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

Tesla Software Update 8.0

tesla.com

Post reply on HN
Print view View original

31–40 of 97 posts

Re: Tesla Software Update 8.0

#31

Post by toomuchtodo »

M_Grey wrote: post #30
speedplane wrote: post #8

So the biggest news here seems to be the upgrades to Autopilot. ( https://www.tesla.com/blog/upgrading-autopilot-seeing-world-... ). They describe a fantastic amount of new technology, including using radar to make a point cloud despite the fact that "[s]omething made of wood or painted plastic ... is almost as transparent as glass to radar." This venture reeks Murphy's Law. I'll let someone else be the guinea pig in…

Nice thought, but they're on the road with everyone else, so it's kind of not our choice about who gets to be subject to this experiment.

New autopilot updates are only pushed out after they've been run against simulated data (in house validation, "the gauntlet" corpus of scenarios) and then real world data (inertly) and perform significantly better than existing algorithms.

https://youtu.be/2vHO1p9717s?t=8m45s

Re: Tesla Software Update 8.0

#33

Post by colejohnson66 »

LeifCarrotson wrote: post #18
rorosaurus wrote: post #9

Earlier quoted context omitted.

"Tesla keeps temp below 40C/105F by automatically venting cabin & turning on AC when needed. Can do this for up to a year on a full charge." https://twitter.com/elonmusk/status/778602694399320065

Well, sure, if it's 72 and partly cloudy out. My car can do that for years by leaving the windows open. This feature only really matters when it's swelteringly hot out. What if I park a black Tesla at the San Diego airport for a week-long vacation, and it's 100 degrees and unmitigated direct sun? Will I come back to a battery that can't get me home? That would be when I would want this feature to avoid damage to my c…

If it can handle a year on a full charge, then unless your battery is below 2%, it'll last a week. And if it was below 2%, you probably wouldn't be able to make it home anyways.

Re: Tesla Software Update 8.0

#35

Post by ghshephard »

toomuchtodo wrote: post #31
M_Grey wrote: post #30

Earlier quoted context omitted.

Nice thought, but they're on the road with everyone else, so it's kind of not our choice about who gets to be subject to this experiment.

New autopilot updates are only pushed out after they've been run against simulated data (in house validation, "the gauntlet" corpus of scenarios) and then real world data (inertly) and perform significantly better than existing algorithms. https://youtu.be/2vHO1p9717s?t=8m45s

Presumably there are a ton of beta testers that put a 100,000 or so miles running this actively in a wide array of road driving scenarios before they roll this out to customers as well. Under beta testing scenarios, a single driver should be, even with time for charging, be able to do about 250 miles/day of driving - figure 10,000 miles/month. So, to get 100K of real-world road experience, would only take 10 drivers/month (or 20 drivers 2 weeks, etc.... Mythical Man month doesn't apply here)

Re: Tesla Software Update 8.0

#36

Post by eth0up »

eth0up wrote: post #13

I wonder if hackers will have to be 6 instead of 12 miles away to take control of it now.

I've added a link in case the above comment makes no sense: https://www.theguardian.com/technology/2016/sep/20/tesla-mod...

The vulnerability has been "patched", but I cannot help but be peeved at any private automobile that can be remotely controlled. Of course, it isn't only Teslas: https://www.wired.com/2016/03/fbi-warns-car-hacking-real-ris...

Re: Tesla Software Update 8.0

#37

Post by toomuchtodo »

ghshephard wrote: post #35
toomuchtodo wrote: post #31

Earlier quoted context omitted.

New autopilot updates are only pushed out after they've been run against simulated data (in house validation, "the gauntlet" corpus of scenarios) and then real world data (inertly) and perform significantly better than existing algorithms. https://youtu.be/2vHO1p9717s?t=8m45s

Presumably there are a ton of beta testers that put a 100,000 or so miles running this actively in a wide array of road driving scenarios before they roll this out to customers as well. Under beta testing scenarios, a single driver should be, even with time for charging, be able to do about 250 miles/day of driving - figure 10,000 miles/month. So, to get 100K of real-world road experience, would only take 10 drivers/…

Tesla is accumulating ~1 million miles of data every 10 hours fleet-wide [1]. With their environment, they could presumably beta test across the entire fleet, as the changes are inert and only data is sent back as to what the vehicle would've done (compared to what it actually did).

[1] http://qz.com/694520/tesla-has-780-million-miles-of-driving-...

Re: Tesla Software Update 8.0

#38

Post by aidenn0 »

koenigdavidmj wrote: post #3

So a touchscreen meant you lose tactile feedback. We knew that already. But now all the buttons move around too?

On my wife's car the play/pause and toggle-repeat "buttons" are both on the touch screen right next to each other and are maybe 1cm in size. It is easily the worst interface I've ever used in a car for anything.

Re: Tesla Software Update 8.0

#39

Post by sliverstorm »

Animats wrote: post #19
speedplane wrote: post #8

So the biggest news here seems to be the upgrades to Autopilot. ( https://www.tesla.com/blog/upgrading-autopilot-seeing-world-... ). They describe a fantastic amount of new technology, including using radar to make a point cloud despite the fact that "[s]omething made of wood or painted plastic ... is almost as transparent as glass to radar." This venture reeks Murphy's Law. I'll let someone else be the guinea pig in…

"It is hard to tell from a single frame whether an object is moving or stationary or to distinguish spurious reflections." What, they're not getting range rate directly from the radar? You can tell if an object is moving towards or away from you from one frame if you have range rate. Bosch automotive radars do return range rate.[1] Even 1990s automotive radars returned range rate. That's one of the huge advantages of…

My recollection is Tesla has been leaning heavily on optical image processing. Perhaps the radar is, or was, not very sophisticated? For example, can a single unit by itself tell, in one frame, if an object is moving laterally to its field of view?

The radar was added to all Tesla vehicles in October 2014 as part of the Autopilot hardware suite, but was only meant to be a supplementary sensor to the primary camera and image processing system.

Re: Tesla Software Update 8.0

#40

Post by tempestn »

colejohnson66 wrote: post #33
LeifCarrotson wrote: post #18

Earlier quoted context omitted.

Well, sure, if it's 72 and partly cloudy out. My car can do that for years by leaving the windows open. This feature only really matters when it's swelteringly hot out. What if I park a black Tesla at the San Diego airport for a week-long vacation, and it's 100 degrees and unmitigated direct sun? Will I come back to a battery that can't get me home? That would be when I would want this feature to avoid damage to my c…

If it can handle a year on a full charge, then unless your battery is below 2%, it'll last a week. And if it was below 2%, you probably wouldn't be able to make it home anyways.

The parent's point is that the circumstances would dictate how quickly the battery drains, so it wouldn't always last a year on full charge. (And therefore not always a week with > 2% charge.) That said, it's not trying to keep it chilly, just not deathly hot, so I expect it would still be fine.
Post reply on HN