[FIXED B729] Why use only one tanker for refueling?

[wild_Willie2]

Hi there,

I am just wondering why all AC will try to refuel at the same tanker while there are three other tankers flying in the same formation? They are all waiting their turn on this single tanker and this takes a lot of time, especially if multiple groups try to refuel at the same time. Can somebody please explain me the logic behind this?

W.

[Rongor]

Just a wild guess: they choose the one with the least remaining time on station available?

[mikeCK]

Maybe it's a simple as each aircraft chooses the closest refueler. Doesn't matter if there are 15 planes there because each plane acts independently?

[Dimitris]

Save?

[wild_Willie2]

Here you go, a save where 61 Hornets are all trying to refuel from the same tanker (there is a cue of 60 Hornets for a single tanker) while there are 24 other tankers available in the tanker formation.

W.

[BradOrbital]

Reminds me of my local fuel station whenever the fuel price rises ha!

[wild_Willie2]

If you live in the US you really cant complain about gas prices, in the Netherlands we pay 1.60€ per liter, this would be $6.64 per gallon for you Yanks

W.

[VilleYrjola]

HA, I thought that Finland was only place with over 1,6 euros per liter.

[MrOrange]

Italy.
Welcome to the club. [:)]

[Sakai007]

lol, yep i am all done whining about gas prices, thanks for that! Gonna bring this up too the next time someone complains about the gas prices around these parts.

[mikmykWS]

Buy a hybrid [:)]

Mike

[Dimitris]

Okay, looking at this now.

First thought: A 61-AC Hornet group? [X(][X(][X(]

Second thought: A 25-AC tanker group? [X(]

I broke up both groups, ordered all Hornets to refuel and sure enough they dispersed and queued up after multiple tankers.

Nothing stops you from pushing the refueling mechanics & AI past the edge of sanity, but don't complain if the sim throws it hands in the air when you do.

Note: This doesn't mean the tanker-selection logic cannot be improved; I'll elaborate below.

[Dimitris]

Okay, a bit of backtrack and history is necessary.

When we first worked on the refuelling priority logic, we went with the simple round-robin principle: Load-balance incoming thirsty aircraft between available tankers. Nice and simple.

This worked fine if all you had was a single tanker orbit area. It became problematic in situations where you have multiple tanker orbits widely separated (e.g. in the scenario "Shamal" IIRC). Planes started wandering off to far-flung tankers simply because these were less occupied, often going wildly off course. Carefully planned ingress legs degenerated to chaos. Bad.

So then we switched to a scheme which gave much more emphasis on tanker proximity. Refuelling queue size still mattered, but now planes were much more reluctant to go out of their way to hook up with a less busy tanker. This has worked much better so far, and is reasonably graceful against tanker withdraws (if a tanker RTBs, it automatically is removed from all best-pick calculations).

However, as your test setup demonstrates well (the jaw-dropping monstrosity of 60-plane groups notwithstanding), the existing logic is vulnerable to abrupt surges in refuel requests. Once I broke up the plane groups and ordered all Hornets to refuel at once, they scattered about and picked the less busy tankers in proximity, as expected. However, the nearest tankers were overwhelmed and outliers (in distance) were not utilized enough. The result was that a few of the Hornets ran out of fuel before they could be serviced.

So I added an extra wrinkle, to prioritize by estimated waiting time. So now when the thirsty aircraft asks around for available tankers, each candidate tanker returns an estimate of how much time will pass before it is able to start refuelling the requester. By necessity this estimate is crude (e.g. each "client" on the existing refuelling queue is assumed that it will receive a full top-up load, the connection type and thus transfer rate [and thus time to refuel] is a best-guess that may or may not hold true at the actual hookup etc.) but still provides a reasonable prognosis. This figure is also compared with the estimated time to actually get to the tanker; a faraway tanker may be very busy right now but may have finished servicing its queue by the time the requester gets there.

Applying this logic to the test scenario, all Hornets made good tanker picks and all tankers were utilized. No aircraft ran out of fuel.

Added in Build 729.

[wild_Willie2]

Thanks for the fix, this will make refueling of medium to large groups of AC go so much smoother in game.
The test scenario was specifically designed to show the scope of this issue, therefore I used a group of 60 AC v. a group of 25 tankers.
Although this scenario would never happen IRL., it did show the problem clearly.

Keep up the good work.

With regards.

William.