One of the disappointments of Infinite Flights is that out of several thousand available airports only a handful have active ATC at any one time. This is understandable since our volunteer ATC are not paid for eight hour shifts and don’t wish to service an airport with little traffic.
In addition GA pilots are sometimes discouraged from using the busy airports where ATC is most likely to be active, and thus miss out on the ATC experience.
A number of other games I play have interactions with AI avatars. AI Controllers could solve these problems. Imagine if every major airport in the world had an active AI Controller ready to serve 24 hours a day. Very busy airports like KLAX might be too much for an AI to handle, but for airports with only one or two planes at a time it seems possible. I am not suggesting that AI controllers replace live controllers, rather that an AI controller be available when a live controller is not.
This idea was mentioned by one pilot in 2019, but not pursued.
1. To request things, please use the #features category. Please note that you need to be Trust Level 1+ to vote, and Trust Level 2+ to create a feature suggestion
2. The thing is, AI can NEVER properly interact or verify with humans, as there are scenarios that you cannot code into it. Sometimes theres human error with a pilot accidentally turning the wrong way. How would AI understand this? How would it be able to cope with high traffic such as at JFK today, along with some situations, such as someone low on fuel. The reason AI isnt used in game, or in real life, is because it is impossible to simulate humans. I would rather have a few actual ATC people, than every airport having some dodgy, impossible to work with, AI
The ATC environment is far too situational to be of any use trying to simulate AI in any sort. Even if everything operated consistently to the point it is impossible for a controller to act on any given situation, it would still take an unimaginable amount of resources to develop anything AI related, given its difficulty to develop including the certainly going-to-be confusing algorithm that goes into the machine learning part on its own so the computer knows how to improve. This does not even include the resources that would go to running many generations at a time thousands of times; all of that for an okay-ish-ish AI controlled ATC system. Personally I can’t say it’s worth the hassle.
Thank you, @NemAir for your response.
I did attempt to submit this post to Features for approval. After two attempts and 48 hours without a response I decided to post to general.
In my post I acknowledged that AI probably could not handle a very busy airport with numerous planes in a holding pattern. My suggestion is for airports that are not so busy, one or two planes at a time, that now almost never have ATC service.
For such less busy airports I think AI could handle the current set of statements available to live controllers.
What about starting with AI centers? Should be less situational and simple won’t it? It’d be so great to have “busy comms” 24/7 all around the globe! It will reciprocate the uniqueness that is IF to always fill the skies and our flight experience with soul :)
Thank you, Superdogs, for your reply. I was not aware that IF once had AI controllers. I agree it would take effort to implement, but don’t all worthwhile new features? AI would need to select an active runway, and use the statements currently used by live ATC. For one or two planes at a time, I think it could be done.
Center does not simply just sit there waiting to respond to radar contacts. It still has to account for many situations such as separation maintenance, covering as a radar controller into airports that don’t have such coverage, the preference-based nature of hand-offs while working with human controllers in general, et cetera. Even that is unprecedentedly difficult on its own.
Never has a feature of this sheer complexity been done before. Cleverbot was launched in 1997 with over 60 million conversations that it has had the opportunity to work from, and still requires a great deal of patience to interact with decently. Imagine this with ATC, I personally find it unlikely that I will have more patience in the cockpit trying to talk to this theoretical AI than I do for Cleverbot.
Just to be clear and to avoid confusion, Infinite Flight has never had AI ATC. Take it from someone who’s been here before the addition of “live” multiplayer. Unicom was used before ATC was implemented.
I agree that limited ATC is a problem. Currently, we have active ATC at very few airports at any one time. Pilots in large parts of the world almost never have active airports. I agree that many of the several thousands of IF airports almost never have traffic. But there are many airports that considerable traffic throughout the day, even if it is only one or two planes at a time. These are where I think AI controllers could perform a valuable service. I think increasing the availability of ATC is a good idea, and it won’t be advanced by live ATC.
Most of the players in the expert server go to/from the main controlled airport anyway. If players feel like their home airport deserves to be represented, im sure whoever organises the ATC schedule is happy to take suggestions
I understand your concerns about the complexity. If ATC in IF involved verbal communication such as in the real world, I would agree AI controllers might be impossible. However, in IF the responses of both pilots and ATC are strictly limited. As long as the AI controller deals with one plane at a time, I think it is doable.
what would happen if the traffic at the airport increased? would IFATC be able to staff then? will there be a limit to the number of traffic there for the IFATC to staff the airport and the AI to disengage with traffic?
That doesn’t change anything considering the complexity in the current system lies in what is the correct command to send at the correct time which to a computer is ridiculously difficult to explain as it is too arbitrary and situational to just give the instruction and make it do its magic. If we somehow want to be aspirational and try it it will take a ridiculous amount of resources to let it develop and learn on its own.
I think the idea is excellent, to be honest, but it’s not realistic, at least not in the world of infinite flight.
AI ATC needs a highly organise airspace, that requires highly professional pilots, although I hate to admit it, but unfortunately, I think we both know how poor are the quality of the majority of players in infinite flight; yes, I’m talking about Expert Server. As said above, developments are also a primary key.
As for now, I think the biggest thing or the biggest concern is how to improve the quality of the players, and how to make our controller life more easier; it’s still a game, the bottom line is how to make it fun and enjoyable, rather than after finishing every session by hitting my device million times.