More reliable lane changing. The Model S can't tell if a car is approaching from the rear, in a nearby lane, at a high rate of speed. Even if you de-fisheye the fisheye, you can't see too far in the nearby lane, you need rear radar for that. I can see that being APv2 hardware. It could also lead, in the future, to automatic lane changing (though I don't know if more hardware would be required).
With all that being said, I already stated my speculation. Model X AP = Model S AP. I can see them adding a 360 degree camera, or something, but that's not AP related.
We'll find out in a few weeks.
Yes, fingers crossed. I suspect you're right.
Advertised 1.0 features will have to be made to work, and made to work safely. They'll be concerned about reliability as a safety issue. To the extent "more reliable lane changing" could be construed as a safety matter, and not seeing far enough in nearby lanes were making it risky enough to make a lane switch, hardware back-porting seems possible.
That said, it may not, and I think you make a good point that they might acquit themselves of "auto lane changing" with a simple implementation in 1.0 and improve it as you describe or in other ways.
For autonomous lane changing, I wonder how much visibility in the rear is the limiting factor, versus processing power and politics.:smile:
No one said anything about bringing back Model S vehicles for sensor upgrades. They will remain with Ver. 1.0 sensors and I'm sure Tesla will deliver on basic autopilot capability with the Ver. 1.0 sensors in the Model S.
What the Ver. 1.0 sensors lack is redundancy as well as breadth of coverage. Other cars with half the advertised ability of the Tesla Autopilot have more comprehensive sensor suites and I feel pretty certain that the Model X will have a more advanced sensor suite.
Only speculating, of course... I think there is a window in which "better" and "safer" coincide - inside of which, I think they may admit defeat on their "released a year ahead of the software" 1.0 sensor package. They don't want safety incidents with 1.0 that could have been avoided in 2.0 sinking their entire investment in the technology, not to mention, their brand. They already know anything that goes wrong will be catnip to the news, and covered all out of proportion to the tens of thousands of people in the US who die of human error yearly.
What you're saying about competitive sensor packages and systems is interesting.
To me it's unusual if, selling the hardware so far ahead of yourself, and giving up the ability to tweak it in tandem with your software until launch, you can yet pull off your launch. So I will be impressed if Tesla can deliver 1.0 without new hardware. Maybe they can.
As for the Autopilot sensors, based on what Elon has said in the past, I feel chances are good that the MX will have a more advanced sensor suite. It does not mean that the Model S will need to be retrofitted or the more advanced sensor suite will be needed to deliver basic lane changing capability promised with Autopilot, but the enhanced sensor suite will offer additional and more robust capability over the version 1.0 sensors. Ver. 2.0 could only lane change when it is safe to do so while Ver. 1.0 sensors require you to ensure it is safe to change lanes...
For lane changing reliability in particular, I think this might be hair-raising for the safety folks, but I agree what you say could come to pass.
There are many creative things you can do with rear radar... The Mercedes S class for example, if you slam on your brakes, it actually checks the rear radar to see if there is a car behind you. If there is a car that might rear end you, the S Class might not apply full braking power. It will apply just enough braking power to prevent an accident with the car in front of you while offering the car behind you, maximum time and space to slow down and hopefully not crash into you.
Another area is cross traffic situations when you pull into a road or back into a road where you can't see around an obstacle. A more advanced sensor array can help detect traffic you can't quite see yet and the current MS sensor suite isn't sufficient for that.
Very cool. And agreed, more good examples of what would be distinctly new features that would justify new hardware.
I would love it if they were progressing that quickly. Who can say for sure? But I end up betting it won't happen, because they're still struggling to ship their 1.0 features. If autosteer slips to 2016, it would not surprise me. Putting a new sensor suite into production, for
new unreleased features, when the existing features still aren't out the door, seems wild to me. It's just intuition and rank speculation, but if we see new sensors, my first thought will be a failure to launch 1.0 on the old ones.
Then add in the reports of private beta testing of AP on the existing hardware on open roads... And given 2.0 sensors (and the integrated system they go into) would have had to be in development since 2014 or 2013...
And that excellent info on the EyeQ... I think Max is right, and the X sensors will either be the same, or not hugely different.
