[Flightgear-devel] Some AI related notes

[Durk T. <d.t...@xs...>]

Today I finally got around to commit a series of patches / changes to the AI 
system. Thomas Foerster contributed some new code that will allow us to 
specify aircraft performance data, which will allow runway length 
calculations on a per aircraft base. He also contributed an implimentation of 
the "Dijkstra" route finding algorithm, which outperforms my own make shift 
algorithm by magnitudes.

In the mean time, I have been working on detecting and resolving higher order 
interaction problems, in the network. A higher order interaction, 
labelled "circular wait" by me is a situation where aircraft a waits for b, b 
waits for c, and c waits for a. Eventually, a situation like this will block 
the entire traffic pattern at the airport. So detecting and resolving is 
crucial. The detection code works, fairly well, but the resolution is 
currently rather crude: remove one of the offending aircraft from the scene. 
Finding a better solution is necessary, but that will take time.

Today, I got my head around implementing different runway assignments based on 
traffic type. There was already rudimentary support for this, but everything 
was defaulted to one type of traffic at the front end. 

Finally, Innis Cunningham sent me a nice traffic demo for the San Francisco 
area, which I have just committed. There's an interesting variety of 
aircraft, right now.

Cheers,
Durk