What are AI people working at the moment?

I'm not seeing very much discussion on AI scripting (in JASS) here or on other boards. I'm curious about how many AI writers there are and what projects they are working on. Let's start from me.

I just finished updating a custom campaign AI made in the TFT AI editor as follows:
1. The AI now attacks certain target buildings in the map (instead of a random enemy town).
2. The AI now changes the target location when certain triggers run (e.g. it does not attack a stronghold gate, if the gate is opened through a lever).
3. The AI now continues its attack when one target is destroyed. It does not return home before launching a new attack, as the AI does normally.

Cheers,

Tommi

I am still continueing to work on AMAI, of course :)
Apart from that I am writing some AIs for Return of the Dragons.

AIAndy: I know you are very expirienced AI scriptwriter. It is possible to ask you about the favor? I'm going to write monsters and mercenaries AI for 'Graveyard' map - but I'm not so expirienced in AI scriptwriting. Please answer some questions...

As for mercenaries I guess their AI to be written as .j - triggers. As for monsters - their AI should possess following features - monsters are created dynamically during game, and in plenties. They should be able to use the their custom abilities, escape when have low health, avoid rockets and grenades... Many of these features require internal core function calls which is impossible from .ai files. What can you advise me?

@AIAndy: I'm wondering if it is due to the large amount of work required in that project, or the difficulty of AI scripting that Darky27 asked you to write scripts for the campaign? As Darky27 definitely knows JASS...

Cheers,

Tommi

@Tommi: The main reason that I am writing the AI scripts is that it is much easier to write them for me since I know how all of common.ai and the special AI natives work and I could use the AMAI common.ai modified to support Campaign scripts. Of course it would have been possible for Darky to learn that but it would have costed time and in such a large projects it is a bad idea to do everything yourself even if you could.

@Cacodemon: If you have no use of the special natives the AI provides which deal mainly with building and using one attack group I'd rather stay in .j . The usage of custom abilities depends a lot on what they are based on and there is no way to change that hard coded behaviour so you do not need AI scripts for that.

Thank you. But it won't be easy to code...

Yes, it is a pity that there is nearly no way to influence unit AI. In the last versions of AMAI I have tried to put in some micro of the units and it is very hard to avoid using up too much performance.

I am currently working on a melee ai, trying to model my own playing style basically. Its currently pretty bad but eventually I want to put in some advanced micro.

Andy- what kind of micro did you get to work in 1.8? Its hard to notice the computer's micro just by playing against it, and I havent looked at the source.

You could do something like loop through all ai ranged units, and have them focus on whatever target in range is the weakest. But that's order units^2, so I can see how performance could be a problem.

I am appointing a major hero for each attack. Around that major hero I do some situation analysis. I loop over all units in a certain radius of that hero. Among other things I check the strength of ally and enemy army, so he doesn't flee when he is actually superiour (I just switch off fleeing in this case since you cannot command him to flee or stay). I also determine the weakest enemy unit and make the units focus fire on this unit. Also focus fire on heroes is controlled in this way. I also micro units that are below a certain health away from the battle to some healing source.

I'm working on a triggering method to hijack computer control from units so that I can make them smarter. I don't like that the blizzard comp has to use cheat resources to win, so instead my idea was to make it much smarter. I'm working on calculating unit value or BE based on hp, armor, damage, attack speed, and then using dimesnsional functions to assess temporary unit value when a unit gain the ability to attack again this is all evaluated and the unit attacks are coordinated to kill of weak units, shoot low-lvl heros, and attack berserker units over meat loaf units. I still have alot of work to do to make them operate a whole game, right now it just makes individual battles hard as hell to win, it doesn't know how to maximize unit production efficiency yet but I will have that too soon enough.

Nice! I have been developing a unit value system for 10 months now, and it's pretty advanced. It's got nice mathemathical foundation behind it, combined with some more judgemental aspects. I used it to balance monsters in the City of the Seven Mithril Golems. It's pretty long to tell how it goes and why, so I don't do it in this post.

I am working on an academic paper regarding whether WC3 is useful for research on human-level AI.

I'm researcher, too. Human-level AI... huh? Sounds pretty hard to code with JASS. :p If your funding covers scripting AIs for your maps that sounds pretty good. :)

I am actually making my master thesis and haven't even graduated yet.

What is the area for you research?

Nothing to do with AIs, though. 8)) I'm working with the evaluation / management of national technology programmes (in Finland), organizing workshops, for example, to help in the commercialization of research.

I don't know, but it seems that I like to get to work with the hardest stuff...