FSX-MS AI Controller 1.4 (SID, Jet Route, STAR and Final Approach) (New) Beta

[Clipper Ocean Spray 189]

Looks like Hot Fix 9 was pretty solid based on feedback, so I've decided to concentrate on performance issues with hot fix 10.

Next step hopefully is releasing 1.4B so I can start working on the next major version.

 

Hot fix 10 adds NO new features or bug fixes (really isn't a "fix").  However, HF10 introduces many performance improvements under the hood.  Thus, it is a bit experimental and will likely have new bugs.

I'm hesitant to link to this version in the main thread, especially since hot fix 9 seems solid.  

 

If you're satisfied with HF9, then HF10 probably isn't for you.  However, if you have ever felt that AIMonitor (AIM)/AIController (AIC) take up too much CPU time and disk I/O, then consider giving it a try. 

Hotfix 10 is targeted to those users running AIM and AIC on the same PC as FSX/P3D.   I would greatly appreciate feedback.

 

HF10 drastically reduces disk I/O between AIM and AIC.  Instead, the two programs now exchange data via memory (note:  AIM/AIC are .exe files and take no virtual address space from FSX/P3D).

 

Hotfix 10 also uses more multi-threading to parallelize the work load, significantly reducing CPU loading.

 

I've also added a "fast jet route calculation" mode (enabled by default) that uses much less CPU cycles when calculating jet routes, especially after the program has been running awhile.  The downside is less jetroute path-finding accuracy.  Because the path-finding is less accurate, there will also be more instances of the program not being able to find a jet route for a particular AI and thus releasing said AI back to FSX during the enroute portion.

 

The new beta version of AIController v1.14B (full version including Hot Fix 10) is available at the following link:

 

http://www.mediafire.com/download/klb71x6826sg5c7/AIController14BFullWithHF10.zip

 

Hot fix 10 by itself is available at the following link:

 

http://www.mediafire.com/download/alc4dr52n1dn3cf/AIController14BOnlyHF10.zip

 

Please do not mix with any version of 1.4A or earlier.  Instead, install into a separate directory for testing.

 

IMPORTANT:  After installation, please select "Options\Reset Options to Default" and run a "Force Update Runway and Taxi Data".

 

Important Revision:  A registered version (4.9x or later) of Peter Dowson's FSUIPC utility is recommended to use the full functionality of AIController.  For example, AIController will not be able to delete problematic AI targeted for removal (e.g., parked at gate after arrival for the specificed parking time).  For networked configuration users, Peter's WideFS (ver. 6.9) is also recommended.

 

Important New NOTAMS!

 

ENTERING FINAL APPROACH COORDINATES.  Do not use FSX when entering final approach coordinates for runway positions (e.g., touchdown, rollout) as the FSX in-game coordinate display (shift-Z) lacks the necessary precision (significant digits).  Use the freeware Airport Design Editor X (or similar) instead.

 

ENTERING RUNWAY DATA.   Ensure the runway number corresponds to the FSX runway description (for whatever airport scenery you are using).  Often, the latest FMS data will reference changed runway numbers that do not correspond to stock FSX runway descriptions.  For STAR files, if the runways do not match (STAR versus FSX), AIController will choose the STAR based on the waypoint having the closest entry point to the AI’s current position (i.e., not based on runway data), which often is satisfactory, but not always.  For Final Approach files, if the runways do not match, AIController might choose the final 

approach file having the incorrect runway data.

 

Please see the Original Thread Post for Additional Details Regarding Ver. 1.4 Including Installation Video.

 

Thanks!

 

-Roland

[FlyTweety 2]

 

Looks like Hot Fix 9 was pretty solid based on feedback, so I've decided to concentrate on performance issues with hot fix 10.

Next step hopefully is releasing 1.4B so I can start working on the next major version.

Here in Germany it is Summertime, so i must admit, i don´t spent much time for testing in the past weeks, yesterday i gave HF10 a chance and watched the traffic with one eye while watching the 24 hours of Le Mans with the second...

 

First, after dozends of Hotfixes for the first time the "MSVCP120.dll" caught me too.

Means, i had to install the 32 bit disttribution as well. Can you tell me why? Did you changed the compiler? Not a big thing, though i would like to continue to use 64 bit calls.

 

What i would like to comment is following:

I still notice that no AI uses the "direct in" Approach.

Also when a Aircraft is on "long Final" and is heading direct to the Runway, AIC uses a left or right base approach, what means a huge detour.

The current angle is 45° for me, but no way that one come in direct.

 

Also, on my well known test ground EDDL, the very most STAR´s for the 05 ending at RONAD which is about half way of the ILS shifted about 5 miles to north.

 

What AIC now very often (allways ?) do is, it sends the AI all the way back to the first downwind entry (Enhanced Autogen Approach). But, in this case, the only logical step would be to skip the first 2 WP and assign the "Base Entry" WP. And of course, the longer i make the final distance (i like the value at 15 miles), the longer the downwind legs will be.

 

Here i would appreciate a (more) steady recognition that the next WP would be in the opposite heading and necessary to skip.

 

Further, i would appreciate to come back to ICAO based settings at some points. E.g the Final Approach height in EDDF need to be 4000 ft, different to 3000 ft at the most Airports.

 

One time i moved to TNCM, there i noticed that the Commuter Airlines sometimes also receive (long) SID´s and STAR´s while hopping between the Islands.

There´s a dilemma, when i use the new option of simplified ICAO based enhanced approaches, the Intercontinental traffic uses them as well.

 

My suggestion would be to add a Parameter that avoid traffic patterns when the the distance between departure and approach is less than XXX miles (I think, 25-30 Miles are fine).

So, no SID and maximum a short final starting at the Base Entry.

Additional, Pistons and Turbo-Props could have a way shorter Final distance for separation purposes, maybe the half of the common value.

 

Another point that end up in chaos seems to be a small bug(?) in the basic options:

 

When i change the minimum separation between AI Arcraft and click apply, then close the window, the value of the minimum separation between User Aircraft and AI is changed too to the same value, which leads to massive Go Arrounds while i hang arround at the sky and watch the traffic.

 

For some reason not only the single setting is changed but both. And more strange:

 

I open AIM and change the Minimum Separation. Open the aicontroller.ini, the distance to the User AI is zero. I close AIM, re-open aicontroller.ini, distance to the User AI is now the same than AI to AI.

I edit to zero again by hand and save the .ini.

Re Open AIM again and open the basic options, the AI to User is still the same than AI to AI. Did it not read the ini?

In short terms, what AIM displays in the box is not what will be saved as soon as i change the AI to AI distance, and what is shown in the aicontroller.ini is not what will be read.

Confusing...

 

Maybe you can look over it.

[Speedbird 217 128]

First, after dozends of Hotfixes for the first time the "MSVCP120.dll" caught me too.

Means, i had to install the 32 bit disttribution as well. Can you tell me why? Did you changed the compiler? Not a big thing, though i would like to continue to use 64 bit calls.

 

 

I also got 3 MSVCP120.dll crashes yesterday during one flight. Could you please share which distribution you installed to fix this?

[chumley 0]

Hi Roland 

 

Not sure if this is related to the operation of AiController.

 

I have just completed a flight across Europe and AiController worked as expected. However, as the flight progressed AI started to disappear from the EFB display until upon landing at LOWW there wasn't a single aircraft displayed - either in the air or on the ground on EFB but AiController showed a fair few to be active and the models where displayed in FSX.

 

I have put up a message on the EFB forum but what puzzles me is that when I stopped AiController and AiMonitor and restarted the EFB display all the AI shown on AiMonitor where present on EFB and in FSX.

 

It is interesting that this phenomenon has only shown itself crossing Europe on two occasions, the first from Dubai to Dublin when I started off with a large number of AI and ended up with none in EFB and the second occasion when flying from Gatwick to Vienna once again when I started off with a large number of AI and ended up with none in EFB.

 

It crossed my mind that simconnect may be having a hissy fit over the network.

[FlyTweety 2]

I also got 3 MSVCP120.dll crashes yesterday during one flight. Could you please share which distribution you installed to fix this?

Oh, i only searched for the Visual C++ Redistributable Packages für Visual Studio 2013 at the Microsoft site. There you choose the x86 package (works only with Internet Explorer).

 

Installing....done. Then AIM worked like before.

[opchicago 0]

Hi,

 

Thank you for this great program, it is a great leap forward for AI and FSX. I am a little confused over how ATC(FSX stock, Radar Contact and PROATC) interact with landing traffic. I tune my radio to Tower at KBWI, one of the airports that I am monitoring using enhanced mode, I hear the take offs just fine and can hear them taxing out on Ground. However, I see but never hear any interaction between landing aircraft as they land and as they taxi. The aircraft show as controlled by ai controller in ai monitor.

 

They seem to be acting correctly, landing on the correct runway etc and avoiding collisions while taxing, but I am a little unsure as to why I can not hear them.

 

Thank You.

[Speedbird 217 128]

Oh, i only searched for the Visual C++ Redistributable Packages für Visual Studio 2013 at the Microsoft site. There you choose the x86 package (works only with Internet Explorer).

 

Installing....done. Then AIM worked like before.

 

Thanks. I already have Microsoft Visual C++ 2013 Redistributable (x86) - 12.0.30501 installed but seem to get the crashes nonetheless with HF10.

[Clipper Ocean Spray 189]

 

 

First, after dozends of Hotfixes for the first time the "MSVCP120.dll" caught me too.
Means, i had to install the 32 bit disttribution as well. Can you tell me why? Did you changed the compiler? Not a big thing, though i would like to continue to use 64 bit calls.

 

Probably has to do with my switch to data transfer via RAM needing very specialized 32-bit libraries (AIC is a 32-bit console app.).  I think it is worth it though...much less disk thrashing.

 

 

I still notice that no AI uses the "direct in" Approach.

 

Looking into this (straight-in approaches not being triggered when appropriate).  

 

 

What AIC now very often (allways ?) do is, it sends the AI all the way back to the first downwind entry (Enhanced Autogen Approach). But, in this case, the only logical step would be to skip the first 2 WP and assign the "Base Entry" WP. And of course, the longer i make the final distance (i like the value at 15 miles), the longer the downwind legs will be.

 

Yep, bug.  Thanks for finding it.  The full downwind (left or right) is always being selected.  Already fixed for the next release.

 

 

Further, i would appreciate to come back to ICAO based settings at some points. E.g the Final Approach height in EDDF need to be 4000 ft, different to 3000 ft at the most Airports.

 

Good suggestion, or I might make all those enhanced approach parameters changeable on-the-fly (i.e., while FSX/P3D is running).

 

 

I open AIM and change the Minimum Separation. Open the aicontroller.ini, the distance to the User AI is zero. I close AIM, re-open aicontroller.ini, distance to the User AI is now the same than AI to AI.
I edit to zero again by hand and save the .ini.
Re Open AIM again and open the basic options, the AI to User is still the same than AI to AI. Did it not read the ini?

 

Ok, that's a bug too.  Many thanks.  Clarification though...AIM doesn't read the .ini file for parameter values, it is this way:  AIM --> .ini --> AIC.  So if you hit "apply" in AIM, those settings will be applied to the ini. and read by AIC.  The bug is when AIM reloads it sets AI-to-user equal to AI-to-AI, so if you want them different manually change and "apply" (as a workaround until the bug is fixed).

[Clipper Ocean Spray 189]

However, as the flight progressed AI started to disappear from the EFB display until upon landing at LOWW there wasn't a single aircraft displayed - either in the air or on the ground on EFB but AiController showed a fair few to be active and the models where displayed in FSX.

 

Doesn't EFB sometimes loose track of AI on the display?  I've heard this before, but it could be wrong and it was some time ago.  

 

 

It crossed my mind that simconnect may be having a hissy fit over the network.

 

It is possible, but AIC doesn't really hit the simconnect link that hard owing to the very slow (in control terms) sample rate I use (1 sample per second) and the amount of AI being controlled always being hard-limited by the reality bubble around the user.  EFB doesn't use simconnect, correct?

[chumley 0]

Doesn't EFB sometimes loose track of AI on the display?  I've heard this before, but it could be wrong and it was some time ago.  

 

 

 

It is possible, but AIC doesn't really hit the simconnect link that hard owing to the very slow (in control terms) sample rate I use (1 sample per second) and the amount of AI being controlled always being hard-limited by the reality bubble around the user.  EFB doesn't use simconnect, correct?

EFB does use Simconnect. The developer is moving to FSUIPC in version 2 of EFB which is somewhere between conception and delivery.

 

As for losing track of AI the developer mentioned it as a Simconnect issue. I have only experienced it twice both times in the thick of things over central Europe. I wondered if AiController might be adding to the load but from what you say it is unlikely. 

 

Next time I'm crossing central Europe I will switch off AiController and see what happens.

[Clipper Ocean Spray 189]

 

 

EFB does use Simconnect. The developer is moving to FSUIPC in version 2 of EFB which is somewhere between conception and delivery.
 
As for losing track of AI the developer mentioned it as a Simconnect issue. I have only experienced it twice both times in the thick of things over central Europe. I wondered if AiController might be adding to the load but from what you say it is unlikely. 
 

 

Interesting, thanks!  I've always marveled at the EFB user-interface, good to hear they're still developing.  Can't wait to see what the next version brings.

 

I don't have a problem with Simconnect losing track of the AI, but I'm probably doing it differently.  I came up with my "tracking" design very early-on and fortunately it seems fairly robust (it had to be really or nothing else would have worked). 

 

 

I am a little confused over how ATC(FSX stock, Radar Contact and PROATC) interact with landing traffic. I tune my radio to Tower at KBWI, one of the airports that I am monitoring using enhanced mode, I hear the take offs just fine and can hear them taxing out on Ground. However, I see but never hear any interaction between landing aircraft as they land and as they taxi.

 

You're exactly right.  Landing and taxi-in air traffic communication is affected.  Fortunately there's some nice work being done by Paul Henty in support of FSUIPC that should allow me to easily update the FSUIPC TCAS tables for all ATC programs reading this table for AI status (e.g., Radar Contact comes to mind).

[Clipper Ocean Spray 189]

 

 

I also got 3 MSVCP120.dll crashes yesterday during one flight. Could you please share which distribution you installed to fix this?

 

That number of crashes makes me think it is a configuration problem.  You also have .NET 4.0 or greater installed too?

 

Did the crashes happen quickly and about the same time after the program loaded, or did they seem to occur more randomly?

 

Is there any sequence of events (they could be a routine way you use the program that never caused a problem in HF 9) that seem to trigger the crash?

 

There were a lot of changes under the hood in HF10.  Potentially the most important (in terms of crash potential) is multi-threading.  When AIC starts controlling AI there are about 4 real worker threads running in parallel.  If anyone experiences AIC crashing (especially in a random one-off fashion) please let me know.  I can always put out a quick revision that dials it back.

[FlyTweety 2]

Sorry for moaning all the time :smile: but i guess i got one more bug.

 

After forcing the AI into a Go Arround the Approach won´t work . In the Short Final the AI moves up and down between 2800 and 2200 ft like a wacko elevator. This happens at least with the Enhanced GA right hand patterns.

 

AI Aircraft never comes down to earth but will freezed at about 2800ft anywhere above the Runway and wait for taxiing (that not happens).

 

Currently i have the odd issue, that at TNCM (FlyTampa) the UT2 KLM 747-400 decellerates after landing, but soon before it stops she accellerate again to hyperspeed and dissappers in the mountain...

Not sure what happens there, actually it is a frequetlly used Model in the sim (e.g. Lufthansa in EDDF)  and i never observed a similar prob. anywhere else.

 

So, not sure what it is related, Scenery, Model (contact points?) or AIC...

[Speedbird 217 128]

That number of crashes makes me think it is a configuration problem.  You also have .NET 4.0 or greater installed too?

 

Did the crashes happen quickly and about the same time after the program loaded, or did they seem to occur more randomly?

 

Is there any sequence of events (they could be a routine way you use the program that never caused a problem in HF 9) that seem to trigger the crash?

 

There were a lot of changes under the hood in HF10.  Potentially the most important (in terms of crash potential) is multi-threading.  When AIC starts controlling AI there are about 4 real worker threads running in parallel.  If anyone experiences AIC crashing (especially in a random one-off fashion) please let me know.  I can always put out a quick revision that dials it back.

 

It seemed quite random. No specific phase of flight, not at similar times into the session. I never had any issues with this in HF9. I report back once I have done another flight.

Thanks for developing this great tool btw

[chumley 0]

Hello Roland

 

is there a way in which aircraft under the control of AiController on the ground can be prevented from stopping every time an AI aircraft is landing or taking off? At busy airports this gets to be a problem.