Tesla's new radar tech predicts an accident caught on dashcam a second later | Hacker News Reader

Tesla's new radar tech predicts an accident caught on dashcam a second later

electrek.co

79 pointsobi1kenobi26 comments

There are going to be haters that come on here and try to down play this video..

But there is something to be said about a car that can gauge the movement of others around it, and a car company that is willing to push for a better type of vehicle. The largest automakers in the world have had the power, finances, and means to progress vehicles for quite some time and have not. Tesla's are not perfect, but it terms of what they are doing I give them a big hats off.

and means to progress vehicles for quite some time and have not.

As a "car guy" for 30+ years, could have fooled me that there hasn't been progress. In fact, there's a 1995 and 2015 Toyota Corolla in my driveway, and the cars are not remotely the same.

EVs are a paradigm shift, but aren't the only form of technological progress.

How are they different?

The major automakers lobbied politicians to prevent mandated front seat collapse prevention that would have cost them $1 per vehicle and saved many lives. They're scum and deserve their downfall - and please next time let's not change bankruptcy laws temporarily to bail them out!

I don't mean to downplay the video, but isn't "predicts an accident" stretching things a bit? It noticed that the vehicles in front were slowing down and applied brakes. That's what it warned for, AFAIK there is no distinct warning signal when an accident is about to happen.

It warned about a forward collision with a car that the driver couldn't even see.

Watching closely, the red car directly in front does not appear to slow down measurably, and certainly, their brake lights do not light up. You do, however, see the black SUV's brake lights light up.

Prediction is a strong word, to be sure. It's great when it's right, and annoying when it's wrong -- like car alarms -- but if it it fires some portion of the time when there is an actual accident imminent -- like a smoke alarm -- then it's clearly useful.

Cars should auto-brake.

Model S has forward collision auto braking. It will be required in most vehicles made after 2020.

> but isn't "predicts an accident" stretching things a bit?

Agree. N=1 here. For all we know this was just a coincidence or what you are giving as a reason.

The largest automakers in the world have had the power, finances, and means to progress vehicles for quite some time and have not

No offense intended, but you haven't been paying attention if you think that's the case. For instance, I've read on this very forum how Tesla's driving-assistance technology is a bit dated (camera and no LIDAR, for instance) in comparison to Daimler, et. al.

As for other aspects of progress, I don't know what you're looking for. Within my lifetime alone we've gone from vehicles that need constant maintenance (ignition points and 5K mile plugs, anyone?) to appliances-on-wheels. My Mom's '16 Corvette puts out 450bhp, still gets 25mpg, and I'll bet you could put your nose straight up the pipe and die of boredom before the exhaust fumes get you. And, unlike something of equivalent horsepower from the 70's, you can drive it in town.

Large brands made hybrids, EVs like the Leaf that made an EV "everyman's car" before the Model S and loooong before the Model 3.

So to play it like Tesla's the only one with the guts to do what needs to be done just means their marketing is working.

Signed,

Owner of Many Shares of TSLA

I'm a huge Tesla fan and am very glad they're pushing things forward... buuuuuut, Audi / VW / Mercedes / Volvo have had adaptive cruise control with auto-braking for over a decade now. The first Audis with Brake Guard were released in 2002. By 2009 you could get ACC on an Audi that cost less than $35k.

The major autos that "had the means to progress vehicles for quite some time and haven't" have literally sold millions of cars with this technology.

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

Can get a Mondeo in AU for $12k with it.

Tesla's one turns, works in peak traffic and can change lanes though.

In my opinion, Tesla's decision to not use LIDAR shows their system is more advanced than the competition's, rather than it being dated. I have two reasons:

- LIDAR is extremely expensive! The LIDAR unit that Google uses costs $80,000 by itself [0], whereas the entire hardware package that Tesla uses is $8,000 [1].

- Path planning, in my personal experience, is significantly easier given millimeter-level accurate readings on millions of points every second, than with 60fps video, even from multiple cameras.

The way I see it, it appears that Tesla has been able to achieve comparable (or better) progress on self-driving cars with significantly lower-cost hardware. The name of the game is making self-driving cars the default option, and that can't happen when the self-driving hardware by itself costs as much as a luxury car.

[0] http://articles.sae.org/13899/

[1] http://www.theverge.com/2016/10/20/13346512/tesla-self-drivi...

Those Google lidar numbers are not reflective of anything that would ship. Much of what Google is doing is capturing all of the ambient data so each subsequent car knows more about the environment than the one that preceded it. You can buy simple solid state lidar units for under $1,000 and there are several in the pipeline targeting $100/unit. It's much more likely to see those on any mass market product.

Delphi is probably the largest auto supplier in the world and they recently invested a significant amount of money into a silicon valley company producing a solid state lidar who are targeting a $250 price point:

https://www.google.com/amp/www.forbes.com/sites/lianeyvkoff/...

If lidar makes self driving safer, there is no reason at all not to use it.

It's worth noting that in the aftermath of the death of a Tesla driver whose autopilot system couldn't see a white semi trailer, Tesla's sensor supplier reiterated the need for lidar and redundant sensors:

https://www.google.com/amp/amp.timeinc.net/fortune/2016/07/1...

I wouldn't hate on it, but my Mercedes equipped with Distronic Plus can do the exact same thing, and I'm sure it was at least an option on the S-class for a decade now. You can even get it on the base barebones A class nowadays. All I'm saying is that there's nothing new here and the number of articles this video caused is almost silly.

This is great! I feel like there is going to be a strong opposition to autonomous vehicles as they become more mainstream and widely available. We need more videos and examples like this that show the benefits. I understand auto pilot isn't fully autonomous but these are good building blocks.

I'm really going to cringe when people want to hold on to their super charged car/truck and drive it recklessly in the name of freedom. We can't publicize enough the current dangers of driving.

I'm very much against the idea that only autonomous vehicles should be allowed on the road, but I'm 100% behind technologies like the one demonstrated in the video,which is essentially auto breaking with radar assist. I think this technology will be required by law within a few years for all new cars in the US actually.

I don't think it "predicted" the accident, it just saw the car two ahead braking very strongly. (And the car one ahead didn't see that. ;)

I think this is a much more appropriate use of the technology present in Tesla cars than allowing it to self drive. I'm firmly in the "shouldn't have worked that way" camp of Autopilot -- no matter what Tesla or Musk says, the first thing I would think of as someone who is not a trained pilot is that Autopilot is a self driving feature of a car, and believe that the wording and marketing, combined with the reckless use of simple technology like cameras, instead of stronger technology like the radars here, caused crashes. I'm glad that even though they have stronger tech they are still toning down the marketing and the use of any self driving features, so kudos to Tesla (it might have been due to bad PR but still).

However, I have to wonder why this is news. Other cars have had similar technology for a long time but when they prevent a crash it's not news. Is the only reason this is news is because it's Tesla and that they've had high profile crashes in the past, so they're spinning up the PR machine?

any guesses as to how the car in front of it could have prevented the accident? what was the human error involved? was it simply a case of being too close to the car in front of it?

Leave adequate following room - at least twice what you think you need, stop fucking with your phone while behind the wheel.

Great, now I'm going to have to keep doubling my following distance forever.

You certainly won't run into the car in front of you until you circle the earth, but then it's the other guys fault!

I'm guessing the LIDAR saw the second car up rapidly slowing. You could catch a peek of it over the smaller car.

Teslas aren't equipped with LIDAR. I think they're the only self driving capable vehicle not to use it.

It uses the radar for detection by bouncing it under the car in front.

Reply on news.ycombinator.com