Rivian May Be Much Closer to FSD Than You Think

[Engi_Nerd]

People get all up in arms about lidar and GPUs but the sensing / inference stack doesn't change US road design, maintenance, and driver discipline being some of the absolute worst. For me, the likelihood of coming across dangerous outliers seems like it will remain too high to go to sleep for the foreseeable future, no matter how good the car is. I think of FSD as "zero pedal driving" and in that role, it's currently peerless.

Sponsored

 

[captainjp]

But the price is insane. If I can do a drive like Dallas to Houston 98% hands-free and only pay a one-time $999 for a Comma 3X, I can't see why I would ever pay a $99/month subscription for something like FSD.

*shameless plug*

 

[gerrylum]

*shameless plug*

How is it a "plug"? You say that like I work for Comma or something.

 

[azbill]

We'll see if that ends up panning out. Volvo just said it's going to start removing LIDAR from their EX90s and ES90s as it just ended their relationship with Luminar. I have a suspicion that, while it sounds good on paper, there just isn't an ROI that makes it make sense to keep including it on the cars.

They announced eyes off driving on the Escalades starting in 2028 models and have already selected the Lidar for it. GM already proved out self driving with Cruise, they used to have the driverless Bolts in Phoenix, along with Waymo.

 

[azbill]

People get all up in arms about lidar and GPUs but the sensing / inference stack doesn't change US road design, maintenance, and driver discipline being some of the absolute worst. For me, the likelihood of coming across dangerous outliers seems like it will remain too high to go to sleep for the foreseeable future, no matter how good the car is. I think of FSD as "zero pedal driving" and in that role, it's currently peerless.

I am working on an SAE committee that is writing the standard for certification of AI in aviation. One of the key issues is that AI always has uncertainty, as you have pointed out.

One of the ways to get around uncertainty is to use dissimilar sensors and dissimilar AI algorithms. Sensor fusion is one of the newer terms for doing this. If you are using two sensors, such as radar and cameras, if they disagree you can hand control back to the driver or pilot, that provides a level of safety but not continued functionality. If you have three dissimilar sensors; lidar, radar and cameras, if one disagrees you can ignore it and continue. The case of three in disagreement at the same point in time becomes much lower in probability (but it does not become zero).

 

[Motoarzon]

I own a Tesla and FSD isn't and hasn't been Full Self Driving, it's a misnomer and shady marketing at best. This is not a benchmark Rivian should be shooting for, namely almost a decade of poorly performing promises on a $5-8K option or a $99/Month subscription.

If that's what December 11th is going to be, it's going to be hot garbage 2.0.

Now if you want to talk about automated driving and how Rivian is going to approach it, I'm in, but if you mention an FSD competitor, that's $$$ down the drain.

Then explain all the hundreds of youtubers with go-pros inside their cars literally recording video's for up to hours on end without ever touching the wheel or peddles? Doing real world random Uber rides, navigating almost impossible situations, going through drive-throughs, finding it's way in and out of above ground & below ground parking lots, driving on "unmarked" dirt roads and even construction site roads. You are either confusing Autopilot with actual FSD or you don't have the latest version 14.2 or you don't have hardware HW4?
This one is from Farzad, there are dozens of his videos you can view. "The whole mars catalog" youtuber only does FSD videos and has hundreds of them too. go watch his latest one's

 

[Motoarzon]

Based on where Rivian is at with their highway assist there is no way Rivian is even close to anything full FSD. Their highway assist is very poor and glitchy with lots of nuisances.
A few months back on vacation I rented a Hyundai Elantra and drove from the the tip of the Florida keys all the way up to Orlando. It had 2 independent buttons, 1 for adaptive cruise control and 1 for lane keep assist. Activating both of those together had me cruising for hours without ever touch the wheel and peddles. The performance on this crappy $25k car was way, way better than what I've ever achieved in my R1S.

 

[Zorg]

My experience with gen2 highway assist is actually pretty good. While it has too many disengagements when it's on it's pretty good. It handles curves quite well, especially compared to gen1.

 

[portdirect]

My experience with gen2 highway assist is actually pretty good. While it has too many disengagements when it's on it's pretty good. It handles curves quite well, especially compared to gen1.

Until around ~3 months ago - comparing gen1 and gen2 highway asset was comparing two generations of mobileye. Its only recently that they have started to actively use the in-house stack for highway assist - which as of now I would 100% say is 'better' than mobileye, but when initially activated was pretty sketchy often crossing lanes etc.

 

[Dark-Fx]

They announced eyes off driving on the Escalades starting in 2028 models and have already selected the Lidar for it. GM already proved out self driving with Cruise, they used to have the driverless Bolts in Phoenix, along with Waymo.

I keep seeing this Bolt locally, this was last Thursday.

 

[portdirect]

Putting this out there as a reminder that we should temper our expections for UHF etc in current R1's:

With the R2 they will actually achieve zonal arch, rather than the halfway house we are in with gen2 R1's where they reduced the ECU count in many cases by colocating domains on the same board rather than fully collapsing them (eg I dont expect to see the an Ambarella SOC present etc). But the significant thing called out in the above diagram is the implication that R2 will be based around Thor, which provides a real horsepower boost to local inference. I fully expect a gen3 R1 to follow the R2 fairly quickly - with minimal changes to gen2 other than a new XMM and steering wheel.
The fact they seem to be keeping the lid so tight on the R2 test mules, and still taking the original hand built R2 to trade shows makes me think they have something they want to keep under wraps for now - possibly to announce the 11th Dec?

 

[Zorg]

Until around ~3 months ago - comparing gen1 and gen2 highway asset was comparing two generations of mobileye. Its only recently that they have started to actively use the in-house stack for highway assist - which as of now I would 100% say is 'better' than mobileye, but when initially activated was pretty sketchy often crossing lanes etc.

Gotcha. We just got our gen2 in September and have been impressed. We live in the mountains so the freeway twists and turns quite a bit. Gen1 is useless there, but gen2 handles the turns like a champ, a la Tesla.

 

[mkhuffman]

Putting this out there as a reminder that we should temper our expections for UHF etc in current R1's:

With the R2 they will actually achieve zonal arch, rather than the halfway house we are in with gen2 R1's where they reduced the ECU count in many cases by colocating domains on the same board rather than fully collapsing them (eg I dont expect to see the an Ambarella SOC present etc). But the significant thing called out in the above diagram is the implication that R2 will be based around Thor, which provides a real horsepower boost to local inference. I fully expect a gen3 R1 to follow the R2 fairly quickly - with minimal changes to gen2 other than a new XMM and steering wheel.
The fact they seem to be keeping the lid so tight on the R2 test mules, and still taking the original hand built R2 to trade shows makes me think they have something they want to keep under wraps for now - possibly to announce the 11th Dec?

My Gen2 R1 is a freaking awesome truck. Call me a fanboy, I don't care. I am very excited about Gen3, as I expect it will be better in many material ways.

I am hoping for a larger battery, and 800V charging. And you just provided yet another area where an improvement is worth the upgrade. (I used to think native NACS was a big deal, but I have no issue with the adapter. And an adapter will be needed with NACS also. That's just the way it is.)

 

[nsuglerio]

Driving under trailers, running into medians, and driving down trains tracks. You’re right.
Derp

What a cogent and thoughtful response. I continue to use driver+ with my Rivian even after it hard veered off the road in the past. Edge cases don't negate the net positives to me. May as well shout into the void but ADAS systems will continue to evolve. Given where RIvian's goals are post Dec 11th, I think you may see yourself out with this brand as well.

 

[Donald Stanfield]

I’m excited for the day humans stop operating motor vehicles & death rates plummet

That day is a long way off. Even if it were perfected today, which it will be another few years at least it's going to take a long time before enough people have one to alter death rates significantly.

Until around ~3 months ago - comparing gen1 and gen2 highway asset was comparing two generations of mobileye. Its only recently that they have started to actively use the in-house stack for highway assist - which as of now I would 100% say is 'better' than mobileye, but when initially activated was pretty sketchy often crossing lanes etc.

I agree with the current state of highway assist. The release of the Rivian system was worse than the Mobileye system that was there before it, but Rivian has since surpassed it by a significant degree. I drove for a few hours using it in heavy traffic and was impressed by how well it navigated the traffic and the decisions it made while switching lanes, adapting its behavior to the situation. Heavy freeway traffic jams are among the most challenging traffic situations to navigate, and Rivian performs remarkably well.

Sponsored