Next generation autopilot - Model S vs XC90

[Gizmotoy]

It's all in the software and processing power. What is Volvo using for processing? Plus, Tesla knows how to highly advance software.
Is Volvo relying on third parties like MobileEye?

It was in the OP, but they're using the same hardware as Tesla's AP 2.0. Software is unknown. Previous generations of their system used MobilEye, just like Tesla. The two are on a pretty similar path, it appears.

In terms of redundancy Tesla's no hardware just falls flat. I think it will have a difficult time getting through regulations

I mentioned this in another thread, but this is a serious concern for me as a stockholder. All of the other automakers are bringing advanced sensor suites to the party. For long-range vision Tesla plans to make do mostly with cameras and its one radar unit.

It's unclear right now whether those sensor suites are necessary, but if you have 10 automakers and 9 of them are telling regulators advanced self-driving systems need these sensors to be safe, Tesla's in trouble. They seem to have anticipated this, resulting in the "Shadow Mode" Elon mentioned. So presumably they're planning to build a set of data that proves a camera-only solution is safe, but I think it's fair to say there's some risk here.

 

[diamond.g]

In terms of redundancy Tesla's no hardware just falls flat. I think it will have a difficult time getting through regulations

Yeah, I am not sure how they can get L4/5 without the redundancy. IIRC Level 3 doesn't need it as the person is expected to take over.


Though now that I am looking at http://www.nhtsa.gov/staticfiles/rulemaking/pdf/Automated_Vehicles_Policy.pdf it appears that the Levels are different? The highest level is 4 there. Based on this PDF it appears there are no talks about redundant systems in case of hardware failure. Maybe it is old.

 

[calisnow]

...Try being limited to complete loss of vision if even 1cm of your windshield has dirt on it. That's the issue...for all other purposes radar is far better than camera. Tesla even admits as much... It's funny that they don't realize that same necessity for all other angles of the car, ESPECIALLY being that only the front camera has any way of clearing the glass in front of it.

Good point. So, obvious question. Assume Tesla's engineers are as smart as you are for the moment and it has in fact occurred to them that dirt gets on camera lenses. What then do you theorize is the reason that Tesla choosing to implement only one radar, facing forward? Cost? Something else?

I think he's alluding to redundancy, because it doesn't seem that the Tesla has much of it in place. From the picture of the placement of the cameras, if one of the cameras goes out, there are portions that are not overlapping which may not be enough to satisfy the full autonomous driving.

Do you think cost is the reason or something else?

In terms of redundancy Tesla's no hardware just falls flat. I think it will have a difficult time getting through regulations

Assume Tesla thought of this. You think it's cost savings? Have the regs been written yet or are we all speculating on what they will say?

 

[Max*]

Assume Tesla thought of this. You think it's cost savings? Have the regs been written yet or are we all speculating on what they will say?

I have seen some sites talk about redundancy. I just checked the NHTSA Autonomous driving page, and it lists nothing about redundancy. Also checked the SAE site, nothing there either.

So you know what? I'll take back my stance until I can find the documents where I originally read about it.

 

[calisnow]

I have seen some sites talk about redundancy. I just checked the NHTSA Autonomous driving page, and it lists nothing about redundancy. Also checked the SAE site, nothing there either.

So you know what? I'll take back my stance until I can find the documents where I originally read about it.

I'll say, however, regardless of regs I will have a hard time falling asleep without solid redundancy. That's just me.

 

[diamond.g]

I have seen some sites talk about redundancy. I just checked the NHTSA Autonomous driving page, and it lists nothing about redundancy. Also checked the SAE site, nothing there either.

So you know what? I'll take back my stance until I can find the documents where I originally read about it.

Yeah me either, I looked and realized that SAE and NHTSA don't even agree on the the available levels of autonomy. SIGH

Anyways, having redundancy seems like a smart idea, not sure why it wouldn't be included. Especially if there is no requirement for human presence in the vehicle while in operation.

 

[J1mbo]

Good point. So, obvious question. Assume Tesla's engineers are as smart as you are for the moment and it has in fact occurred to them that dirt gets on camera lenses. What then do you theorize is the reason that Tesla choosing to implement only one radar, facing forward? Cost? Something else?

Do you think cost is the reason or something else?

Assume Tesla thought of this. You think it's cost savings? Have the regs been written yet or are we all speculating on what they will say?

Err... didn't they just massively increase the price of AP? Don't think cost came into it much. If they thought rear-facing radar was necessary, even if it added another $1000 to the BOM, then they would have put it in and adjusted the customer price accordingly. Point is, they don't think it is necessary, and to some of us at least, this is a curious decision.

Elon said in the press call that they were careful to site the cameras, and that they are all heated. However, the reversing cam hasn't been re-located, so what have the smart engineers done to address the fact that it gets wet and dirty from time to time?

 

[u00mem9]

As for your example it boils down to "Non-existent-vaporware-x would be superior to actual-product-Y in the following scenario."

It seems more like "non-existent-vaporware-x vs non-existent vaporware-y" at the moment.

My money says we are going to be sitting here having the same conversation a year from now...The only difference will be that the people that have provided $8000 interest-free micro-loans to Tesla are going to have smoke rolling out of their ears!

 

[calisnow]

Being a Californian I will be satisfied with "sunny day autonomy only" on the AP 2.0 cars. For those poor souls living in Seattle - I'm sorry.

 

[whitex]

This, like all other comparisons thus far is not valid because Volvo is vaporware and Tesla is shippable product. A limited run 100 car test in specially designated zones of Copenhagen or London does not count as shippable product in my book.

AP2.0 as shipped today does less than AP1.0. What exactly is the definition of "shipable product" in that book of yours? Taking pre-orders for future software product, which is what ordering AP2.0 functionality today is, does not qualify as "shipable product" to most people.

The OP's point about lack of redundancy is a very valid point. Given that the current AP1.0 can't even reliably figure out the speed limit after 2 years of being "shipable" doesn't inspire my confidence that I'll be able to reliably have my car pick me up at the airport any time soon.

 

[calisnow]

AP2.0 as shipped today does less than AP1.0. What exactly is the definition of "shipable product" in that book of yours? Taking pre-orders for future software product, which is what ordering AP2.0 functionality today is, does not qualify as "shipable product" to most people.

The OP's point about lack of redundancy is a very valid point. Given that the current AP1.0 can't even reliably figure out the speed limit after 2 years of being "shipable" doesn't inspire my confidence that I'll be able to reliably have my car pick me up at the airport any time soon.

Nobody said the OP's point about lack of redundancy is not valid. The more redundancy the better. But Volvo is pure vaporware. Tesla is shipping product. If you want to argue that the software is not turned on so it doesn't count as shipped product while it is being refined, running in shadow mode and being validated through the fleet - that's your call. To me it is and the larger point still stands - it is nonsense to compare Tesla's system to anyone's R&D project which is not orderable in any form.

AP 2.0 as shipped today is in the same boat AP 1.0 was in as of October 2014. It was an entire year before the shipped hardware was turned on in 7.0. To me, Tesla proved themselves with AP 1.0 - so I am taking their word that AP 2.0 will do 90% of what they claim it will and they will indeed gradually turn on the features.

As for speed limits I'm sorry about your experience in Seattle - my experience here in Southern California for 16,000 miles of driving has been that AP 1.0 is reliable in detecting speed limits - I just drove 150 miles last night on I-10 to Hollywood and back from Riverside and it nailed every sign.

With Tesla's model I believe every new hardware generation is likely to be "turned off" while the fleet verifies, checks, etc.

Also - sorry about your speed limit experience up in Seattle. Down in L.A. at least I personally have found speed limit detection to be reliable over 16,000 miles of driving - including last night when it nailed every change on I-10 over 150 miles of driving round-trip to Hollywood and back. Maybe it's bad where you are.

 

[acarlos1000]

My wife has a 2016 XC90, this is the first car from Volvo's next generation of cars after Ford left the company. This is an incredible vehicle and it does have some limited capabilities of autonomous driving up to 30 mph. But the new S90 Sedan will allow for autonomy up to 80mph.

From all traditional automakers, I think Volvo is the one that is the closest to the direction of Tesla (besides having a full EV option). The comfort and interior design on new Volvos is way superior than what I see on Tesla (both model S and X). But Tesla just moves faster than traditional car makers, Volvo updates it's system maybe 2 times a year (And I need to go to the service center to do that), you can't compare that with Tesla's continuous improvement.

 

[acarlos1000]

Here is a more recent video that compares the new XC90 with the new Audi Q7, for the Emergency Breaking Test go to 3 minutes on the video.

 

[calisnow]

My wife has a 2016 XC90, this is the first car from Volvo's next generation of cars after Ford left the company. This is an incredible vehicle and it does have some limited capabilities of autonomous driving up to 30 mph. But the new S90 Sedan will allow for autonomy up to 80mph.

That's great to hear about the S90 - the more competition in the marketplace the better for all of us as consumers.

 

[green1]

What a concept, AEB that avoids a crash. Tesla could learn something from that!

 

[calisnow]

What a concept, AEB that avoids a crash. Tesla could learn something from that!

Does 8.0 improve on the AEB? I thought Tesla claims it is 80% quicker to react.

 

[Muhammad]

Sounds like someone's been following this thread:

Tesla is getting self-driving cars to market first by being imperfect, but better than humans

 

[Ohji]

I think Tesla likely has great hardware for AP 2.0, but I still wonder why they didn't just add a few more radar sensors for redundancy. They aren't expensive, and they would have acted as a failsafe in case of side/rear camera malfunction. Maybe these enhanced ultrasonic sensors are significantly better than their predecessors? I only bring this up because Elon Musk himself has spoken about the critical need for redundancy.

I'm not questioning Tesla's choices here -- I just know that there is a good chance that any regulatory body that approves full autonomy is likely going to require a some decent amount of redundancy.

AP 2.0 is far more than I could have dreamed of, but Tesla is setting expectations really high with this update. The caveats are there in the fine print, but a lot of people are going to be unhappy if a new suite of sensors is necessary to achieve the autonomous driving hinted at with FSDC.

And one thing we haven't considered... Because autonomous driving hasn't been fully tested and no one knows what regulatory bodies will require, Tesla may have prepared for the possibility that additional sensors will be needed and provided a path to add additional sensors in the future. I know... this may be a pipe dream, and if I were Tesla, I would have shouted this contingency plan from the rooftops when announcing the new Autopilot, but hey, I can hope right?!

 

[calisnow]

...Because autonomous driving hasn't been fully tested and no one knows what regulatory bodies will require, Tesla may have prepared for the possibility that additional sensors will be needed and provided a path to add additional sensors in the future.

I would say you are crazy - except for the fact that Tesla is really laying it on the line with this announcement, talking Tesla network of self driving rideshare, etc. These are not small claims - these are huge claims. So, perhaps you are right - maybe there is an upgrade path built-in "just in case" regs require more later...

 

[MP3Mike]

The only reason I can think is that the more sensors you have the more processing power you need, and maybe the power, cooling, and space needs for the additional processors was just too much for what they need now.