Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The long, winding road for driverless cars

economist.com

Post reply on HN
Print view View original

11–20 of 61 posts

Re: The long, winding road for driverless cars

#11

Post by XorNot »

wavefunction wrote: post #9
jacquesm wrote: post #6

Earlier quoted context omitted.

Drivers deal with having the sun shining in their eye using sunshades, visors and a whole raft of movement tricks to avoid having their retinas burned out. A passive sensor (which is where we're headed) has no protection at all, it just sits where while it is being fried. It's true that the sun low near the horizon at sunup or sundown, especially on clear winter days causes accidents but I'm not all that sure whether…

What if you put an intensity sensor next to the LIDAR sensor and came up with some solution to position the LIDAR like sunshades, visors, and 'movement tricks.'

HDR photography can be applied practically to let you look directly into the sun and still see objects next to it perfectly clearly.

There's no actual problem for a video camera to drive while staring straight into the sun, beyond the need to have a camera shutter.

Re: The long, winding road for driverless cars

#12

Post by hwillis »

ams6110 wrote: post #5
daveguy wrote: post #3

Earlier quoted context omitted.

> There are about 500,000 people who have placed orders for the Model 3 sight unseen. * There are 500,000 people who have put down a fully refundable $1,000 deposit on the option to buy a Tesla Model 3. That's a long way from 500,000 orders placed. (But I agree, "less than 2% of all new car sales in most countries" is a weasely stat, because it is growing fast. They are still a small percent of total sales)

It's because EVs only work for a subset of driving needs (local commutes). If you need to drive long distances, even occasionally, you either need to own another car, or rent one. Either is at best suboptimal compared to owning one car that can do it all. When I can own an EV that will go 300-400 miles per charge and can be recharged in 5 minutes at any corner "station" I will think about it.

It works for any "local commutes" up to 250 miles (base model). That's a shitload of driving. Taking a half hour break after over 3 hours of driving is not that weird.

Imagine if we all had electric cars, and had to switch to gasoline. It would seem so inconvenient to have to go all the way to a gas station to refill, when you're used to having a fully charged vehicle every morning. It would seem so inconsequential that you could make 400-500 mile journeys a whole half hour faster. Why would you not want to stop for lunch while on a 6-7 hour trip?

Re: The long, winding road for driverless cars

#13

Post by moxious »

When I hear "Most carmakers have plans to start testing the market with Level 3 or possibly Level 4 autonomous vehicles around 2021"...

It makes me think of what a stupendous amount of time that is in engineering terms. And then I think that basically a suit somewhere has a product roadmap, but no real idea of how he's going to get there.

Which in turn tends to make the time estimate questionable at best, wishful thinking at worst.

Re: The long, winding road for driverless cars

#14

Post by wcummings »

daveguy wrote: post #10
wcummings wrote: post #7

Earlier quoted context omitted.

I thought the idea was that you could swap batteries, not sit there charging.

As far as I know no one has implemented that yet. Tesla has superchargers around the country, but those are charge based not exchange based. They give +170 miles for a 30 minute charge, so not exactly a quick stop. The size and weight of the full battery pack may make it infeasible to change quickly. A half pack or 1/4 pack may be useful, but then the car has to be standardized/designed for that and you have to worry…

Agreed, I can see how that would be a hassle.

That said, I cant even remember the last time i traveled more than 85mi (half the capacity, so depleting thr battery in a round trip) in a car, and I dont think im that exceptional, so theres probably still a pretty big market for these things.

But personally, Ill stick to the bus. I do all my best thinking on the bus.

Re: The long, winding road for driverless cars

#16

Post by sliken »

Seems like a pretty biased article which seems to largely ignore quickly increasing sensor ability (cameras, pairs of cameras, radar, and lidar). They also ignore quickly increasing amount of CPU power for vision like tasks (parallel) like the Nvidia TX2. No reason that shouldn't keep improving over the next few years. The article also repeats the popular refrain that lidar is the only reasonable solution. Funny how human eyes (just 2) aren't bad. I've also read reports that lidar is particularly susceptible to things like fog and dust.

I can't see why a dozen or so cameras can't do quite well compared to a lessor number of lidar units. Especially if you use pairs of cameras which makes it much easier to pick out 3D features. Human eyes are only separated by a few inches, no reason pairs of cameras couldn't be separated by 6 feet if depth perception is a limiting factor.

Machine learning has been increasing quite quickly (things like recognizing human speech, identifying images, computer vision, tracking objects, etc).

Current trajectories seem to imply pretty large improvements in the next 3-4 years. Things are somewhat delayed because Tesla just switched from pretty poor sensors to processors to a significantly improved hardware last year. The software is still in the process of catching up.

I expect quite nice autonmous driving solutions will be available in the discussed 3-4 year time frame and the roads will be significantly safer because of them.

The article also seems to exaggerate the difficulty in testing. Tesla (a fairly small car company) already has quite a few miles under it's belt, even if you ignore all the miles on the previous generation hardware. The previous reported number of dropping airbag deployments by 40% is a huge step in the right direction. With even a pessimistic number of model 3's shipping there will be ample chance to show if tesla's system is better than the average human driver.

I expect pretty much all the large car manufacturers to update their current systems that provide predictive braking to include similar capabilities and catch up and surpass Tesla's 100's of millions of miles driven under computer supervision.

Re: The long, winding road for driverless cars

#17

Post by frgtpsswrdlame »

sliken wrote: post #16

Seems like a pretty biased article which seems to largely ignore quickly increasing sensor ability (cameras, pairs of cameras, radar, and lidar). They also ignore quickly increasing amount of CPU power for vision like tasks (parallel) like the Nvidia TX2. No reason that shouldn't keep improving over the next few years. The article also repeats the popular refrain that lidar is the only reasonable solution. Funny how…

If this article is biased because it shows some pessimism towards self-driving hype, what are the dozens of articles that blindly accept Musks near-term prophecy?

Here's another good source for some pessimism:

http://www.nakedcapitalism.com/2016/10/self-driving-cars-how...

Here's the best parts:

Contrary to Musk and many of the most prominent advocates of autonomous cars, Shladover insists that so-called Level 5 vehicles—robocars that require no human input—are not on the horizon. “I tell adult audiences not to expect it in their lifetimes. And I say the same thing to students,” he says. “Merely dealing with lighting conditions, weather conditions, and traffic conditions is immensely complicated. The software requirements are extremely daunting. Nobody even has the ability to verify and validate the software. I estimate that the challenge of fully automated cars is 10 orders of magnitude more complicated than [fully automated] commercial aviation.”

Herman Herman, director of the Carnegie-Mellon University Robotics Institute, disagrees as well: "With autonomous cars, you see these videos from Google and Uber showing a car driving around, but people have not taken it past 80 percent. It’s one of those problems where it’s easy to get to the first 80 percent, but it’s incredibly difficult to solve the last 20 percent. If you have a good GPS, nicely marked roads like in California, and nice weather without snow or rain, it’s actually not that hard. But guess what? To solve the real problem, for you or me to buy a car that can drive autonomously from point A to point B—it’s not even close. There are fundamental problems that need to be solved."

Re: The long, winding road for driverless cars

#18

Post by kbos87 »

I for one hope driverless cars take a hell of a lot longer than people like Elon Musk expect them to.

First off, I enjoy driving. I identify with the brand of car I just bought. Yes, I have to deal with traffic and competing for parking in my congested city neighborhood... but that's a choice I made because driving and the freedom that comes along with it are, to me, things I value. If you think I'm a small minority, look at all of the car owners in my neighborhood who still choose to own cars despite an Uber being available in 120 seconds, and affordable public transit being a 3 minute walk away. I use both of these services, and I still choose to own a car.

Second - who is to say that autonomous vehicles are the cure for all of our ails? Yes, we'll eliminate some traffic fatalities. We'll also eliminate one of the biggest sources of employment in our society, which I worry will set an entire chunk of the population down a negative path toward bad outcomes like depression, suicide, and drug addiction when they can't sustain the meager existences they already live.

Lastly, I don't like the idea of some self aggrandizing tech entrepreneurs in SV who have decided that they know what the future of the place where I live should look like.

Re: The long, winding road for driverless cars

#19

Post by sowbug »

This section is completely false:

"Tesla continues to include the Autopilot sensors and software in its cars, but has deactivated the system while further testing is undertaken. The company plans to re-activate it in 2019 or thereafter."

The author might have been confused about the reason for the switch from the Mobileye-based v1 hardware of Tesla Autopilot to in-house v2, which indeed represented a regression in functionality. It is true that the death of the driver was a catalyst in the breakdown of the relationship between Tesla and Mobileye. But there were plenty of other reasons, most related to hardball business tactics. It's false to say that Tesla removed existing Autopilot functionality.

Moreover, even v2 of Autopilot is currently approaching v1's level of functionality. Autosteer works and TACC works. Those are the main parts. I think only minor features like automatic windshield wipers are still missing. (I have only v1 so I don't know first-hand.)

Incidentally, it's rare for The Economist to get facts so wrong. Their articles are usually very well-researched.

Re: The long, winding road for driverless cars

#20

Post by Fricken »

kbos87 wrote: post #18

I for one hope driverless cars take a hell of a lot longer than people like Elon Musk expect them to. First off, I enjoy driving. I identify with the brand of car I just bought. Yes, I have to deal with traffic and competing for parking in my congested city neighborhood... but that's a choice I made because driving and the freedom that comes along with it are, to me, things I value. If you think I'm a small minority,…

>I don't like the idea of some self aggrandizing tech entrepreneurs in SV who have decided that they know what the future of the place where I live should look like.

Does Henry Ford count as a self-aggrandizing tech entrepreneur deciding what the future of the places where people live should look like?

Post reply on HN