Supercharger routing to avoid waits

[Simavon]

Yesterday we collected some people from Luton. It’s a 340 mile round trip and the ABRP plan was for two charges on the return.

Tesla decided we could do one stop at Membury Westbound. But we stopped for a coffee break at Reading Services and then it decides to direct us back to Wokingham. We decided this wasn’t a good idea but realised on the way to Membury that it was because of no availability of chargers with a “Medium wait”. As we got further along it changed to Membury Eastbound as recommended stop which was going to add probably 15-20 miles. By then it had become “Long Wait”

In the end we got to Membury and the chargers were barely 50% occupied. Would have been pretty wasteful and times consuming to have done as told. And we would not have known.

Anyone else observed this? Obviously the wait time is a lagging indicator but is it accurate? It’s possible that the directing of cars away from there reduced the wait to zero which does end up doing load sharing, though.

 

[Kansalis]

Anyone else observed this? Obviously the wait time is a lagging indicator but is it accurate? It’s possible that the directing of cars away from there reduced the wait to zero which does end up doing load sharing, though.

It's possible the wait time calculation is only done when you start the trip, or doesn't update very often. It's not beyond the realms of possibility that all the chargers were full 10 minutes before you got there, so the nav may have been correct at the time. Most SuC charging is quite quick, so turnover on the chargers doesn't take long. Unless of course there's a long queue of cars already.

I suppose this will always be a guessing game until they implement some kind of queueing system for the chargers. It would be quite easy for them to add a realtime queue with a "you're number 3 in the queue" type system for cars within in a certain range of the chargers, who have set the charger as their destination. The nav knows with decent accuracy when you'll arrive, so it would be almost trivial to have a fair system of prioritisation programmed into the process. Unfortunately that would all go out the window if/when the network is opened to other cars unless you had to sign up to the app in some way to register. I'm quite surprised they haven't implemented something like that already to be honest.

 

[WannabeOwner]

Might have been that all the cars heading there were part of the "Long wait" prediction and, as a result, they all went somewhere else?!

I'm quite surprised they haven't implemented something like that already to be honest.

What about people who turn up at the Charger without having put it in their SatNav - perhaps they don't want the pre-condition energy usage and/or cost

 

[Overworked]

… they dont get a place in the queue until they notify arrival by setting it as their destination and then it gives them a place.

 

[nufan]

Last weekend I noticed charger occupancy updates on the nav were only happening infrequently - perhaps only when I clicked the Supercharger ‘info card’ up. Data seemed ‘sticky’. Not sure if real or imagined. If real - would explain what you saw….the Long Wait was just old news.

 

[Simavon]

Might have been that all the cars heading there were part of the "Long wait" prediction and, as a result, they all went somewhere else?!



What about people who turn up at the Charger without having put it in their SatNav - perhaps they don't want the pre-condition energy usage and/or cost

Well indeed one could foresee a feedback loop forming where cars heading for the westbound charger diverted to the eastbound. If enough vehicles do this the state flips. They are then all told to head back to westbound. Repeat until all batteries are flat.

Of course we realised if we went to the westbound charger and it was a disaster we wouldn’t have lost anything anyway as we could just rejoin the motorway to loop back to the eastbound. Unless the queue was so bad the services were gridlocked.

 

[Wol747]

I assume that Fremont has access to the same data as in your car, so when plugged in and the screen showing "XXX minutes remaining" it would seem a fairly trivial coding to be able to show on everyone's screens what the actual occupation would be YYY minutes in the future, and/or prioritise cars enroute to charge.

 

[WannabeOwner]

… they dont get a place in the queue until they notify arrival by setting it as their destination and then it gives them a place.

Complicated (not to you or me ...)

the screen showing "XXX minutes remaining"

I almost never leave then

 

[Neilio]

I assume that Fremont has access to the same data as in your car, so when plugged in and the screen showing "XXX minutes remaining" it would seem a fairly trivial coding to be able to show on everyone's screens what the actual occupation would be YYY minutes in the future, and/or prioritise cars enroute to charge.

This won't work if people charge correctly for long journeys. Unless you actually set the precise SoC you need to leave every single time. I often just pop in for a splash and dash. I use ABRP to calculate the best trip so the car doesn't know when I'm leaving

 

[WannabeOwner]

Although ... with enough data Computer will be able to figure out trends. It knows how many people are parked at that spot ... no way to tell who is waiting for a charge, and who not (although someone who has finished charging and moves to nearby could be assumed to be done ... but I've charged, parked up ,and then moved back to charge from 90% to 100% before departure).

But half a dozen cars milling about is pretty difficult to get a guestimate for, although tens of thousands, across all superchargers, would provide the data for "trends".

I'm still sceptical this would work in practice. Tesla have much bigger wait problems in California, especially on American holidays, and sales there have been huge, compared to here, given that new models come out there years before RHD is available here. And yet Tesla haven't done anything there (they do implement cheap rates out-of-core-hours on Holiday days though). As folk say the software engineering solution appears trivial ... I reckon there is a reason why they haven't done it, and complexity for the user and lack of sufficient numbers (at individual Superchargers) may mean that the prediction has been found to be hopeless.

Tesla can, of course, "shadow" test this, as with AP. Try to predict waits and queueing and see how that compares with reality. Maybe they then decided that "Medium" / "Long" wait messages was the best compromise solution.

Along time ago I had a mate who did software for traffic light linking in cities. "How did you test a new hypothesis?" "We put it live and went out into the streets to see how much more/less horrendous the tailback was"

 

[Simavon]

Indeed part of the problem with this information was we had no idea what these medium and long waits actually were. A number would be informative but as already said, the validity of the prediction may well be suspect. And “waits are long” does not necessarily equate to “your wait will be long”. I don’t think it’s remotely trivial.