Someone already build an Air Traffic Controller Action Extraction-Prediction Model Using Machine Learning Approach
3 Likes
Very interesting approah (no pun intended) and would definitely improve MSFS ATC. I like the idea.
It’s nice, but to computate it for 100-thousands of players simultaniously…
it doesn’t have too. Only for the region you are flying in on that server. ATC only control their airspace. Don’t forget that all NVIDA user have Tensor Core that can be used to do this or that Miicrosoft is a big provider of Machine Learning services on Azure.
This would, just as the ATC speak synthesis, be done online. And so Azure must calculate it for ALL players currently playing. I’m not saying it’s impossible, it just not (yet) probable 
Again, Just for traffic coming into that specific airspace just like real life ATC. Don’t forget that your client will only request data bout your area.
If you request it for all traffic in your area. And all online players do that. The server has to calculate it for all players. Since it will be done online, just like the ATC voice synthesis now is happening online also.
The bulk of the processing is done way before the end user interact with the machine learning. In the same way where all the procedural scenery is “baked” all in advance. That same needs to done to feed the Machine Learning equations. They feed millions of ATC interaction to build the predictive model.
I don’t believe an AI based ATC would be more taxing than current process and would be 10000 millions time more effective than the current design. There is also real life application in doing something of that nature. The link I gave you was an model for real life application.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.