Driver+ tried to put me in the wall today!

[OrthoBlock]

Today's incident was a bit of a surprise. Was in the fast lane on the freeway with only the wall next to me on the left. Freeway was curving ever so slightly to the left. The truck all of sudden tried to make a hard left. Was certainly far too close for comfort for me. While my hands were already on the wheel and I was ready, already being close to the wall as it was, there wasn't a lot of margin for error. I caught it in time, but did certainly put a scare into me as it was so unexpected.

For what it's worth, the exact same thing happened to me, on exactly the same situation you describe: freeway, left lane, next to wall, slight left turn --> truck suddenly tries for a hard left.

EDIT: ...also saved it, and still love the driver assist feature, but will definitely need to pay attention all the time.

Sponsored

 

[ironpig]

Having had a Tesla with "autopilot" and now the Rivian, I never use these systems. They are not ready for primetime. The only one I'll both with is the distance controlling cruise control.

 

[iansriv]

While I agree with your comments in principle, I will challenge you on the plane analogy. Modern aircraft have sophisticated autopilot features that not only can fly the plane when it reaches cruising altitude but can also take off and land in emergency situations. Pilots spend a significant amount of their flight time in autopilot because it has been shown to be safe and effective over millions of flight hours.

Airplanes aren't a great analogy as FAA requires 3 miles of separation and there are no "walls" in the sky to run into.

Personally, I would never fully trust autopilot in vehicles because the technology simply isn't there yet. Nearly all accidents occur because of operator (human error). The human brain is capable of 63 quintillion operations per second, which as of a ~ decade ago, constituted all known computing power worldwide. It's also the most efficient information processor ever conceived and can be powered by a ham sandwich and a glass of milk.

We've made strides since then in computer vision, neural networks, and power efficiency but it is light years away from replicating the human brain.

Keep your attention on the road, hands on the wheel, and keep those fingers away from your phone and you'll do fine.

Here here! I absolutely second that. When did it become so difficult to drive your own car.

 

[kizamybute']

Please Submit a service ticket to Rivian so they can try and track it down.

Hitting the flashers for a second creates a log point for them to look at as well.

Good to know for future reference. Not sure it's worth submitting a service ticket. Don't think there's anything "wrong" with the truck. It just apparently saw something that wasn't there. Tesla's system does it too. The only real different I notice between the Tesla and Rivan systems is that it does take more force in the Rivian to take over and knock it out of Self Drive mode. The Tesla, I could easily do without the car moving. With Rivian, definitely have to give it more of a tug, which so far, has resulted in me jerking the truck by taking control back. Hoping as I learn the right touch over time, can minimize that. Still all new, so giving it some time. But will say, it would be nice if it took less force to get control back.

 

[kizamybute']

Here here! I absolutely second that. When did it become so difficult to drive your own car.

I'm with you. I had FSD in my Tesla, almost never used it. Lane keep assist on the freeway was about the only place I found it useful on occasion. The Rivian offers that, so don't feel like I'm missing much by not having the Tesla anymore.

Plus, airplanes have a lot less traffic to deal with. Certainly tougher to design an "auto-pilot" system to handle the unique circumstances that come up in every day driving. The first 95% was pretty easy for the programming geniuses, the last 5% has been really, really difficult. Understandably so.

 

[SeaGeo]

Good to know for future reference. Not sure it's worth submitting a service ticket. Don't think there's anything "wrong" with the truck. It just apparently saw something that wasn't there. Tesla's system does it too. The only real different I notice between the Tesla and Rivan systems is that it does take more force in the Rivian to take over and knock it out of Self Drive mode. The Tesla, I could easily do without the car moving. With Rivian, definitely have to give it more of a tug, which so far, has resulted in me jerking the truck by taking control back. Hoping as I learn the right touch over time, can minimize that. Still all new, so giving it some time. But will say, it would be nice if it took less force to get control back.

Unfortunately the service ticket is how they can track and work on software bugs. That's where my guide landed with me, but I've also confirmed that with someone who works on D+. Flash the hazards, and submit a service ticket. I've done this for it misidentifying gantry structures that go over the road as tunnels as an example.

I agree, I do not like the force required to break highway assist.

 

[lostpacket]

Good to know for future reference. Not sure it's worth submitting a service ticket. Don't think there's anything "wrong" with the truck. It just apparently saw something that wasn't there. Tesla's system does it too. The only real different I notice between the Tesla and Rivan systems is that it does take more force in the Rivian to take over and knock it out of Self Drive mode. The Tesla, I could easily do without the car moving. With Rivian, definitely have to give it more of a tug, which so far, has resulted in me jerking the truck by taking control back. Hoping as I learn the right touch over time, can minimize that. Still all new, so giving it some time. But will say, it would be nice if it took less force to get control back.

the service ticket is mostly so the software engineers can investigate and improve the system. It might not fix anything for you in the near term but the more information they know about bugs the more likely they are to find the real reason and improve/fix it.

Edit: @SeaGeo is fast

 

[thrill]

Good to know for future reference. Not sure it's worth submitting a service ticket. Don't think there's anything "wrong" with the truck. It just apparently saw something that wasn't there...

That's basically the definition of "something wrong". While Rivian continues to suffer from consistency in assembly quality, the assembly team is not the self-driving team (we hope), and since there's multiple teams in the company all becoming better daily, the team that works on self-driving would quite likely be delighted to have the data to inspect knowing something didn't behave as it should.

 

[SeaGeo]

That's basically the definition of "something wrong". While Rivian continues to suffer from consistency in assembly quality, the assembly team is not the self-driving team (we hope), and since there's multiple teams in the company all becoming better daily, the team that works on self-driving would quite likely be delighted to have the data to inspect knowing something didn't behave as it should.

The "issue" is it definitely feels weird to submit a service ticket for your specific truck for something that's not specific to your truck. I really wish they had a process specific to submitting bugs and feature suggestions.

 

[thrill]

The "issue" is it definitely feels weird to submit a service ticket for your specific truck for something that's not specific to your truck. I really wish they had a process specific to submitting bugs and feature suggestions.

Software that makes decisions (ex: driver+) is applicable to any given vehicle on any given day in a similar situation.

For example, having taken formal classes in machine learning, and specifically the type generally being used for various levels of self-driving, I think some decisions made by Rivian, likely following Tesla's lead since that's the way they did it, are incorrect, specifically the removal of lidar, in the belief that they could get a sufficient answer simply with optical cameras. Lidar, while relatively expensive at perhaps a grand per vehicle these days, presents a lot of information from frequencies that radar, ultrasonics, and optical cameras do not. It's possibly too much information to be processed by the particular hardware configuration they chose, and so that'd be another expense, plus the engineering time to put it to good use from the multiple sensor inputs. Regardless, the self-driving team needs to know, and as I have no idea if the vehicle uploads all video (doubtful) or perhaps all video when it initiates a hard turn (possible, but I doubt that too), then it's likely up to the owners to flag misbehaviors.

Tesla's continued pushing of the envelope in self-driving, especially with their optical-only system limitations, is likely to eventually financially bite them in the ass (and really already should), as evidenced by nearly 300 autopilot crashes over the last year, and the recent fatalities in Florida seem to finally be getting the NHTSA more involved.

Rivian is obviously(?) continuously assessing their software's performance, and if it's inadequate in specific situations, then better, even though more expensive, hardware would likely be welcome, though it's more likely that they'll engage other limitations (disabling self-driving at night, for example), but at least they'll be able to try to make such decisions based on data before injuries or worse becomes the impetus.

Sorry for the diatribe.

 

[NY_Rob]

Tesla's continued pushing of the envelope in self-driving, especially with their optical-only system limitations, is likely to eventually financially bite them in the ass (and really already should), as evidenced by nearly 300 autopilot crashes over the last year, and the recent fatalities in Florida seem to finally be getting the NHTSA more involved.

I'm not even impressed by their regular/free Auto Pilot system w/radar no less the $12K FSD option with cameras only. Our Model Y (with radar) decided to suddenly slam on the brakes from 40+mph while approaching a 4-way intersection where only the left and right intersecting roadways had stop signs! We narrowly avoided getting rear ended by the truck behind us and it scared the shit out of everyone in the car. I got yelled at by everyone.. "turn that thing off!!!" That was the last time I used Auto Pilot on anything other than a major limited access road.

 

[SeaGeo]

Software that makes decisions (ex: driver+) is applicable to any given vehicle on any given day in a similar situation.

For example, having taken formal classes in machine learning, and specifically the type generally being used for various levels of self-driving, I think some decisions made by Rivian, likely following Tesla's lead since that's the way they did it, are incorrect, specifically the removal of lidar, in the belief that they could get a sufficient answer simply with optical cameras. Lidar, while relatively expensive at perhaps a grand per vehicle these days, presents a lot of information from frequencies that radar, ultrasonics, and optical cameras do not. It's possibly too much information to be processed by the particular hardware configuration they chose, and so that'd be another expense, plus the engineering time to put it to good use from the multiple sensor inputs. Regardless, the self-driving team needs to know, and as I have no idea if the vehicle uploads all video (doubtful) or perhaps all video when it initiates a hard turn (possible, but I doubt that too), then it's likely up to the owners to flag misbehaviors.

Tesla's continued pushing of the envelope in self-driving, especially with their optical-only system limitations, is likely to eventually financially bite them in the ass (and really already should), as evidenced by nearly 300 autopilot crashes over the last year, and the recent fatalities in Florida seem to finally be getting the NHTSA more involved.

Rivian is obviously(?) continuously assessing their software's performance, and if it's inadequate in specific situations, then better, even though more expensive, hardware would likely be welcome, though it's more likely that they'll engage other limitations (disabling self-driving at night, for example), but at least they'll be able to try to make such decisions based on data before injuries or worse becomes the impetus.

Sorry for the diatribe.

Rivian is taking a pretty different approach to it than Tesla.

Tesla also removed radar, not Lidar. Rivian still has Radar. I suspect that they also rely heavily in high accuracy GPS.

 

[lostpacket]

I'd imagine they have some observability and alerting built in for the software team to review. E.g. if driver+ is on, and the steering changes by more than 2% in 300ms (made up numbers but you get the idea). Plus I'd hope they are constantly reviewing disengage events.

But there isn't a substitute for explicit user reports. User reports are imperfect too, but they are explicit and the more information the better.

Note this is all speculation on my part, I'm not an ML engineer but have worked with quite a few.

 

[yizzung]

All of these systems are Level 2, which means you’re supposed to babysit them constantly and only cede control to the vehicle momentarily. It’s kinda dumb that people think they are autopilots and that the manufacturers play along with the public misperception. I have no intention of using these features and prefer to just drive.

 

[photontorque]

While I agree with your comments in principle, I will challenge you on the plane analogy. Modern aircraft have sophisticated autopilot features that not only can fly the plane when it reaches cruising altitude but can also take off and land in emergency situations. Pilots spend a significant amount of their flight time in autopilot because it has been shown to be safe and effective over millions of flight hours.

Airplanes aren't a great analogy as FAA requires 3 miles of separation and there are no "walls" in the sky to run into.

Personally, I would never fully trust autopilot in vehicles because the technology simply isn't there yet. Nearly all accidents occur because of operator (human error). The human brain is capable of 63 quintillion operations per second, which as of a ~ decade ago, constituted all known computing power worldwide. It's also the most efficient information processor ever conceived and can be powered by a ham sandwich and a glass of milk.

We've made strides since then in computer vision, neural networks, and power efficiency but it is light years away from replicating the human brain.

Keep your attention on the road, hands on the wheel, and keep those fingers away from your phone and you'll do fine.

Great post!

Unlike @thrill I do not have formal experience with the technical details behind machine learning, but I have some related experience and classes. I know to never underestimate technology, but I think safe and reliable self-driving cars are still a ways off. I hope it happens, it could just be a while.

My armchair thinking is that a human being driving a car applies not just "car-driving" knowledge while at the wheel, but also all the other life-experience of living in this world. Maybe that knowledge isn't tapped when the driving situation is "standard": well defined lanes, well-behaved traffic, no random non-vehicle objects or dynamics in the environment. But "standard" is in quotes because how much of the time is real driving in that kind of environment? Maybe mostly for some, very little for others.

Training a car to drive by collecting data about a road and other cars is necessary, but not sufficient. What about how kids chase after balls or ride bikes? What about how tree branches appear in different light and different seasons? Is that light through the mist and rain at dusk a porchlight or another car with a headlight out? It's going to take a long time for self-driving algorithms that just use data from driving to back fill all those other life experiences that we apply when driving. Again, my speculation.

Plus, the computational capacity of the human brain is . . . impressive. Getting a silicon equivalent in a car will take a while.

In any case, I am thankful for this thread, and related threads, as a heads up. My car is old enough that even the most basic driver assistance features didn't exist when it was made. I didn't plan on using anything from the Rivian in this regard, and now for sure I won't.

Sponsored