Problematic range estimations outside of cell service areas

[e-FTW]

These last two weeks, we took the car for two road trips in National Parks in CA: Lassen Volcanic and Yosemite.
In both cases, when planning the drive back from the park, the nav/energy graph initially showed a solid arrival charge near 20% at the Superchager.
But at some point in the case of Yosemite, while descending from the mountains the predicted arrival charge level was adjusted drastically. For Lassen, it forced us to hypermile a bit. For Yosemite, it forced us to add a stop.
In both cases, the initials route planning was done in the nav while out of cellular connectivity. Am thinking that once the car got online, it adjusted the metrics based on some unknown factor. And it clearly needed to do that, as the initial estimations were clearly off in hindsight.

This is the result for Yosemite:

As you can see, about 10 miles down the mountains, the grey bar suddenly went off the screen as if it suggested we should have left the Fishcamp Supercharger at something like 150%...
(We could still probably have made it home, but a lunch stop was in order anyway)

In Lassen, it predicted we would arrive at Corning SC with 20% charge after leaving from Lassen Peak (plenty of regen to be had). We got there at 7% by hypermiling it as the graph’s lines were parting ways like they had a big falling out. No sudden reset like Yosemite, just a clearly unattainable goal.

So initial calculations were way, way off. Is this something you have experienced? What would be the best channel to report this to Tesla (I did not use the bug reporting feature in the car at those times, and I should have I guess)?

 

[Tam]

...the grey bar suddenly went off the screen as if it suggested we should have left the Fishcamp Supercharger at something like 150%...

This is what I understand that it has nothing to do with cell coverage but current consumption projecting for the current trip.

The grey chart went up because it senses the regen. Instead of consuming energy, your car was now producing energy.

When your car had 86% but the grey went all the way up 100%, it was guessing if your car kept abstaining from using energy, then it would credit you as if you got 100%.

It's a predictor based on your driving conditions and your current driving condition was adding energy. It guessed that if this condition keeps ongoing, you should arrive at the destination with about 60% left.

But of course all good things must end and your will consume energy at some point and its guess was just wrong.

...What would be the best channel to report this to Tesla (I did not use the bug reporting feature in the car at those times, and I should have I guess)?

You can send a message through your Tesla Account page.

 

[e-FTW]

This is what I understand that it has nothing to do with cell coverage but current consumption projecting for the current trip.

The grey chart went up because it senses the regen. Instead of consuming energy, your car was now producing energy.

When your car had 86% but the grey went all the way up 100%, it was guessing if your car kept abstaining from using energy, then it would credit you as if you got 100%.

It's a predictor based on your driving conditions and your current driving condition was adding energy. It guessed that if this condition keeps ongoing, you should arrive at the destination with about 60% left.

But of course all good things must end and your will consumes energy at some point and its guess was just wrong.

You are right, although the nav should know pretty well how much regen I am about to get (see the bump up to 86%). It did at all other times, except these two examples.
My point is the actual arrival charge was way off in both cases, grey bar or not. For Yosemite, the green line got reset much lower than it was at the outset.

 

[Tam]

...actual arrival charge was way off in both cases...

Tesla is using EPA ratings so please notice the theoretical and actual differences to project your future trips.

Unless of course,if you can drive like EPA does as graphed below with the average speed of 48.37MPH for the High Speed Test. (It's high because it topped at 80MPH) and only for a distance of 8.01 miles:

Detailed Test Information

 

[e-FTW]

Tesla is using EPA ratings so please notice the theoretical and actual differences to project your future trips.

Unless of course,if you can drive like EPA does as graphed below with the average speed of 48.37MPH for the High Speed Test. (It's high because it topped at 80MPH) and only for a distance of 8.01 miles:

Detailed Test Information

Yup, aware of that. And I drive the least efficient version of the 3.
But no other segments or trips had these types of issues.

 

[AlanSubie4Life]

It's a predictor based on your driving conditions and your current driving condition was adding energy. It guessed that if this condition keeps ongoing, you should arrive at the destination with about 60% left.

This does not make much sense to me. The predictor is known to take elevation gain/loss into account. In addition it does take current efficiency into account as you say. Maybe that starting point of efficiency was way off. But the overall result here seems buggy.

I’m not sure what happened - whether it was an issue with lack of connectivity (seems possible if the data is not fully cached but I thought it was?), or whether it is some other bug, I don’t know.

Would be interesting to see if others have seen similar results. Pretty sure @Perry has but he is a very special case.

 

[gecko10x]

Did you take a different route than initially planned?

 

[Trekboy]

Too bad the results couldn't have been compared to "A Better Route Planner".

 

[CharleyBC]

The meaning of the green line in the graph is easy, but what does the grey line mean?

 

[Tam]

...This does not make much sense to me...

The meaning of the green line in the graph is easy, but what does the grey line mean?

I am not sure but my guess is:

The green and yellow line is the predictor calculated at the start of the trip that includes many factors including elevations but not driver's driving habit (such as fast or slow or lead foot...). It's static.

The grey is the predictor calculated at the present time. It's dynamic depending on what your car is doing right now. When you go downhill, it would readjust itself as if you have 100% charge and would predict that you would still have lots of charge at the destination like 60% but that's false because the car will go uphill sometimes too.

 

[e-FTW]

Did you take a different route than initially planned?

I did not.

 

[AlanSubie4Life]

I did not.

It looks like you do have custom wheels & possibly sticky tires (wasn't able to identify the dimensions from your picture other than they are 19" Michelins).

It's possible if those are a lot stickier than stock it could also disrupt the estimate. That's been reported elsewhere, though it's weird that the algorithm can't deal with it reasonably.

In any case, it still doesn't seem like that difficult a problem for the car to be able to keep track of the basic vehicle efficiency and topography and get VERY good estimates of the range (notwithstanding rain, wind, etc. - though those potentially could be dealt with in the future but would require internet access of course).

 

[e-FTW]

I am not sure but my guess is:

The green and yellow line is the predictor calculated at the start of the trip that includes many factors including elevations but not driver's driving habit (such as fast or slow or lead foot...). It's static.

The grey is the predictor calculated at the present time. It's dynamic depending on what your car is doing right now. When you go downhill, it would readjust itself as if you have 100% charge and would predict that you would still have lots of charge at the destination like 60% but that's false because the car will go uphill sometimes too.

The green line is what you are actually doing, the grey line is the prediction.
Typically, the green line is pretty close to the grey one, and adjusting one's speed allows you to underperform/overperform depending on what you want. It is really handy!
I have used this graph since 2015 with my S and have not seen a behavior like these last two trips here, ever.

To further clarify, the shape of the green line never changed, so the car knew about elevation changes and regen and speed limits. What suddenly changed is the arrival charge % dropped a lot when the grey line went offscreen. That is when we realized that the prediction was way off, and we would have to make another stop so that we/my wife would be comfortable with our expected arrival charge.

 

[AlanSubie4Life]

To further clarify, the shape of the green line never changed, so the car knew about elevation changes and regen and speed limits. What suddenly changed is the arrival charge % dropped a lot when the grey line went offscreen.

Kind of hard to understand how this looked without the picture of the "before" - though obviously you wouldn't have been thinking it would be necessary to take such a picture. If the green line never changed it's weird that the final estimate changed a bunch. But I believe you!

Looks like nice light traffic out of Yosemite. Good times.

 

[e-FTW]

It looks like you do have custom wheels & possibly sticky tires (wasn't able to identify the dimensions from your picture other than they are 19" Michelins).

It's possible if those are a lot stickier than stock it could also disrupt the estimate. That's been reported elsewhere, though it's weird that the algorithm can't deal with it reasonably.

Yes, I have downsized to lightweight 19-inch wheels but have slightly wider tires (245 vs 235) although they are all-seasons and likely less sticky than the 20-inchers that came with the car. Thats seems unlikely to be that big of a problem as other segments of these two road trips did not suffer from such events.
One segment on I-5 with lots of head wind and a 70mph speed limit understandably did show my green line dropping bellow the grey as expected. I do not believe Tesla takes whether into account yet. And I did not have my browser set to Tesla Winds and Elevation Web Browser App as I should.

 

[e-FTW]

Kind of hard to understand how this looked without the picture of the "before" - though obviously you wouldn't have been thinking it would be necessary to take such a picture. If the green line never changed it's weird that the final estimate changed a bunch. But I believe you!

Looks like nice light traffic out of Yosemite. Good times.

When we started the drive, both lines matched up. So you could really only see the green one. And it was green all the way, as it predicted we would arrive well above 20%.
So look at the picture above and remove the dark area above the green line, and boost the arrival charge to something north of 40% IIRC.

 

[AlanSubie4Life]

Thats seems unlikely to be that big of a problem as other segments of these two road trips did not suffer from such events.

Yeah, I have no explanation for the behavior other than some sort of connectivity issue as you suspected. Doesn't make any sense that it would be so wildly off, but still...not surprising...

So look at the picture above and remove the dark area above the green line, and boost the arrival charge to something north of 40% IIRC.

I see. My guess is the scale on the left hand side actually must have changed as well (otherwise the arrival charge would not be able to change). So basically the green line actually changed a LOT while the gray line stayed the same. But it looked the way you describe because of the scale change. It will rescale the plot to keep the basic plot of charge vs. distance traveled looking similar at all times I suppose.

EDIT: Though, the gray line is indicating arrival at about ~60% SoC. So that doesn't quite line up either. I guess it changed too. In any case, the green line must have changed otherwise the arrival charge would stay the same as whatever it was originally.

 

[e-FTW]

Yeah, I have no explanation for the behavior other than some sort of connectivity issue as you suspected. Doesn't make any sense that it would be so wildly off, but still...not surprising...



I see. My guess is the scale on the left hand side actually must have changed as well (otherwise the arrival charge would not be able to change). So basically the green line actually changed a LOT while the gray line stayed the same. But it looked the way you describe because of the scale change. It will rescale the plot to keep the basic plot of charge vs. distance traveled looking similar at all times I suppose.

EDIT: Though, the gray line is indicating arrival at about ~60% SoC. So that doesn't quite line up either. I guess it changed too. In any case, the green line must have changed otherwise the arrival charge would stay the same as whatever it was originally.

I see what you mean. The green line's shape is what never changed, matching with the route's elevation changes and speeds.
But I think its angle changed to drop to a much lower arrival number suddenly.

Of course, if I had printed (!) out some routes on pick-your-trip-energy-planner, I would have spotted the overly optimistic initial numbers from the car.
But the result would be have been the same: what the heck did this screen do? And why?

 

[AlanSubie4Life]

what the heck did this screen do? And why?

Have no answers for you. Seems buggy. I've seen the same thing on a more minor scale when I was in a Tesla with limited connectivity. Step change in the gray line.

The solution is to do all range calculations including net elevation changes in your head, including a simple model for velocity as well.

 

[e-FTW]

Have no answers for you. Seems buggy. I've seen the same thing on a more minor scale when I was in a Tesla with limited connectivity. Step change in the gray line.

The solution is to do all range calculations including net elevation changes in your head, including a simple model for velocity as well.

Sure, but what if I am singing show tunes and chewing gum?

(Nice avatar btw, mine is also Subie)