State of AI Traffic in MSFS (Update: May 2022)

Around 4 months after the last Status Update on the state of AI traffic within MSFS and one Sim Update (SU9) later we have checked again the AI system in MSFS and post our findings here.

Live Traffic (Online AI Traffic – Default MSFS)

In this Section we will focus on features only used by live traffic.

Model and livery Matching

The correct algorithm for livery/model-matching is still unknown. The SDK is indicating icao_model and icao_airline as the values to go here, but no word on what happens if only one of them is matching or there are multiple possible matches.

The best results can be achieved when having liveries for the correct model (should be obvious). For Models without the correct livery, it seems that MSFS will fallback to some other livery for the correct model. This be good to have the correct model, but you might end up with European regional airlines in the US. In case no model matches (even before checking the icao_airline value), it will use a generic model.

The ICAO database used by FligthAware and MSFS does not differ between engine types or winglet/sharklet variations of the model. Some airlines use different variants for the same aircraft-type, which can not be represented in MSFS.

MSFS only matches icao_model and icao_airline, but not atc_id (it is actually not even transmitted into the Sim), this results in missing special liveries for specific airframes.

What should be improved

There needs to be a way to use different variants for the same model (engines, winglet/sharklets) to match the actual airplane operated on this flight. While using the FA Database for this is not enough the usage of the atc_id in combination with a new database that has the needed information (like


Airplanes spawned parked at the airport, do not have the front gear deployed, but they do depart. At the time of pushback the front gear gets visible.

Airplanes only spawn shortly before takeoff in the Sim, leaving the airport quite empty since no turnaround airplanes are shown in the game.

What should be improved

Since the front wheel issues has been there since some time, and the corresponding issue with not departing AI as already been fixed, this should get fixed soon as well. Overall, it is an immersion killer at the moment.

A better usage of the data provided by FlightAware would allow to spawn every airplane that is currently at the airport within the Sim. In combination with the atc_id the correct airplane at the airport would get the then departing flight assigned.

Offline AI (SimConnect based)

In this section we will focus on features specific for offline AI that is injected via SimConnect.


Airplanes spawning enroute are not spawning at the correct altitude they should have at the stage during their flight. The spawning altitude is related to the user’s airplane altitude, meaning that if the user is on ground all airplanes will spawn at altitudes around +/- 500ft of the user. Resulting in fast low fliers in the reality bubble. Combined with the issue that AI airplanes tend to not climb correctly they will remain at the low altitudes.

SU9 introduced a new major issue when it comes to following waypoints. The major benefit of using SimConnect based AI is that we can provide a flightplan with different waypoint to each flight, but since SU9 the airplanes would simply turn around and fly straight back to the first waypoint in that flightplan even if they are injected between point 11 and 12. For AIGFP based AI Traffic we have added a workaround for this already.

Besides spawning issue for enroute traffic we do have a minor issue on the ground as well. Compared to older Simulators MSFS does not block the injection of new airplanes when there are no free parking spots left at the airport. This results in “flickering” airplanes at the airport. We might be able to add a workaround for this to AIGTC.

AI System

Go Arounds

Airplanes send for a go around do not start a second landing attempt. They will fly circles around the airport. Workaround added to AIGTC already.

Rollout after landing

A major issue at busy airports at the moment is the fact that AI airplanes are breaking down to 10-15 kts after landing and then taxing with that speed to the next taxiway. This results in long blocking times of the runway and causes go arounds for the following airplanes.

Flight Model

Most of the flight model is still hardcoded and far from the quality we have seen in FSX.


The overall taxi speed is far too low compared to real world operations and is causing long taxi times when following an AI airplane. Airplanes taxing at the airport and required to stop by ATC (or when reaching the end of a queue) are starting “spinning”. 


AI airplanes turn directly after takeoff to their next waypoint, resulting in low flybys at the airport.


The bug that called regular to expedite the climb/decent has been fixed and ATC now handles AI airplanes as it does the user airplane. The ATC system is still far from perfect but is now a small improvement over FSX.

Sometimes at airports the ATC seems to give priority to vehicles and not airplanes when it comes to usage of a taxiway. (See more in the airport section)


Well, this could be an own report on its own, but since it has a major impact on AI I will add here the major bullets.

Most airports designed by the generic airport algorithm are far from suitable for Ai use. Overlapping parking spots, missing hold short nodes, missing vehicle networks, missed placed spots and so on.

All these small issues have a major impact on AI immersion at the airport and need to be fixed before getting a fully working AI system there. With the increase of live Ai traffic, we will see more and more airports that have issues and need to be fixed.

A major downside is the lack of big parking spots for the A380 at Dubai for example. ASOBO has decided to reduce all spots by 6m in Radius to add the ground services, with already not many default 40.1m spots in FSX or other Sims, we now have almost none for the A380.

With each new update and 3rd party content the bad practice design for airports is getting more and more common and new “unusable” airports are added to the Simulator.

Similar Posts


  1. Perfect and thank you !!!

    Hopefully ASOBO is reading along. And takes AIG finally in`s boat.

    AIG keep up the good work and don’t get discouraged.

  2. It is a mystery to me why Asobo do not use the data from the published SIDs, STARs, and Approach Procedures to control the flight paths of AI aircraft taking off and landing at airports, or is that just too obvious?

Comments are closed.