P3D v4.5 (or 4.4) Crash with AIseparation or RC4

[Ray Proudfoot]

After multiple tests and sending the results to Pete and John Dowson we appear to have found a way of avoiding the CTDs when running RC4 but crucially only when running RC4 via WideFS. This was tested with Ai at 100% at EGLL.

In the wideclient.ini file locate the [User] section and add this line:-

Deny=2900,12

For those running RC4 on the same PC as P3D a solution has yet to be found. The problem appears to be related to RC4 trying to slew multiple aircraft all at once even those not in any danger of colliding with the user's aircraft. The above command prevents that happening. All the requests come at once causing a problem with P3D. The best way of avoiding the crashes it to limit Ai% to less than 50.

As that code is in RC4 and John Dekker no longer supports the program why that is done will never be known. Hopefully a fix can be found but after 6 hours of testing I'm knackered and there's a certain football match I want to watch this evening.

[robschofield]

I appear to have similar problem with FSAIS v1.1, the AI traffic smoothing programme (which I still use!)

It started crashing FSX shortly after loading it up at busy airports, and as soon as it starts slewing AI planes using FSUIPC to achieve desire separation P3d4.5 HF crashes. Repeatable at a number of locations. 

Am sure Pete will find a similar fix for insertion in FSUIPC.

[Pete Dowson]

2 hours ago, robschofield said:

It started crashing FSX shortly after loading it up at busy airports, and as soon as it starts slewing AI planes using FSUIPC to achieve desire separation P3d4.5

Sorry, the same happens in FSX as well? So it isn't just a weakness we've come across in P3D4?

2 hours ago, robschofield said:

Am sure Pete will find a similar fix for insertion in FSUIPC.

Not I. It will be John. And if slowing down the requests sent to P3D works, it would apply to both local and WideFS clients. But we don't yet know is slowing the requests down will be the solution. Stopping them altogether, as is done by the temporary fix for RC, would render your traffic smoothing useless too I should imagine.

Which program is it, anyway? AIseparation or AIsmooth or something else? I thought Ray said those old FSX programs didn't work with P3D4, though if they a 100% using only FSUIPC I don't see why not.

Pete

 

 

[Quoth]

Hi John

Well the sweet spot for me appears to be 40% AI, tried a couple of flights with no crashes which is nice 😉 and the amount of AI isn’t too bad although it’s always nice to have more 😀. One other thing I meant to ask, does it matter how many programmes utilise fsuipc at the same time?

Hope you’re able to find a fix and thanks for your help, if you need any more logs let me know

Regards

Alan

ps this is using RC4 on the same computer as P3d 

 

[Ray Proudfoot]

3 hours ago, Pete Dowson said:

Which program is it, anyway? AIseparation or AIsmooth or something else? I thought Ray said those old FSX programs didn't work with P3D4, though if they a 100% using only FSUIPC I don't see why not.

Hi Pete, I'm only going off what I've read elsewhere. Not tried either since FSX or even FS9 days. Just to report I had a successful flight this morning from Atlanta to Raleigh-Durham. Not as many Ai at KATL as I thought but probably due to the dawn departure. Maximum of 126 in the 80nm bubble and no problems at all. When I lined up to take-off I noticed two Delta aircraft crossing my runway to taxi to the terminal. I just waited until they cleared and then pushed the throttles forward.

[robschofield]

Pete - its ai separation which I use - and apologies, I'm running p3dv4.5 HF not FSX!! FSAIS has worked with every p3D version I have had, and probably for the reason you describe which it is solely using FSUIPC as its link into any simulator. Works fine in 4.5HF with FSUIPC 5.151 as long as it is not trying to slew an ai aircraft to maintain separation.

It has only begun manifesting itself in P3dv4.5.

Best regards

 

Rob

[Pete Dowson]

2 hours ago, robschofield said:

It has only begun manifesting itself in P3dv4.5.

It does seem to point to a weakness somewhere in P3d4.5. What about previous P3D4 versions?

Currently we don't know of any solution other than the drastic one of stopping FSUIPC applications using the facilities in FSUIPC for controlling/slewing AI aircraft.  I don't like that idea, and for AIseparation it destroys its purpose altogether.

There are some things which we might try, but it is in John's hands. I am on holiday soon and won't be able to assist till july.

If you know it was okay in P3D4.4, or 4.3, say, it would give us a bit more leverage with L-M. Currently to prove it is a P3D problem and not one or more addons is not easy -- in fact they'll want us to demonstrate it with the most simple of tests, possibly not even with FSUIPC (i.e. a small program calling the same functions in P3D as FSUIPC does). Of course they'd also need a decent number of AI too.

How many AI aircraft were you getting at the times of the crashes? Is AIseparation going to be really busy at the time? One of the tests we thought we'd do is to try and slow the request rate down, but if it occurs with few AI being controlled that might be doomed.

Pete

 

 

[robschofield]

Pete - have re-installed the P3dv4.4 client application last night and will do some testing this week - is there any way of getting access to the legacy versions of FSUIPC - eg 5.15 instead of 5.151 or 5.151b which were updated to accommodate v4.5? Didn't keep by prior downloads!

Regards

 

Rob

[Pete Dowson]

28 minutes ago, robschofield said:

is there any way of getting access to the legacy versions of FSUIPC

It wouldn't make any difference to this problem.  If an earlier version of P3D4 doesn't fix it, the only interim solution for you is not to use AIseparation for the time being.

There have not been any changes in FSUIPC5 to accommodate any P3D updates recently. The only "accommodations" have been when P3D4.0 was released and then with 4.1 to take advantageof new facilities (in the PDK). Unlike FSUIPC4, which had to be updated for every release, FSUIPC5 is designed to be free of such hassle.

Pete

 

[Ray Proudfoot]

Hi Pete,

Given this problem never occured with MT6 or Ultimate Traffic Live I was curious about how much more Ai was injected into P3D by the AIG Ai. Sitting at 414 at EGLL at 14:17 Local (13:17 GMT) this is the count for each product with 100% Ai set in P3D and no restrictions in the AIG airline %. Total / Ground / Airborne.

AIG = 284 / 216 / 68

UTL = 199 / 170 / 29

I have no idea what the maximum number of slewed Ai P3D can cope with but thought those numbers might be helpful.

[Drumcode]

My understanding is there's no work around yet for the user running a local version of RC?

[Ray Proudfoot]

6 minutes ago, Drumcode said:

My understanding is there's no work around yet for the user running a local version of RC?

Correct although John has just installed RC4 so once he's generated the crash he'll be working on a (hopeful) fix.

[Drumcode]

Excellent. Fingers crossed.

[robschofield]

Did testing last night having found a copy of v 1.5 in another post on this topic.

 

Outcome - FSAIS works fine in either P3Dv4.4 or P3Dv4.5 with FSUIPC v5.1.5, but as soon as you use 5.1.5.1 or 5.1.5.1a or b then irrespective of whether you are using P3dv4.4 or 4.5, it will crash when ai slewing becomes active. I think John has picked this up in the other post on this topic, has got a copy of RC4, and suspect whatever fix does the job for RC4 will similarly work on any other application which makes use of FSUIPC as its interface into the simulator.

[Pete Dowson]

1 hour ago, robschofield said:

Outcome - FSAIS works fine in either P3Dv4.4 or P3Dv4.5 with FSUIPC v5.1.5, but as soon as you use 5.1.5.1 or 5.1.5.1a or b then irrespective of whether you are using P3dv4.4 or 4.5, it will crash when ai slewing becomes active. I think John has picked this up in the other post on this topic, has got a copy of RC4, and suspect whatever fix does the job for RC4 will similarly work on any other application which makes use of FSUIPC as its interface into the simulator.

Hmm. Curiouser and curiouser ... I can't imagine which of the relatively small changes since 5.150 could make any such difference.

I would like to assist John with this, but I have precious little time now before going on holiday.

As for reproducing it, it may be better (quicker) for John to use FSAIS (AIseparation) instead of RC4, as the latter needs rather more setting up and learning.  Is this the right download?

Category: Flight Simulator 2004 - Utilities
Flight Simulator AI Separation

File Description:
Flight Simulator A.I. traffic separation. With this tool you will be able to separate the traffic inbound to an airport defining a set of paremeters like minimun separation, desired separation, reduced separation zone, reducing the annoying go arounds that the AI engine of FS2004 can cause.

Filename:	fsaisv1.1.zip
License:	Freeware, limited distribution
Added:	17th March 2005, 13:41:42
Downloads:	31540
Author:	Armando Di Francesco
Size:	497kb

 

I'll try it here if I get time over the next two days.

Pete

 

[Pete Dowson]

Further to the above, I've installed AIseparation and I am running it connected to FSUIPC5.151 on P3D4.5 under win7.

I am at EGCC with over 300 AI about -- about 1/3rd in the air.

How do I encourage AISmooth to actually do something? I have left it with "airborne" selected, because I can't make sense of the Ground" option -- how does it separate Ground AI using the parameters -- obvious distances of 7nm or up don't make sense.

So far it hasn't done anything at all with AI, nor logged any 'conflicts'.

Pete

 

[Pete Dowson]

Okay, I see I have to tell AIsmooth to retrieve data for it to do anything.

Both John and I have been running in with 5.151b and so far there's no crashes. Maybe we just don't have enough aircraft? I'm at EGCC with around 300, john also at EGLL but with 380. AIseparation seems to be trying to do things but still no crashes,

Oddly, monitoring AIseparation's efforts in the relevant offsets, it appears to set a selected aircraft into SLEW mode with SLEW on, then it sends a SLEW AHEAD MINUS (to move it backwards) followed by a SLEW off. BUT it then continues to adjust that same aircraft with more SLEW AHEAD MINUS and SLEW OFF pairs, with no further SLEW ON for that aircraft. So it really isn't doing what it sets out to do -- just one brief adjustment is made.

Anyway, we are still concerned as to why it crashes your P3D4 but neither of ours. Could you tell us how many AI you have at the time? It might be useful also for you to monitor the relevant offsets so please do this as follows:
FSUIPC Logging Tab
Right hand side, enter offset 2900 type U32, 'hex' selected, and 2904 type U32 not hex.
Check 'normal log' below.

When the crash occurs please sned the FSUIPC5.LOG file. Also the AISmooth settings please (is there an INI file?).

Pete

 

[Drumcode]

@Pete Dowson I never get crashes on the ground, it's either during roll or mid-air. The other day I left P3D running for two days straight at ORD and it would not CTD, but as soon as I was about 40NM west of it, it crashed.

[Pete Dowson]

17 minutes ago, Drumcode said:

I never get crashes on the ground, it's either during roll or mid-air. The other day I left P3D running for two days straight at ORD and it would not CTD, but as soon as I was about 40NM west of it, it crashed.

That's with RC, AIseparation, so I'm not surprised. It will only be trying to use the facilities to handle aI in certain curcumstances.

We are concentration on trying to reproduce it with AIseparation, which at a busy airport should be doing this stuff much more frequently. We need to be able to get it happen reproducibly.

Pete

 

[robschofield]

Pete - I can get it to occur at EDDT, LTFM and a few others I have tried, as soon as the first 'conflict' is managed in FSAIS. At Berlin, there were less than 130 AI aircraft active at the time according to AI Traffic Manager and significantly less than 300 at LTFM. Am back 0n 1.5 as that works fine, but will but 1.5.1 back in at the weekend and send some log files as requested.

 

Regards

 

Rob

[Pete Dowson]

2 hours ago, robschofield said:

I can get it to occur at EDDT, LTFM and a few others I have tried, as soon as the first 'conflict' is managed in FSAIS. At Berlin, there were less than 130 AI aircraft active at the time according to AI Traffic Manager and significantly less than 300 at LTFM. Am back 0n 1.5 as that works fine, but will but 1.5.1 back in at the weekend and send some log files as requested.

I think it must be very timing dependent, and it still appears to be a P3D problem. With 5.15 it may just be very infrequent. Yesterday I had AIseparation running for hours at Heathrow with lots of conflicts being listed and treated, and no crash. Before that I managed to generate one crash quite early one, but unfortunately that wasn't the Debug version and wasn't running under the debugger, so I couldn't track it to source.

John is pondering what to do. He might write a separate program doing the same thing directly into SimConnect, not via FSUIPC, because then if we can force the crash we can present it to L-M as evidence.

Note that all FSUIPC does when receiving these SLEW mode controls (or any controls for that matter), aimed at an AI aircraft, it simply sends it on to P3d for execution in exactly the same way as it does for any assigned controls for the user aircraft. The only difference is the ID of the 'object' being commanded.

 

I'm off on holiday soon so John will continue this when he can.

Pete

 

 

[Quoth]

Hi Guys

I’m getting confused, is the problem with RC4 crashing fsuipc the same problem as ai smooth?

Regards

G

[Pete Dowson]

1 minute ago, Quoth said:

I’m getting confused, is the problem with RC4 crashing fsuipc the same problem as ai smooth?

Yes, It certainly seems so. It's related to trying to control AI in Slew mode via FSUIPC offsets 2900 & 2904.

Pete

 

[Ray Proudfoot]

Using RC4 at EGLL I can guarantee a crash either when lining up on the runway or as soon as the wheels leave the ground.

I haven't heard from John since he got RC4 installed but the actions required to re-create the crash take just a few minutes.

[John Dowson]

Hi Ray,

I started with RC4 to try and reproduce the crash but didn't get too far and switched to trying with FSAIS instead, but still no joy.

I can go back to trying with RC4 - can you provide more details on how to re-create the crash you get with this?

Thanks,

John