Learner - reasons for 5 minutes minimum interval?

[eddl 11]

Hi,

the forum search did not help me, but it might have been answered in the past already.

Is there any reason for the Learner to be limited to a minimum 5 minute interval between airports? I think I could collect way more data with an interval of 1 minute.

Still trying to understand the underlying algorithms to get the best output from of the learner.

Many thanks in advance! 🙂

 

[kiek 1,213]

Yes there is, not technical, but what are the chances that you capture an aircraft arriving at a gate and the pilot switching off its transponder,  or an aircraft that has set its transponder and is about to leave a gate,  if you monitor less than 5 minutes minus your delay time (default 30 seconds).

It was just to protect the user from making the wrong choice. It could be set maybe a minute shorter, but not much.

I have not done that but it would be interestiing to see what captures more, two airports each one hour in sequence or one hour but alternately each only 5 minutes. And that in the same time frame (you need a professional RT license for that). And you have to repeat that a few times for different hours  to get valuable results.

I never go below 15 minutes.

Like I said, there are no technical constraints. Over here I have a special version of Learner  called Learner_Crawler that changes airports every 5 seconds...But his job is to read the RT data stream only once (full lateral width) and to gather real live info about registration codes+airlines+types. You can find that info in data\dbase.xml. it can crawl the whole world (major airports) in an hour. I let it run every now and then to capture new aircraft and airline changes. It's like the  Google crawler.

 

 

Edited January 31, 2023 by kiek

[eddl 11]

Thank you.

With US airports the stream snapshot could be quite effective, as I often get several updates within the first seconds of scanning.

On some European airports this will work too, but not as effective. I have monitored the update counter while continuously pressing +1min in RT Pro. Often a lot more updates within a short timeframe that way.

Of course learning a single airport 24/7 for a week gives the most output, but it takes two years for 100 addons... and often quite frustrating with smaller airports like Skiathos, requiring a whole week of learning catching 30 movements in total.

Quickly browsing though RT historic data can be more effective, but of course it is manual work and not 'set and forget'.

But I get your point.

[kiek 1,213]

3 hours ago, eddl said:

Of course learning a single airport 24/7 for a week gives the most output, but it takes two years for 100 addons... 

Normally people do not fly at more then 10 airports. And they fly at certain hours of the day only. Thus you do not need to let it learn the whole day.

Besides, not everybody wants as real as it gets. With a few hours learning an airport you will already see aircraft that fly there and get an impression of immersion.

There are airport files available at my website too.

We could ofcourse more share our airport files. The problem is however who is setting up and maintaining a library of airport files for commercial addon airports? There were some trials in the past, but they all did not succeed.

 

Edited February 1, 2023 by kiek

[eddl 11]

Currently learning a LOT of addon-airport files (200+). Not planning any library at the moment, but I will happily provide them later in one batch to have them included in the PSXT installation if you like.

[kiek 1,213]

1 hour ago, eddl said:

Currently learning a LOT of addon-airport files (200+). Not planning any library at the moment, but I will happily provide them later in one batch to have them included in the PSXT installation if you like.

Be careful you separate default  airports, Asobo airports and third party airports, because they all have the same file name.

[eddl 11]

Thanks for that. Will have to manually sort them anyway as some need to have the developer added manually to the file.