Ketchum: Wait for LiDAR? - The Rivian R2 LiDAR story isn’t what you think

[Thebandit]

I've said it before and still stand by my earlier comments- Rivian is not developing LIDAR for direct sale to the public (us) in the near future.

They're developing it for full autonomous taxi use through Uber. They're planning on rolling out Uber branded Rivian full autonomous vehicles by 2028.

I don't get the strategy folks seem to subscribe to, in which if we just wait a little bit longer we can totally buy a Gen 3 R1x or R2 with LIDAR.

I think Gen 3 hardware will fork - LIDAR systems for Uber partnership vehicles. Non-LIDAR for the rest of us.

The software stack to run full autonomous driving for Uber is going to be nearly 100% unique to that platform, and will not easily plug into a vehicle meant for sale to the general public. I do not believe Rivian can pick-and-choose what sections of software they write to leverage LIDAR data to perform full autonomous driving and graft that into something that looks like Autonomy + they sell to us.

At some point many years down the road, (Gen 4+) things may be different, but given how fundamentally different the R2 software is from R1, I just don't see Rivian writing code in a modular format that would enable LIDAR technology on things we can actually buy any time soon.

What? They have specifically said that R2 with LIDAR and RAP1 will ship late this year or early next year.

Sponsored

 

[Alan in Tempe]

Since the Lidar/RAP equipped R2s are now rumored to add a real FM radio, I will wait for it.

 

[shandering]

Doubtful... I was on the way home, with FSD driving me through my neighborhood, when it stopped for a squirrel running across the road. A human is much bigger than a squirrel...

And it will actively dodge road debris at highway speeds (shredded tires, branches, whatever).

Go put a dummy under the tire of your car and start FSD. They're instantly dead

That's my point. FSD has blind spots (bumper camera being one of them) and this is theoretically unsafe at scale

Robotaxi software build does seem to use the bumper camera but not while driving. It only uses the bumper camera to decide whether it needs to back up or not before starting a drive. the issue is the car often needs manual confirmation to start a drive and even then it still ran over something in tests from AI drivr

As a human you can both hear screaming, and have much more awareness of what's around you before you start a drive.

 

[justinkitswa]

I think "shipping vehicles with LIDAR sensors" is much different than writing software to allow the general public to take advantage of them. I don't see software existing in the next 12-24 months to leverage LIDAR sensors to be fundamentally different/better than the existing vision/radar systems.

I think the first several thousand of these to ship with LIDAR will be for generating data for developing their full autonomy models. And I suspect once they learn how large and complex those models are, they'll need additional compute resources in the vehicle to take advantage of them.

So I am not saying we will never see useful gain from the LIDAR sensors, I just caution against interpreting "shipping with LIDAR later this year" as meaning we will be buying vehicles that will be monumentally better at self driving in the near future.

They will be more expensive, however.

 

[Thebandit]

I think "shipping vehicles with LIDAR sensors" is much different than writing software to allow the general public to take advantage of them. I don't see software existing in the next 12-24 months to leverage LIDAR sensors to be fundamentally different/better than the existing vision/radar systems.

I think the first several thousand of these to ship with LIDAR will be for generating data for developing their full autonomy models. And I suspect once they learn how large and complex those models are, they'll need additional compute resources in the vehicle to take advantage of them.

So I am not saying we will never see useful gain from the LIDAR sensors, I just caution against interpreting "shipping with LIDAR later this year" as meaning we will be buying vehicles that will be monumentally better at self driving in the near future.

They will be more expensive, however.

Rivian has specifically said that LIDAR will not yield any immediate self driving benefits. This isn't some biting analysis. It's what the company line has been all year.

 

[macb00kemdanno]

Go put a dummy under the tire of your car and start FSD. They're instantly dead

That's my point. FSD has blind spots (bumper camera being one of them) and this is theoretically unsafe at scale

Robotaxi software build does seem to use the bumper camera but not while driving. It only uses the bumper camera to decide whether it needs to back up or not before starting a drive. the issue is the car often needs manual confirmation to start a drive and even then it still ran over something in tests from AI drivr

As a human you can both hear screaming, and have much more awareness of what's around you before you start a drive.

I'm not understanding the point of your scenario. Are you saying that you wouldn't notice someone lying across either in the front or back of your car, lying on the ground -- up against the wheels -- when you approach to get in? Like, WHAT?

What's next? Testing for someone asleep under your car before you back away? I'm sorry, but this whole line of reasoning is silly.

 

[rvnxyz001]

I don't see software existing in the next 12-24 months to leverage LIDAR sensors to be fundamentally different/better than the existing vision/radar systems.

Once Rivian gets lidar, they can just ask Claude or Codex to build the world’s best self-driving system. Should take about a week. Jensen Huang and many others already said coding is basically solved.

Then owners can beta-test it for another two or three weeks, report the bugs, AI will fix everything overnight, and by the end of the month we’ll all be sleeping in the driver’s seat.

 

[shandering]

I'm not understanding the point of your scenario. Are you saying that you wouldn't notice someone lying across either in the front or back of your car, lying on the ground -- up against the wheels -- when you approach to get in? Like, WHAT?

What's next? Testing for someone asleep under your car before you back away? I'm sorry, but this whole line of reasoning is silly.

my point is for unsupervised

As I said, a light amount of supervision (recognizing this and not pressing the start FSD button) would prevent a scenario like this

Tesla would also know the likelyhood of something like this happening and they would calculate risk with robotaxi

 

[NinjaWrap]

Go put a dummy under the tire of your car and start FSD. They're instantly dead

That's my point. FSD has blind spots (bumper camera being one of them) and this is theoretically unsafe at scale

Robotaxi software build does seem to use the bumper camera but not while driving. It only uses the bumper camera to decide whether it needs to back up or not before starting a drive. the issue is the car often needs manual confirmation to start a drive and even then it still ran over something in tests from AI drivr

As a human you can both hear screaming, and have much more awareness of what's around you before you start a drive.

All of the unsupervised cars have bumper cameras (and I don't think any without the bumper camera will ever be unsupervised).

There is still a massive line spot beside each front fender though, and in certain sharp angles turns (which are more frequent in the mountains where I live) I notice FSD has issues where it obviously can't see something at a certain angle and then as it's moving further into the turn it sees it and quickly adjusts to avoid it.

 

[llninja]

People also told us the R2 backlog was years out and new reservations wouldn't get their cars till 2028 or later. If the order backlog is thinner than they want, Rivian could conceivably keep launch edition going to get these higher dollar sales. Rivian is hungry for R2 sales. Just look at the way they are prioritizing expiring leases ahead of people who waited 2 years in line.

If prices change, consumers change behavior too. I bailed out on buying a Model Y when prices started increasing rapidly a few years ago. A lot of people are waiting on Premium trim just because of the $4k difference regardless of any freebies with the Launch Edition.

I was late with my reservation in Nov 2025 and just got an email that my reservation should convert to a configuration sometime between Dec 2026 and Feb 2027. That‘s not too bad.

 

[s4wrxttcs]

He's not wrong. It's a good video that everyone should watch.

It's so obvious to me from a business perspective that Rivian is including Autonomy+ for free on LE so people will buy it and not wait for RAP1/LIDAR. LE ends when RAP1/LIDAR begins IMO. Ironically, I have a feeling some of those waiting for RAP1/LIDAR will regret that decision later on. You will have to pay $2,500 for Autonomy+ and maybe more for the car if Rivian raises prices. Anyone seen the prices of RAM these days? Even Apple isn't immune from price increases, so no doubt Rivian will increase prices too. "When" is the ultimate question.

To each his own and some want to hedge against future autonomy changes, but it seems to me Rivian is a LONGGGGG way away from that; they have to catch where Tesla is today first. They said the current chipset has plenty of headroom for point to point/hands off autonomy like Tesla has now. Good enough for me.

YMMV

I completely agree with you, but that's not the only reason to hold out.

It's generally good advice to hold out on a new model for at least 1 year.

A lot of people also have their own unique circumstances that factor into the timing decision.

In my case I have a 2022 R1T which works great for my needs. but the warranty will be expiring soon due to mileage. This creates out of warranty cost concerns so this might ultimately factor into an R2 buying decision earlier than I had planned.

Or I might get an extended warranty if Rivian offers it for my state.

In any case over the next few months I'll be keeping an eye on extended warranty threads.

Lidar is intriguing, but like you said I think its too far away of utilizing it. It might be best to bag some savings initially. Maybe get one of the last LE editions.

I will miss the truck aspects of the R1T though. Sure I can use the back of the R2 if I'm really vigilant about protecting it and cleaning it after each home depot trip. Or maybe I'll supplement it with a slate. Haha.

 

[Jeremy3292]

I completely agree with you, but that's not the only reason to hold out.

It's generally good advice to hold out on a new model for at least 1 year.

A lot of people also have their own unique circumstances that factor into the timing decision.

In my case I have a 2022 R1T which works great for my needs. but the warranty will be expiring soon due to mileage. This creates out of warranty cost concerns so this might ultimately factor into an R2 buying decision earlier than I had planned.

Or I might get an extended warranty if Rivian offers it for my state.

In any case over the next few months I'll be keeping an eye on extended warranty threads.

Lidar is intriguing, but like you said I think its too far away of utilizing it. It might be best to bag some savings initially. Maybe get one of the last LE editions.

I will miss the truck aspects of the R1T though. Sure I can use the back of the R2 if I'm really vigilant about protecting it and cleaning it after each home depot trip. Or maybe I'll supplement it with a slate. Haha.

I suspect R2 will have much better build quality bc it’s a much simpler vehicle optimized for cost but time will tell. For example, no air suspension which is a headache for R1 owners.

 

[KineticKev]

The debate regarding LiDAR with the new RAP1 chip is fascinating. Some will choose to hold out while others will jump right into LE.

Most of us have probably weighed both sides. I know I went back and forth on it, but I ultimately decided that the LE perks are worth it.

I've done an exhaustive amount of research on the R2, and am happy that both versions share the exact same updated cameras and the next-gen infotainment chip. The primary differentiator, as many have pointed out, is that the Gen 3 platform will introduce LiDAR paired with RAP1. While that hardware will eventually boost the roadmap toward higher-level autonomy, Rivian's initial engineering efforts are still heavily focused on enhancing the overall vision-and-radar flywheel for all Gen 2 models, including the R1S. It's also a long term cost cutting measure.

The fact that R2 launched without fundamental features like Pet Mode strongly suggests to me that we will see plenty of standard incremental updates before we get anywhere near actual "eyes-off" Level 3 testing. By the time that level of autonomy is fully realized, we will likely already be looking at RAP2 and LiDAR 2.0.

 

[macb00kemdanno]

My only reservation surrounding point to point is how quickly can Rivian establish a stable baseline, i.e. something comparable to FSD 14.3.x?

Tesla has millions of vehicles feeding the machine and over a decade of data. With fewer inputs, it will be interesting to see the update cadence.

 

[Zathras]

I thought they said UHF would work on any roads with clearly painted lines, not just mapped areas. I'm I missing something?

No, you're not missing anything. All Gen 2 vehicles will have point-to-point driving on 3.5 million miles of roads that have painted lines.

Sponsored