Electric Vehicle Thread!

[AlphaWolf]

I think it's very different running a nvidia GPU full time on end, then just running it for a couple simpler items.

You can bolt on self driving, that doesn't mean it's a good fit economically for it to be widespread among ICE vehicles. Trucks, sure. Lots of room there to support that type of thing. A mini cooper, Corolla, or Civic? These things get harder to do.

These sensors are tiny, they are in the headlights, and the mirrors the main nav system could no doubt host the innards from a high end laptop.

I think you are just underestimating how much of a modern vehicle is already 'drive by wire'.

 

[iroboto]

These sensors are tiny, they are in the headlights, and the mirrors the main nav system could no doubt host the innards from a high end laptop.

I think you are just underestimating how much of a modern vehicle is already 'drive by wire'.

Well, then there’s no reason for ev and hybrids to represent a majority of self driving features then.

I guess we will chalk it up to no one wants it

 

[AlphaWolf]

Well, then there’s no reason for ev and hybrids to represent a majority of self driving features then.

I guess we will chalk it up to no one wants it

I wouldn't pay much for what's currently available and allowed. When it allows my car to operate as a driverless taxi in my downtime. Then it's a significant feature.

 

[Silent_Buddha]

I would disagree, because the alternator would not provide enough consistent power regulation compared to a battery. We're talking about running a nvidia gpu right? https://blogs.nvidia.com/blog/2022/09/20/drive-thor they are going to be up to 2000 TFlops of self driving power in 2025. This isn't running some 20W card. If we use the alternator, your alternator has to be able to provide consistent power from stand still to highway speeds while still powering your auto, audio and air conditioning systems for up to hours at a time. Full self driving matters most when you have to do road driving and stopping for lights and avoid pedestrians. Limiting it to highway lane control is barely scratching the surface of what full self driving would require. I don't think the technology is there honestly. By 2025 we're going to be closer. But I think ICE vehicles will have challenges with the power component.

I totally think there would be self driving cars on ICE, the fact that there hasn't been, and we've only seen it on EVs and Hybrids sorta of shows that it's probably not economically feasible (there are exceptions in the luxury market). There isn't some agenda here to keep self driving on EV cars only. It's just simpler and therefore cheaper for EVs to run self driving systems. Otherwise we would see it everywhere. Nvidia has been doing this a long time.

Self driving alone is already an expensive feature. it's not cheap for the hardware. It is very much at this point in time a luxury feature, bolting it onto ICE will cost more than bolting it onto EV. That's my TLDR.

You shouldn't underestimate the ability of car alternators to generate the necessary power. People have been repurposing regular car alternators for decades to power their homes in the case of emergency outages. I'd be incredibly surprised if the right sized car alternator would have any trouble at powering anything needed for self driving. The alternator in my Ford Focus is rated for 150 amps. Most modern vehicle alternators (at least in the US) are rated for over 100 amps. They can easily run multiple PCs with 3090 Ti GPUs if required.

The alternators will have regulators that limit the power that they can transfer to the battery and rest of the vehicle based on the load that is being generated and the system that is being used. They have as otherwise it would fry the battery. Just like the electrical system in a house, the electrical system in a vehicle is designed to limit how much power goes to each system (the fuse box is one such system).

If you don't know what you are doing and bypass those systems you can easily fry all the wiring in your car as well as the systems they are connected to. I did this once when I was a teen decades and decades ago when I used to install car stereo systems for friends and family. That was a cost and time intensive lesson to learn as I had to replace all of the wiring in the car. Ouch.

So, for example, I have a 350 watt inverter for my Ford Focus that I use on trips that could drive one of my desktop computers. You can also install car stereo systems that draw in excess of 6000 watts if you wanted.

Regards,
SB

 

[tuna]

I think Fully self driving cars are like decades away. But ≈95% self driving cars are possible today (like Waymo and Tesla have shown) and that is real progress.

 

[iroboto]

You shouldn't underestimate the ability of car alternators to generate the necessary power

I don’t. But I don’t believe in conspiracy theories so that either means there is no demand for self driving on ICE or it’s not economically advantageous to do so.
It would appear; I am correct about AVs having better alignment with the properties of EVs.

General Motors:

Why it Matters: All-electric autonomous vehicles have many advantages over their gasoline- or hybrid-powered counterparts:

• Cleaner Environment: By 2040, it is estimated that there will be 40 cities with populations exceeding 10 million. All-electric shared autonomous vehicles will be ideal for dense cities that need solutions for congestion and noise pollution.
• Stable Power: The advanced sensing and computing hardware on an autonomous vehicle needs a lot of electric power. Compared to an internal combustion engine, an all-electric battery pack acts as a more stable power source that can enable higher-powered AV components.
• Low Latency: Whether you’re a human driver or an AV navigating public roads, reaction time matters. By the nature of their design, electric propulsion systems have lower latency and more consistent response when accelerating. As a result, when compared to internal combustion counterparts, an all-electric AV will have a lower delay between the time it decides to make and the time it completes a maneuver.

The Bottom Line: Electric vehicles allow for simpler integration of the advanced technologies required for the cleanest and safest operation of autonomous vehicles. That’s why we believe that in the long term, building all-electric vehicles with autonomous capabilities integrated from the beginning (rather than through retrofits) is the most efficient way to unlock the tremendous potential societal benefits of self-driving cars.


Source: https://www.gm.com/stories/all-avs-should-be-evs

 

[Silent_Buddha]

• Stable Power: The advanced sensing and computing hardware on an autonomous vehicle needs a lot of electric power. Compared to an internal combustion engine, an all-electric battery pack acts as a more stable power source that can enable higher-powered AV components.

That's the part that makes me laugh. Does it need greater than 10 kw of power?

It's already possible and common to run 1000's of watts of power worth of electronics in addition to everything else that comes standard on vehicles now-a-days. Most of the sensing equipment is already available on many if not most ICE vehicles. That leaves mostly the computing unit. Does that actually require multiple 1000's of watts of power?

It may or may not be easier, but it's certainly a selling point to attempt to convince enough people to buy an electric vehicle to meet potential future government requirements on percentage of a fleet of sold vehicles being all electric. That's not a problem for an ICE vehicle.

That's what most people want to use and are comfortable using, so self driving isn't exactly a selling point that's going to move appreciably more ICE vehicles. Especially when faced with the more pressing need to convince your average joe schmoe consumer that they should spend significantly more money on an EV versus a much cheaper ICE vehicle. And that's where at least some form of limited self-driving capabilities might convince someone to buy an EV instead of an ICE vehicle.

BTW - In 2019 an ICE vehicle (a freight truck) using self driving AI drove across the US.

A Self-Driving Freight Truck Just Drove Across the Country to Deliver Butter

It made the daunting 41-hour trip from Tulare, California to Quakertown, Pennsylvania.

www.popularmechanics.com

We hear more about EVs, however, because it's easier to get government grants to assist with EV R&D for SDVs.

Regards,
SB

 

[iroboto]

That's the part that makes me laugh. Does it need greater than 10 kw of power?

It's already possible and common to run 1000's of watts of power worth of electronics in addition to everything else that comes standard on vehicles now-a-days. Most of the sensing equipment is already available on many if not most ICE vehicles. That leaves mostly the computing unit. Does that actually require multiple 1000's of watts of power?

It may or may not be easier, but it's certainly a selling point to attempt to convince enough people to buy an electric vehicle to meet potential future government requirements on percentage of a fleet of sold vehicles being all electric. That's not a problem for an ICE vehicle.

That's what most people want to use and are comfortable using, so self driving isn't exactly a selling point that's going to move appreciably more ICE vehicles. Especially when faced with the more pressing need to convince your average joe schmoe consumer that they should spend significantly more money on an EV versus a much cheaper ICE vehicle. And that's where at least some form of limited self-driving capabilities might convince someone to buy an EV instead of an ICE vehicle.

BTW - In 2019 an ICE vehicle (a freight truck) using self driving AI drove across the US.

A Self-Driving Freight Truck Just Drove Across the Country to Deliver Butter

It made the daunting 41-hour trip from Tulare, California to Quakertown, Pennsylvania.

www.popularmechanics.com

We hear more about EVs, however, because it's easier to get government grants to assist with EV R&D for SDVs.

Regards,
SB

I've never built an AV. But the same patterns crop up. I get that yes, on paper an alternator can produce the power you are looking for. But I've yet to see an article that specifically states that the modern ICE car can take on AV upgrades easily without much of an expense. As I said earlier, I don't believe in some grand conspiracy theory here. There are many autonomous companies out there that aren't car companies. So the only reason is that either there is no demand for it on ICE vehicles or they are economically unfeasible or disadvantageous in modern car manufacturing to take it on. I've cited several reports now all pointing to similar reasons here. I KNOW very well that AVs started as gas cars for testing and prototyping, but that is NOT the same as FSD AV as feature being sold to the public on a gas car. There are very few exceptions to that and only a tiny fraction of gas cars out of the many millions of models that actually support it.

Electric Cars Make Excellent Autonomous Vehicles—So Why Are Most AVs Gas-Powered?

These engineering students are making an autonomous electric car.

www.vice.com

Unlike gas cars, most EVs come with “drive-by-wire” functionality, which allows people to run a car from a laptop using a USB cable. This, combined with the fact that EVs have bigger and more powerful batteries to help run power-sucking advanced driver-assistance systems, makes them ideal bedfellows for autonomous technology.
“Powering sensors and computers is much easier when you have access to a high-current power source, which is readily available in electric cars. In a gasoline car, you would try to power everything off of the battery/alternator, but this just isn't enough power,” Burnett explained.

Because most gasoline cars aren’t drive-by-wire, autonomizing them involves expensive, bulky, and heavy upgrades that can cost upwards of $100,000, Steven Waslander, an associate professor at the U of T Institute for Aerospace Studies (which also does autonomous-car research), told me on the phone. “Electric cars would be easier and less expensive to upgrade,” he said. “It’s literally just exposing the interface to allow control over [acceleration], steering, and braking.”

Another Source from another AV Company:

From EVs to AVs. | Ambarella

Electric vehicles are playing a significant role in driving the shift to autonomy. The past five years have seen an extraordinary investment toward the

www.ambarella.com

here are a number of reasons why AVs are typically EVs.​

• Cost per-mile. Current AV business models center around the deployment of robotaxis and autonomous trucks—for fleet applications like these, cost per-mile is critical. Today, a majority of these costs are related to driver/operator expenses, but under a driverless AV business model, the major costs are related to two factors: fuel and maintenance, both of which represent tremendous advantages of electric vehicles.
  • In the summer of 2019, for example, the average cost of a gallon of gas in the U.S. was approximately $2.80, while over that same period, the average cost per kWh was around 15 cents. Direct comparisons are difficult, but it is generally accepted that the cost to power an EV is approximately 50% less than the cost to power a gas-powered vehicle.
  • As for maintenance, EVs have the advantage here as well. With no internal combustion processes, lower temperatures, less corrosion, and fewer scheduled services, the average maintenance savings for the current generation of EVs on the road is nearly $1,000 per year.
• Complementary technologies. The strengths of AVs offset the weaknesses of EVs—and vice versa. For instance, AVs have the potential to drive more efficiently, which has the potential to increase an EV’s range. In turn, most EVs are built with drive-by-wire systems that replace traditional mechanical control systems with electronic controls, which are more compatible with an autonomous approach.
• Simpler integration. Drivetrains in conventional vehicles can include as many as 2,000 parts, while electric drivetrains are composed of fewer than twenty. Fully electrified vehicles have no need for components such as radiators, fuel tanks, exhaust systems, or multispeed transmissions. Simply put—implementing autonomous features on EVs is more straightforward because of the base vehicle’s minimalist design

Another Source: https://www.embedded.com/industry-bets-on-avs-and-evs/

The EV’s foremost virtue is a lot fewer moving parts. Its three major components are battery, inverter and electric motor. In contrast, an internal combustion engine contains thousands of tiny pieces that must be maintained. From an engineering point of view, the EV matches AV technology perfectly because an AV needs more electrical brainpower to manage vision, sensor fusion, mapping and path-planning functions while processing an ever-growing volume of software.

 

[orangpelupa]

Maybe for more context, according few users that share their experience, comma ai / open pilot works great with ICE cars.

The computing power only in a smartphone.

It already more consistent/predictable than tesla autopilot and fsd.

It have way less feature domain than fsd tho.


Unfortunately open pilot still need a dedicated hardware (a phone) to work. It cant be installed into car's on board computer. It also cannot reuse the cameras already in the car

 

[iroboto]

Maybe for more context, according few users that share their experience, comma ai / open pilot works great with ICE cars.

The computing power only in a smartphone.

It already more consistent/predictable than tesla autopilot and fsd.

It have way less feature domain than fsd tho.


Unfortunately open pilot still need a dedicated hardware (a phone) to work. It cant be installed into car's on board computer. It also cannot reuse the cameras already in the car

Stretch to call them the same. FSD reads street signs and speed bumps, identifies pedestrians, motorcycles etc. It’s does full automation with traffic A to B. Adaptive cruise control with lane centring is among the most basic feature of autonomy and can be found in a lot cars prior to the teslas coming out.

 

[Albuquerque]

Yeah, from what I'm able to tell iroboto is exactly right: the OpenPilot project is mostly gearted to intelligent lane keeping and cruise control. Major callouts include no current support for navigation-following functions, many cars cannot support turning while also stopping, many cars also limit how much the steering wheel can be turned, it cannot read road signs (speed limits, stops, yields) and seems to have an even worse time at roundabouts than Tesla's 10.x build release of FSD.

There was a video published yesterday of Tesla FSD driving in broad daylight and heavy traffic from a residential street in Los Angeles to a residential street in San Francisco -- six whole hours of real time video posted -- and the driver had to intervene once to go through the tool booth on the SF Bay bridge. Here's the time-lapse version of it:

Not quite the same.

 

[iroboto]

Yeah, from what I'm able to tell iroboto is exactly right: the OpenPilot project is mostly gearted to intelligent lane keeping and cruise control. Major callouts include no current support for navigation-following functions, many cars cannot support turning while also stopping, many cars also limit how much the steering wheel can be turned, it cannot read road signs (speed limits, stops, yields) and seems to have an even worse time at roundabouts than Tesla's 10.x build release of FSD.

There was a video published yesterday of Tesla FSD driving in broad daylight and heavy traffic from a residential street in Los Angeles to a residential street in San Francisco -- six whole hours of real time video posted -- and the driver had to intervene once to go through the tool booth on the SF Bay bridge. Here's the time-lapse version of it:

Not quite the same.

Yea. I’m not a staunch defender of Tesla. Just trying to put into perspective the real challenges poised by full FSD running at regular human speeds, adapting and adhering to changing conditions etc.

All of it very challenging and a tiny cellphone SoC has no chance at being able to do it.

 

[Albuquerque]

You and I both. Tesla has made some pretty lofty statements about how far they think their FSD is going to get; I'm rather curious to see how they're going to make good on it all. Still, Tesla is inarguably further along (and substantially so) than anyone else in the consumer space right now, even if FSD still isn't ready for prime time nor capable of delivering half of what they've promised.

Our Model Y Performance should be here by the end of October (lol until it's delayed another two weeks again) and I was one of those suckers who added in FSD. You know why? Because at least six times per year we drive to Lousiville KY and I wager a free cup of coffee it can drive itself from my house to my parents-in-law's house without breaking a sweat. Can I do it by myself without having to rely on a fancy shmancy playtoy? Yeah, I can and I do. You know what though? Letting the car do it for me after I've worked a nine hour day sounds like an absolute lifesaver.

 

[orangpelupa]

Stretch to call them the same. FSD reads street signs and speed bumps, identifies pedestrians, motorcycles etc. It’s does full automation with traffic A to B. Adaptive cruise control with lane centring is among the most basic feature of autonomy and can be found in a lot cars prior to the teslas coming out.

That's why I specifically say open pilot have way less feature domain.

Also not sure they have added more traffic signs recognition or not. Iirc they works with traffic lights and stop signs but nothing else.

 

[wco81]

Saw this quoted from a WSJ paywalled article.

Toyota said hybrids, including plug-in cars, accounted for around 20% of U.S. sales in September, but has said that could have been double if not for supply constraints.

Overall, hybrid vehicles represented 5.6% of U.S. sales through September, while EVs were 4.9% of the market, according to data from research firm Motor Intelligence.

“If it has a hybrid engine on it, it has a gold star in the customer’s mind,” said Adam Simms, CEO of California-based dealership group Price-Simms Inc. He estimates that he could sell 6,000 more hybrids if he had the supply."

Hybrids are the most popular type of cars selling now, more than EVs and far more than ICE cars.

U.S. car buyers on average have been snapping up hybrids within just 12 days from their arrival on dealership lots this year through August, compared with 16 days for EVs and 23 days for internal-combustion vehicles, according to data from consumer-research site Edmunds. com.

 

[orangpelupa]

and toyota planned to bring hybrids to indonesia in 2023.

the road to EV is long

 

[Sxotty]

You and I both. Tesla has made some pretty lofty statements about how far they think their FSD is going to get; I'm rather curious to see how they're going to make good on it all. Still, Tesla is inarguably further along (and substantially so) than anyone else in the consumer space right now, even if FSD still isn't ready for prime time nor capable of delivering half of what they've promised.

It seems weird that your say inarguably when they're are plenty who will argue against your statement in the press and in society.
Example and the mapped roads have doubled since that review.

If you said that Tesla allows it to behave more riskily and operate in more conditions people would not argue. Super cruise is just as good as FSD IMO and many say it is better, however it is much more limited in what it is allowed to do. Tesla keeps removing sensors to increase profit margins.

 

[iroboto]

Letting the car do it for me after I've worked a nine hour day sounds like an absolute lifesaver.

So I've been running enhanced Autopilot since 2017 and it's certainly gotten much better. In general stay aware of the roads as though you are the front seat passenger and take over whenever you feel like driving defensively and you'll be safe 100%. Really helpful when you're tired because it takes no energy to watch the car drive, but it takes significantly more to actually steer and drive yourself.

If you said that Tesla allows it to behave more riskily and operate in more conditions people would not argue. Super cruise is just as good as FSD IMO and many say it is better, however it is much more limited in what it is allowed to do. Tesla keeps removing sensors to increase profit margins.

Yea, it's certainly more dangerous to run the tesla variants, mainly because they allow you to do well, anything really. Which also leads to Tesla's biggest advantage in getting to the higher automation tiers earlier, they have the data on it, unfortunately we are at risk for it.

I think driving assist (level 2) is pretty good for most people who are very capable of driving, most probably don't need full FSD (though I would say I've seen some really awful drivers that I'm sure I would prefer there to be a tesla FSD driver instead). But tesla's autopilot still does highway on ramp to exit ramp with navigation on its own, even without the full FSD package. Depending on what you're trying to do, this may still be a desirable feature here. I've never had full FSD do road driving, but it wouldn't bother me if I was doing a long trip like Albuquerque is doing here, sometimes the roads are just empty and you're not in a rush to go anywhere, let it take over and rest your eyes and conserve that energy.

 

[iroboto]

edge cases for AV are pretty nuts; watching the waymo deal with emergency vehicles and a car crash was pretty good moment. I would like to see the Tesla's identify road debris. There was a moment on the on ramp I thought there could be a hit

nvm, looks hella tight though. It works mediocrely right now lol. Don't trust it!

 

[Sxotty]

Depending on what you're trying to do, this may still be a desirable feature here. I've never had full FSD do road driving, but it wouldn't bother me if I was doing a long trip like Albuquerque is doing here, sometimes the roads are just empty and you're not in a rush to go anywhere, let it take over and rest your eyes and conserve that energy.

Remember you are definitely not allowed to rest your eyes.

Stay safe

And I freely admit that having all the automatic bits is safer, but it is not as capable as tesla let's it try to be and the user is responsible for being aware and paying attention. Personally all I want is on ramp to off ramp and many systems are pretty capable for that but none allow them to be used like that without driver paying attention.