[JoeReal]

They consume the equivalent of 50-100 laptops of today. Sensors and processing power of redundant systems in autonomous drive systems consume 2-4 kW of electric power! Surely the next priority improvement of efficiency would be for the processing chips and sensors for autonomous driving.


https://www.bloomberg.com/news/arti...-are-giving-engineers-a-fuel-economy-headache

 

[srmarti]

So has any Tesla owner noticed a range difference when driving manually vs autopilot? Or does more efficient driving compensate for the extra energy required for processing?

 

[JoeReal]

So has any Tesla owner noticed a range difference when driving manually vs autopilot? Or does more efficient driving compensate for the extra energy required for processing?

I don't know about Tesla's power consumption for autonomous drive, but they have far fewer redundant systems and sensors than what's used in autonomous Chevy Bolt, so I expect them in the low end of power consumption.

A 2-4 kW overhead consumption can reduce the range by 30-60 miles for the Chevy Bolt.

 

[saghost]

So has any Tesla owner noticed a range difference when driving manually vs autopilot? Or does more efficient driving compensate for the extra energy required for processing?

Not sure you'd see the difference. Tesla has the AP sensors and computers running at all times for safety features - might or might not need the same amount of processor time.

Also note that NVidia is launching a SoC solution for self driving, which should greatly reduce the computer power consumption.

 

[JoeReal]

Not sure you'd see the difference. ...

Does anyone have some energy consumption data to share between the autonomously driven Chevy Bolt and those that aren't? It would truly be a good comparison.

 

[viking79]

So has any Tesla owner noticed a range difference when driving manually vs autopilot? Or does more efficient driving compensate for the extra energy required for processing?

The current Tesla system does not have near enough computing power for a self driving system. The current AP2 hardware uses NVIDIA Drive PX 2 I think, which is pretty low power, maybe 300 watts? The hardware powering the Autonomous Bolt EV (which is actually autonomous, today) is much higher than that, and the computing power takes up almost the entire hatch back area based on their graphic showing where the autonomous systems are added.

A draw of 2-4kw is a lot, but about half of what the electric heater draws, so it isn't that big. It will definitely shorten the range some, but it isn't like it is a huge power consumer, and I would expect that to drop by a factor of 5 or 10 in a few years. The new NVIDIA Pegasus system looks pretty impressive in terms of power as well. The fact that a Bolt EV has a 60 kwh rated battery gives it a huge advantage for autonomy vs a an ICE as it can easily provide reliable power for the self driving hardware.

If you want to see a huge consumer of power, look how much energy it takes to process a single bitcoin transaction...

 

[DonC]

Meh. The article completely mistakes the CAFE standards by confusing the old EPA standard with the new. For CAFE purposes the EPA uses the old standard. That would be a much higher number than the new standard, which is based on more challenging drive cycles. So for CAFE purposes the Honda Fit would likely need to hit something like 45 MPG rather than the 59 MPG the article is suggesting. (Not sure of the exact numbers here but I'm not writing the article).

In any event, this is why Mark Reuss has said that electrics are the most obvious route to autonomous vehicles. Not only do their MPGe numbers obliterate the targets, the larger batteries allow for more power to be used for running the electronics.

And yes, power consumption will come down. Big time. And battery capacity will go up.

Not sure you'd see the difference. Tesla has the AP sensors and computers running at all times for safety features - might or might not need the same amount of processor time.

AP is not remotely an autonomous system. You need more sensors, and likely Lidar and maps, to get to autonomy. The data and computing demands are going to be high.

 

[saghost]

AP is not remotely an autonomous system. You need more sensors, and likely Lidar and maps, to get to autonomy. The data and computing demands are going to be high.

I didn't say it was. I was responding to a post that asked if Tesla owners saw a difference in consumption when running in AP and not...

 

[edk-austin]

I have to wonder if 2-4KW is a peak power, which might include power steering. Or is an increment to a typical ICE car without big bright LCDs. In which case, maybe there's not a big increment over Tesla or V/Bolt already.

Or the 2-4KW is estimated from increased air drag at speed...

 

[DonC]

I didn't say it was. I was responding to a post that asked if Tesla owners saw a difference in consumption when running in AP and not...

Oh. Didn't understand that. I'd be surprised if they did. I have a friend who is putting in an AP like system for the Volt. When I asked him about the power draw, he said it wan't a big deal.

 

[Norton]

I have to wonder if 2-4KW is a peak power, which might include ....
Or the 2-4KW is estimated from increased air drag at speed...

This does seem like a lot of power to run cameras, 2 Lidar units and whatever small sensors there might be.
Maybe this test vehicle is loaded down with multiple 'Flight Test' PC's gathering data and running beta OS for the sensors.
And a 120 VAC inverter system to power the PC's.

 

[hellsop]

I have to wonder if 2-4KW is a peak power, which might include power steering. Or is an increment to a typical ICE car without big bright LCDs. In which case, maybe there's not a big increment over Tesla or V/Bolt already.

Or the 2-4KW is estimated from increased air drag at speed...

Doesn't really matter. The power steering (and power brakes and electronic transmission shifting and all the rest) are already THERE and being used. So we're only really talking about the power required to actually process outside conditions into inputs to the control systems that already exist as is.

 

[srmarti]

Does anyone have some energy consumption data to share between the autonomously driven Chevy Bolt and those that aren't? It would truly be a good comparison.

Don't think any of those are in the hands of the general public. It's not truly autonomous but Tesla data should be available and might tell something. A system still in development might be an energy hog. but might be substantially more energy efficient by the time it's in regular production.

 

[Loboc]

It is likely that an autonomous vehicle would drive more like a hypermiling human. This could potentially save 1/3 to 1/2 of the power a normal driver would demand. In addition, reducing collisions (aka 'accidents') would keep traffic moving and use less power.

So, net zero to run the bigger computers and sensors.

 

[DonC]

It is likely that an autonomous vehicle would drive more like a hypermiling human. This could potentially save 1/3 to 1/2 of the power a normal driver would demand. In addition, reducing collisions (aka 'accidents') would keep traffic moving and use less power.

So, net zero to run the bigger computers and sensors.

Good point I hadn't thought of.