Research and the AI.

I have been working for quite a while to get the AI to research in a semi-intelligent way. The problem is that i hardly understood how the research mechanism worked for the AI and was basically blind shooting. Now i understand it a lot better and can share some of my discoveries.

Basically, the AI research in a sequencial way. I will research 1 item then the next. But each turn it reset it's research. To continue researching what it had the previous turn it use a list that pinpoint where it was last turn in the research queue. The only exeption is when it run on a tech it can't research yet and will try to queue up the requirement if some research "slot" is free (but most of the time, there is none). All this make the AI very predictible and totally inflexible. It will try to research things that it doesn't need yet and won't adapt to changing situations.

I'm currently working on a way to make this a thing of the past. The problem is to permit the AI to loop on it's research queue and take different things from turns to turns depending on it's current priorities. Patching the old system won't help as making conditionnal research just screw it up. It remember where it was the previous turn, but if the line was conditionnal and it happen to be false that turn, it will just jump to the next entry and research it instead. It will jump back to the previous research once the condition is true, but if the other item is fully researched by then, it will never resume it. And once it reach the bottom of the research queue, it never go back to the top. Once the last item is researched, it just stop there.

In the stock code, there are 2 routines at the end that will make the AI research things in alphabetical order until the all reach the maximum level. Again it's not helping the AI at all, it just make it spend it's research points on something...