ai separation not resolving conflicts

[prtkilo]

Hi

I just updated to p3d 3.4 and fsuipc 9.957c (free function), but after update aiseparation seems not be resolving any conflicts. Prior to updating to 957c I ran briefly with only p3d updated to 3.4 and had the same issue. Previously I was using p3d 3.3.5 and fsuipc 9.955 where aiseparation mostly worked, occasionally failing to act on some conflicts.

Would appreciate any info on resolving this.

Regards

[Thomas Richter]

Hi,

what ever that function is 'ai separation to resolve conflicts' is not done by FSUIPC. You can read via FSUIPC Ai traffic info and control Ai Traffic range (in FS/P3D) but FSUIPC doesn't control or interfere with Ai Traffic.

Thomas

[Pete Dowson]

39 minutes ago, prtkilo said:

Prior to updating to 957c I ran briefly with only p3d updated to 3.4 and had the same issue.

In that case it certainly sounds like a change in P3D. Do you know for sure that AISpeparation uses FSUIPC? It does have a facility to send controls to AI aircraft, provided the correct ID nmber is retried first via FSUIPC's Traffic Tables (TCAS tables).

Tell me, does TrafficLook show all the local traffic? That's where the information comes from. The ID is extracted from the TCAS tables, accessed via FSUIPC offsets, and controls sent to AI aircraft via offset 2900, which instigates normal SimConnect controls to affect the AI unit.

Alternatively programs can interface directly with SimConnect to do the same thing, more efficiently (one less interface to go through).

The only real facility FSUIPC offers in this area which cannot be done via SimConnect is the AI Traffic Zapper. Normally AI deletion can only be done by the program which created them.

Pete

 

 

[prtkilo]

AIseparation is an old utility (last updated 2005) so pretty sure it works through fsuipc only. It has a table for showing AI in 40-50 mile vicinity and another to show conflicts (based on set criteria). AI information (including ID) is being shown normally in its tables as before, only that it is not doing anything to the conflicts i.e. aircraft are not getting slewed away to maintain required set distance. I haven't looked into TrafficLook but I think traffic is being detected fine, only the control message not seem to be reaching p3d.

[Pete Dowson]

27 minutes ago, prtkilo said:

traffic is being detected fine, only the control message not seem to be reaching p3d.

So, it seems that  since the 3.4 release, P3D has a problem sending controls to specific AI Traffic by ID. It would be worth reporting this to L-M on their website -- but they'd probably need to know which specific controls are being ignored. It may not be all of them.

I can only think of determining which controls work and which now don't by trying them. With FSUIPC, controls are sent via offset 2900. you'd need the ID, from Traffic Look. You could have a Lua program sending a different control each time you pressed a key or clicked a button and try to observe the effect of each, but really the first thing to find out is which controls AISeparation actually uses. Maybe it's then only one, crucial, one of those which is ineffective -- like trying ot control airspeed maybe. Not sure what control it would use for that.

I also seem to remember that, at least at one time, you could only really control AI if you put then into SLEW mode. I think Radar Contact used to do that in order to stop AI runway encroachments when the user aircraft is on finals.

I'm afraid I am away after this evening, until October 17th, so I can't really do much to assist in this at the moment. Maybe an appeal on the P3D Forum, or the L-M website, will elicit some feedback.

BTW, in FSX days,  the Traffic Explorer add-in, supplied in the SDK, had a facility to send controls to specific AI aircraft. I don't know if this is carried through into P3D, but it might be worth checking. Note that the traffic IDs used by that add-in are not the same as those in FSUIPC -- the latter ones are the 2's complement (i.e. the negative, treating the ID as a signed integer). This is historical, starting way back in FS9 days.

Pete

 

[prtkilo]

Found the TrafficLook utility, running it shows the airborne traffic normally as does the aiseparation table. However the IDs shown in the latter are of the type "&HFFFFFF7C" whereas TrafficLook shows the AI flight name as the ID, which is shown in the aiseparation utility under "flight" column. But this shouldn't indicate any problem as it has been like that since before.

The aiseparation utility does not try to fly/control AI in any way, only aims to shift it to desired distance. This is what the aiseparation readme says: "the planes are slewed backwards for a period of time calculated upon the speed of the SLEW and the distance to be moved" (capitals as original in the manual), which hints a simple slew might have been employed.

Edited October 2, 2016 by prtkilo

[Pete Dowson]

19 minutes ago, prtkilo said:

However the IDs shown in the latter are of the type "&HFFFFFF7C" whereas TrafficLook shows the AI flight name as the ID, which is shown in the aiseparation utility under "flight" column.

The traffic ID is NOT the same as either the flight number or the tail number, which TrafficLook can display as well. The Traffic ID is the internal number used to uniquely identify that FS object. All objects have IDs.

20 minutes ago, prtkilo said:

The aiseparation utility does not try to fly/control AI in any way, only aims to shift it to desired distance.

How do you think it can possibly do that without exercising any sort of control?

21 minutes ago, prtkilo said:

This is what the aiseparation readme says: "the planes are slewed backwards for a period of time calculated upon the speed of the SLEW and the distance to be moved"

Right, so its puts them into SLEW mode, as I suggested -- which is a specific FS control ("toggle slew"). Then it can move them about using the usual commands which you can do on your user aircraft in slew mode.

All this is "control" and involves sending controls using the aircraft's ID to select it.

Pete

 

 

[prtkilo]

I think I mixed up Traffic ID with TCASid then. In any case this should not be the issue area since there hasn't been any change to the aiseparation-fsuipc interface so the utility must be calling fsuipc correctly as before.

Sorry for the confusion there, I just meant to say aiseparation does not attempt to maneuver AI in flying mode which would involve multiple commands and difficult to narrow down to the problem one as you mentioned. But it is using slew mode which should make it simpler. Unfortunately I do not have experience working with fsuipc or simconnect commands so I will wait till your able to get back and look at it again, or someone knowledgeable in this matter may chip in.

[Pete Dowson]

4 hours ago, prtkilo said:

IUnfortunately I do not have experience working with fsuipc or simconnect commands so I will wait till your able to get back and look at it again, or someone knowledgeable in this matter may chip in.

We cannot solve it. It is a change in P3D 3.4. and needs reporting to L-M. 

Try sending traffic commands through Traffic Explorer. Try putting them into slew mode for example. The commands you can use are those listed in the List of Controls document in your FSUIPC Documents folder.

Last message from me now for 2 weeks ...

Pete

 

[prtkilo]

Installed p3d 3.3.5 client again and found that aiseparation is able to resolve conflicts with fsuipc 9.955 but not with 9.957c. Doesn't with either, on p3d 3.4. So something must have changed in both fsuipc and p3d.

[Pete Dowson]

On 10/3/2016 at 4:51 PM, prtkilo said:

Installed p3d 3.3.5 client again and found that aiseparation is able to resolve conflicts with fsuipc 9.955 but not with 9.957c. Doesn't with either, on p3d 3.4. So something must have changed in both fsuipc and p3d.

Version 9.xxx? I assume you mean 4.xxx?

I'll probably need some logging to investigate this.

The facility to send commands to AI aircraft certainly hasn't changed and works fine. There was a change to move the AI data reading, from SimConnect, into a separate SimConnect connection.  This was done because it was discovered that on heavily loaded systems the notifications of AI being added and removed stopped arriving from SimConnect. Separating them seemed to help a lot.

Please can you run TrafficLook (a utility provided in the Useful Additional Programs subforum) and see if you are actually getting traffic data read into FSUIPC?

Then, before I look at providing a logging version, could you try my latest internal update, FSUIPC4957d.zip. Also try adding "UseAIClient=No" to the [General] section in the FSUIPC.INI file. This makes the AI data reading revert back to the way it was.

Pete

 

[prtkilo]

Running p3d 3.3.5, with fsuipc 4957d aiseparation did not resolve conflicts, but did after adding UseAIClient=No entry. Ran TrafficLook which showed traffic in both cases.

[Pete Dowson]

52 minutes ago, prtkilo said:

Running p3d 3.3.5, with fsuipc 4957d aiseparation did not resolve conflicts, but did after adding UseAIClient=No entry. Ran TrafficLook which showed traffic in both cases.

Very strange. Here it works fine, just as well either way, and FS/P3D definitely runs a tiny bit better with the separated AI traffic client. If TrafficLook shows the traffic either way, that should certainly be enough for AIseparation too.

I'll see if making the command system for the traffic use the Traffic Client helps. Expect another version to try later today, to test without the UseAIClient parameter. Okay?

If that doesn't help I'll have to consider hich default to use. Maybe for safety have the separation off, but document it as a possible improvement to try.

Pete

 

[Pete Dowson]

I think I found the problem, but need it checked if you can plrease. I think the facility to send commands to AI traffic was using the separate client ok, but the commands needed had not been registered, so none worked!

Please download  FSUIPC4957f.zip. Please test with that "UseAIclient=No" parameter removed!

Let me know. But I think this must be it. I can see no other way it can be wrong!

Pete

 

[prtkilo]

aiseparation resolving conflicts now (without that parameter) with 4957f, in both p3d 3.3 and 3.4.

On other note could you tell more on the simconnect not reporting ai addition/removal? I am using UT2 and start having some duplicate and dead (which never move) AIs parked after letting sim run for a while. UT2 log shows unrecognized_id error when this happens and seems to create AIs that it is unable to track. Sometimes a couple of AIs and their plans also get mixed up. I was able to trigger these by overloading the cpu by running a file compression program in background that caused heavy stutter in the sim. If so is this something that could be fixed by LM in a future release? I have nagle disabled through simconnect.cfg but not sure if it has improvement as I still get these problems sometimes.

[Pete Dowson]

4 hours ago, prtkilo said:

aiseparation resolving conflicts now (without that parameter) with 4957f, in both p3d 3.3 and 3.4.

Ah, good. Thanks for testing. 

4 hours ago, prtkilo said:

On other note could you tell more on the simconnect not reporting ai addition/removal?

I don't know what causes that. I cannot make it happen on either of my systems here, in FSX-SE or P3D, but my friend Ray was getting this problem, which is why I started experimenting with ways around it. The separate Client helped, but didn't actually solve it.

He eventually solved the problem by disabling the time sync option in FSUIPC. Seems his system is running at a different clock rate to FS and the time was being reset at intervals -- only by seconds, but somehow this must have been enough to mess up the traffic notifications. That was with FSX-MS though.

4 hours ago, prtkilo said:

I am using UT2 and start having some duplicate and dead (which never move) AIs parked after letting sim run for a while. UT2 log shows unrecognized_id error when this happens and seems to create AIs that it is unable to track. Sometimes a couple of AIs and their plans also get mixed up. I was able to trigger these by overloading the cpu by running a file compression program in background that caused heavy stutter in the sim. If so is this something that could be fixed by LM in a future release?

Sorry, I don't know. I use UT2 together with MyTraffic6, and have not noticed a problem specifically. I do know that if you reload a flight or even just "refresh scenery" (there's a control for that) the AI Traffic can get re-generated and produce duplicates. I've had up to 3 AI aircraft sharing gates -- actually drawn inside each other!

Pete