7 minutes ago, jabloomf1230 said:

The rules of ATC communication are myriad but they haven't changed that much over time. Since they are all written down somewhere, it just takes perseverance to code an AI ATC app, not any fancy neural network computer, like which might used in designing the algorithms for self driving cars. The former is just a massive collection of IF and CASE statements. The latter has to process sensor data in real time. Is that an object? Is it a dog? How  does it move? Etc. 

Also those known rules cater for very rare non-normal events, which probably won't occur often enough for machine learning to assimilate.

Using an IF THEN ELSE is the old way of doing it. I am talking about a predictive model that would take decision based on what it has learned.  Machine Learning is perfect for that kind of application. I am surprised that it doesn't exist already. 

2 minutes ago, fogboundturtle said:

Using an IF THEN ELSE is the old way of doing it. I am talking about a predictive model that would take decision based on what it has learned.  Machine Learning is perfect for that kind of application. I am surprised that it doesn't exist already. 

IF THEN ELSE is still the way human controllers operate, I wonder if you would you advocate real ATC being passed over to machine learning?

Just now, kevinfirth said:

IF THEN ELSE is still the way human controllers operate, I wonder if you would you advocate real ATC being passed over to machine learning?

Let me give you a different example that might make more sense.  In the radar detector world, since the first radar detector, people have been using the analog way of detecting a signal. If signal is within this KA or KA band therefore beep.  It was very effective until car manufacturer started to use radar technology for blind spot monitoring an anti-collision system. Not only that but more city are using traffic signal radar on every  traffic light so they can study traffic pattern. This new problem created issue that it will cause false alarm within radar detector. Of course, radar detector company started to try to build analog filter to try to ignore frequency. Of course, even after doing that, False alarm is still plaguing radar detector. Until recently, a company called Radenso decided to use Machine learning to look at signal fingerprint and amplitude instead of just using frequency. Now only can they now detect effectively radar gun but they can also see the difference between a radar gun in the K Band and a Traffic radar on top of traffic light. This would not be possible as they would be overlap in just using frequency. 

Machine Learning predictive model is ideal for this type of application as yes it needs to know a set of rules but it's how it applies the rules by looking at all the traffic in a specific area and providing instruction. I am very surprised that one one has come up with something like this already. I am pretty sure the FAA is probably running some prototype in parallel not to replace human but to help them making better decision. 

1 hour ago, fogboundturtle said:

Like I said, I am too old for that now. I wouldn't want to start a project. 

I can try to explain to you in laments terms how Machine Learning works.  For each event that requires and action, You feed the equation the right answer.  You need to feed the equation millions of transaction that tells the AI what is the right answer. What the machine learning does is approximate the right answer.  The more data it has, the better it gets at predicting the right decision. So it's not your typical If <plane does this> visual basic logic.  The AI will determine how to control all the traffic in it's region based on what it learned.

Good luck with all that. But an ATC program is not just about controlling other aircraft. Its primary task is to monitor your aircraft and ensure you follow instructions issued all the way from getting clearance to taxiing off the landing runway to the gate.

Such as clearing you to taxi to the departure runway and half way there handing you over from Ground to Tower who then has to check for inbounds and whether there is enough time for you to line up and depart before the landing traffic gets too close. And you might be number 5 for departure.

Thats just one tiny element of what an ATC program has to do. That’s assuming you know the layout of the airport and weather conditions so instruct the pilot to the correct runway. How do you get all that data? Not easy is it?

6 hours ago, kevinfirth said:

Yes, here's how Voxatc voice config page looks like with many addon voices.  You can select pilot and controller voices, filtered, with static or as provided.  Gives a much more authentic experiences with a lot of variety.  I picked up all my Ivona and Cereproc SAPI voices when they had sales on.

Thanks Kevin, I’ve downloaded vox to have a look at, if I keep using it then I think these voices will be the way to go, maybe a couple to start with and look out for those sales you mentioned.

I noticed they have a novelty voice which is cheaper than the others “Nicole the sexy french voice”  actually listening to that it’s very realistic for french ATC.

5 minutes ago, jon b said:

I noticed they have a novelty voice which is cheaper than the others “Nicole the sexy french voice”  actually listening to that it’s very realistic for french ATC.

So is "sexy french" actually English in a french accent? Only ask because some of the European voices I have installed to p2a occasionally launch into their native languages! 🤣😂

 

G