European IFR without oxygen, and Eurocontrol routings generally

tomjnx: I've also noticed this with the FPP routes. If I change my departure airport to ESMS, which is very close to my home base and has published SIDs, I get a completely different route. Interesting logic.

Achimha: I use FPP and eurofpl.eu to find my routes then do the filing via our Swedish flight planning center, which is free of charge. Great website for filing and incredibly helpful staff when you call them to get assistance.

Interesting thread, thanks for posting everyone - you have some great insights into the ec-matrix :)

I'd be interested to know how you can hack the eurocontrol system...

The easiest way is to enter RMK/IFPS REROUTING ACCEPTED under item 18 of the flight plan. If it gets rejected by CFMU in Brussels, some poor guy at AIS will manually adjust your plan until it gets accepted and send you the new routing by SMS and/or email. You can get the most nonsensical routings to work that way and the service is still free of charge, at least here in Germany.

I believe you that the MID000000 trick worked in the past. But I'd be interested in an example that still works now.

What does Deadalus do? If you do something like route suggestions, with only airway routing, no Y/Z, just one level, easily giving up, then yes, that is fairly easily tractable.

If you want full support of DCT's, level changes and allow low altitude operation, it becomes resource consuming.

The routing graph for a typical routing problem for a 300nm flight including DCT's and 5 routing levels has several hundred thousand edges. Then you need to compute the shortest path within the routing graph. If you're lucky, you've got a route. But most of the time, this route violates some eRAD or other flight plan rule, so you'll need to compute the second best path, and so on. Typically, I compute about 100 routes before I get one that passes.

Many eRAD rules do not directly map to routing graph changes.

Worst are the mandatory rules, something like if you cross X, then you need to cross Y at any later time. This is something that doesn't map to routing graph modification.

Or the ED3115A rule, Frankfurt is forbidden unless you fly elsewhere within Germany above FL135. Not codable in the graph structure.

Or flight plans may not incorporate the same point (or a different point with the same name) more than once. How do you code that into the routing graph structure?

Currently, we have about 8200 eRAD rules that can be obtained from Eurocontrol. After removing those that can be easily determined to not apply (because they only apply to different aircraft types, to different departure or destination airports, to different class of aircraft, to different altitude bands, to regions on the globe sufficiently away from the likely route etc.) there are still over 1000 potentially applicable rules.

I haven't been around in IFR circles in 2005, but I suspect the problem size has grown considerably since then.

Right now, I can read and preprocess (remove rules which are obviously not applicable) within about 10-20s, construct the routing graph in about 10-20s. Evaluation of all remaining rules for a single candidate flightplan is several 100ms, computation of a new candidate flightplan is also several 100ms.

The setup process can certainly be amortized over many route computations, I fail to see how the router iteration can be made much faster, without throwing a lot of hardware at it or making the problem simpler.

Good that FPP worked for you - it didn't for me. Switzerland and Austria seems hard to do flight planning on a low level. Also, last time I checked, FPP couldn't generically handle DCT's. That's why even early versions of my router generated much better routings than FPP for my use cases. Also, FPP often failed to get a route at all, and I think it doesn't handle level changes.

I don't know about Autoplan.

"Maze through airway" - yes exactly! If you don't handle DCT's generically, the problem is suddenly an order of a magnitude simpler!

And I don't quite see how a maze router fits the problem. Maze routers are good if your problem is rectangular, like a VLSI grid router. It doesn't seem a good fit to an airway graph. Or to put it differently, a maze router when implemented over an airway graph probably degenerates to something similar than Dijkstra or worse, depending on the visitor strategy.

But the problem remains, you cannot map many of the eRAD rules to changes in the maze structure. Or the graph structure.

The database issue is indeed a bit of a problem. I cannot evaluate a certain amount of eRAD rules because I am missing the airspace definitions they refer to. Eurocontrol says they cannot publish their airspace data, because the NAA's don't allow it to, and EAD says they cannot add the definition because it hasn't been published. I tried to track a few of those airspaces down, and indeed they weren't published by the NAA's/ANSPs.

If you actually read the description of what my router does, you'll notice that my tool also does remote (IFPS) validation. It actually started out with only remote validation. But this solution has two problems. One, the error messages are often imprecise or not understandable at all, leading to the routing graph (or maze) being modified in a too pessimistic way, thereby losing potential solutions. The other problem is that it's hard to do good DCT support without understanding the DCT rules, at the time I had DCT implemented but no local rule validation, it would often do 500 remote validations until a plan worked.

So really, you need to do both. You need to do remote validation, because local validation is not 100% accurate because some rules cannot be tested due to missing reference data, and local validation, because otherwise you'd be generating excessive server load and it's just needed to actually understand what the Eurocontrol validator is actually telling you. And the DCT rules need to be evaluated locally to intelligently add the DCT arcs to the routing graph/maze.

If you actually read the description of what my router does,

Which is your router?

Personally I think IFPS is hardly a problem anymore. Tools like RocketRoute make it very simple for the user to come up with a halfway direct routing. It is not more difficult to file an IFR flight plan in Europe than it is in the US (although you have to pay extra for the router software).

This one:

It is working for me. The remaining piece is essentially the legal stuff for the distribution (or rather limiting the distribution) of the eurocontrol data.

I would like to compare the RR router with mine, is there an easy way to do this (for less than 230 pounds)? I compared the only route from my homebase I could find on their webpage to what my router would give me, it's 375NM vs. 350NM and mostly FL110 but FL90 to FL120 vs FL90 for LSZK->LFMD.