I find autopilot helps me become more aware and less fatigued while driving. It’s a level 2 system. There is never a question of losing engagement as an active driver like what happens with level 3 self driving.

It will keep a clearly marked lane better than I can (it’s very precise with staying in the exact middle of the lane and doesn’t cut corner like humans tend to). For stop and go traffic it’s a blessing.

It sucks at zipper merging. I’ve had to brake for merging cars more than once, even though they are visible coming into my lane on the HID. That part needs work. And I don’t really trust it to change lanes as I feel like the blind spot detection isn’t quite there either.

All said it’s pretty much the most amazing driving tech ever, and the TM3 is most definitely the most amazing driving machine I’ve ever had the pleasure of owning. Just orders of magnitude better handling than I realized was possible. Curves rated at 25mph that I used to take at 45mph in my Infiniti previously with some tire squeal, it can silently tear through at 65mph. I don’t know what the skidpad rating is, but it feels unflappable and body roll is just nonexistent.

Actually it’s a bit tricky because you literally feel no complaints or judder or roll all the way up to pulling 100% of traction the tires have to offer. It is quite well mannered once it does start drifting though.

Sorry, sorry, I can’t help it. You can guess I’m thrilling at parties.

I rave about how awesome it is too. In stop and go traffic it’s wonderful and rarely asks me to touch the wheel. I don’t like it for lane changing either so I always take over.

> It will keep a clearly marked lane better than I can (it’s very precise with staying in the exact middle of the lane and doesn’t cut corner like humans tend to).

Anyone have experience with the lower Westchester County section of the Taconic Parkway? There are some serious curves, and the lanes are very narrow. In the left lane, you're maybe 2-3 feet from the box-beam center divider.

I just saw someone reverse their TM3 into their garage with their phone. Pretty neat! Definitely will be my next car to make my long commute bearable via autopilot tech.

The ratings on corners (at least in Australia) are rated based on visibility, not some complicated traction calculation.

They factor how far a head you can see, and how much time you need to react. Lots of cars/bikes can obviously tear through them faster. But that means they are just taking much more risk.

That said, perhaps auto driving can respond quicker, but I’d imagine that doesn’t change most of those ratings give. The time needed to avoid accidents is largely due to physics not reaction speed.

Pretty much every critical system in a vanilla Toyota is already software-controlled, not to mention the legion of critical systems involved in air travel, elevators, medical equipment (e.g., pacemakers), etc. Not sure that this should bring comfort, but at least you need not be particularly frightened by your Tesla. :)

Elevators use simple physical safety systems.

Which is why you see headlines like: "Skyscraper elevator plunges 84 floors — and everyone survives" https://www.cbc.ca/news/world/elevator-plunge-chicago-1.4913...

Every link i’ve seen to this story annoys the hell out of me because they don’t explain HOW the elevator eventually stopped. That’s literally all I want to know.

It reached the bottom. Probably hit one of these: http://elevation.wikia.com/wiki/Buffers

Note that it wasn't descending particularly quickly; one cable snapped but other safety systems were still operable.

The article says it went from the 95th to the 11th floor. It appears it stopped before reaching the bottom buffer.

The best book ever for understanding how Elevators work is a wonderful late 20th century work of fiction called The Intuitionist by Colson Whitehead, which was described by its author as what you would expect to happen if you let a child grow up reading elevator manuals.

Entering “elevator safety mechanism” into google will bring up several links with explainations. The first link for me references this story.

There was an elevator hacking talk that went into the buffers at the bottom of the elevator. You may have seen them in elevators that have glass on the outside of the shaft like in malls and hotels. The elevator never rests on those. Those are to take the impact of an elevator that crashes. Inspectors will often test the buffers as well as part of routine maintenance:

https://www.youtube.com/watch?v=ZUvGfuLlZus

The basic theory is that between the car and the wire there is a spring connected to brakes. When the car hangs in the wire this spring is loaded and the brakes disengage, if the wire breaks, the spring returns and brakes engage. All this is mechanical.

What I never understood is how elevators can travel downwards so quickly without such a system engaging. Can they block it somehow or are they never accelerating fast enough down?

I’d also like to know what they meant by “plunges”. How fast did it descend? Was it in free fall?

You can easily tell that it wasn't in free fall, because if it fell 84 floors in free fall and then stopped, the passengers would have died.

It could have slowed down gradually and stopped. The articles I’ve read don’t do a good job describing anything.

If it slows down gradually, it's not in free fall.

Guess again. The cable could break at floor 84, and the elevator car could travel through the next 50 floors, and not slow it’s descent until the 34th floor, at which point, given ten vertical feet per floor, it would still have a total stopping distance of nearly 200 feet, or 2/3 (66%) of a football field.

So, 500 feet of freefall is still substantial. That puts the situation on par with a bungee jump.

What's deadly is the other type of failure, where the counter weight causes the elevator to shoot upwards, like what happened in that elevator in Chile. There's a video of the guy confused as the doors continued to open/close at each floor. He had severe injuries.

Apparently if that ever happens, your best bet is to lay flat on the floor with your hands covering your head, as the impact will launch you upward.

Thats how I usually approach stressful situations, so I should be fine.

Aren't brakes and steering mostly manual? I don't think it's the right advice to tell people to trust Autopilot when it's failed in several egregious ways.

They are manual in the sense that there is a physical connection: the steering wheel is usually mechanically coupled with the steering rack, brakes usually through through hydraulics.

That said there are probably 5-20 systems in a car that can completely control these systems without any human input. Think of automatic emergency braking, traction control, vehicle stability control for the brakes, and autopark + powered steering for the steering.

Most of these systems have multiple redundancy and isolated monitors (eg, if autopark is on and the vehicle speed is > 10mph, the monitor will intervene) but there are still probably some theoretical situations where this could occur

Steering is electronic.

Electronically assisted, but still mechanically coupled for 99% of consumer cars. I've completely lost 12V on a car before and still managed to bring it safely to the side of the road.

Inifinity tried steer-by-wire but ended up retrofitting traditional hydraulic steering back.[1]

[1] https://en.wikipedia.org/wiki/Drive_by_wire

sterring is electric, but there's always a backup system in case the electronics fail: https://youtu.be/em1O8mz7sF0?t=276

Driving a car on the road with other cars is - factoring in the age of the tech, lack of regulation, involvement of humans and the consequences if it goes wrong - more difficult that what's mentioned on your list.

there's a huge complexity difference, however.

I have used Tesla autopilot on twisty cliffside roads (my hands hovering an inch away). The question in my mind was, when there is "nothing" there, is that a place it thinks is okay to go? Or is it a place a place where it should absolutely not go? I wish it could just tell me that.

It doesn't know about dropoffs. It only knows about lane lines. It's really good at following lane lines. Sometimes right into an obstacle.

It knows about the driveable areas even outside the lane lines where it can go in the case of an emergency.

See: https://www.youtube.com/watch?v=il6LV4064bI

>what if there is a bug or if it turns off

Right? I have an irrational fear of laser eye surgery because I'm afraid some random component in the machine will fail and cause some sort of over-powered pulse.

It's far more realistic a component would fail in a car though and cause the 'brain' to take a shit on itself. Still pretty out there but it would be a very real concern of mine.

You let it go round a curve at 75mph, or is this hypothetical?

I think he's referring to a curved piece of highway, not a turn.

The downside risk is never worth it for me. I'm always hands on, and curves I treat it like a drivers' ed situation: You better do the right thing.

Yeah, I tell people that I treat it like a well-mannered 15-year-old driver - I expect it'll do the right thing, but I'm here to take over in case it finds itself out of its depth.

Consider not going on a modern airliner. Every commercial passenger jet since the A340 has been fly-by-wire.

(military jets have been so longer).

But you have more time to react on a plane and switch to manual override. It's not like people fly with just a few meters distance to other airplanes. There's also a good distance to the only other thing you can crash into, the ground, as opposed to driving between walls or with a cliff to one side.

OP said "fly by wire", meaning there's no non electronic systems to fall back on. If the electronics fail there's no "manual override".

Of course there's a fallback. If the fly-by-wire system fails, there's another whole fly-by-wire system ready to take over. Usually this is safer than the old redundant hydraulics systems where all the lines ran close together, making them vulnerable to localized impact damage and fire.

I specifically said "non-electronic" fallback.

Reliably implementing a few control loops and exhaustively testing all possible input/outputs for bugs - difficult but manageable.

Reliably implementing image recognition from multiple cameras and AI that should respond to ever changing weather, inconsistent roads, signs, maps and other human drivers erratic behaviour within milliseconds - Not even in the same league.

Not saying aircrafts are simple but self driving is just orders of magnitude more complex. Otherwise we would have seen them on the roads already.

Aircrafts also have much stricter maintenance schedules, failure investigations and completely different economic, media and moral incentives.

Aircraft autopilots and fly-by-wire systems are vastly more simple than self-driving cars.

Depending on what fails how in a car, you may have less than a second to recover.

Apart from being hit with a high-explosive anti-air missile, I can't think of any failure mode in a plane that has such a narrow recovery window.

737 Looks like cables, air, and hydraulics to me.

737 is a much older aircraft. The A320 and subsequent Airbus jets and Boeing jets since 777 have been electronic.

The Embraer ERJ series does not have fly-by-wire, just to name one. So the assertion that every "Every commercial passenger jet since the A340 has been fly-by-wire" is simply false.

... So don't use it?

Not only that, but you can adjust the “springiness” by having a variable spring constant.

Using css animation-timing-function!

You can even make it jump by itself.

Maybe you can use regenerative braking to charge a battery, to avoid having wires and keep a bit closer to the spirit of the pogo stick.

Alternatively a lever system or non-circular gear arrangement could convert the force from fixed magnets into a linear response entirely mechanically.

...edited for sense

> you are now essentially hopping onto a railgun.

No, you're riding a coil gun. But your point stands.

https://en.wikipedia.org/wiki/Coilgun

You could have controls in the handgrips to change the shape and amplitude of the curve.

As far as bugs go, you could clamp the output in the good old fashioned analog way (i.e. via circuitry) so the worst case is you just fall/jump off which is already the case for the sprung ones.

Honestly, someone might want to hop on a railgun. Extreme pogo

Could you not hardcode limits in your circuit, such that the maximum power it supplies is never enough to be deadly?

Or hardwire? A fuse, just not having enough voltage, etc?

This is the best comment I have ever read

You don't have to think about buggy software to come up with death scenarios. What if a driver in the other lane is drunk or not paying attention? Another driver could kill you at any moment and there's nothing you could do to prevent it.

The dangers of operating a pogo stick on vehicle congested roadways are significant, but remain consistant across differing pogo technologies. This post makes no mention of making a pogo stick safer to operate on public roadways.

What are you talking about?

Reply on news.ycombinator.com