Allegiance Player AI

[ThePhantom032]

I've thought about adding (clever) AI to allegiance before and came to the conclusion that AI-players would - just like humans - need different skill levels.

Some would be better at flying bombers, some would be better at intwhoring and some would be better off probing
etc.

QUOTE I honestly think the commanders in Allegiance tend to do the same thing which is why AI comms would work brilliantly against them because AI comms can do opponent modelling where it can take an educated guess on what the opponent will do and it can model that on any number of commanders.[/quote]
+1

I actually planned on beginning this after my last exam on september 6th, starting with a (client sided) DM-bot for beta as 'AI-programming practice'.
Drop me a PM or tell me on IRC if youre interested in what i thought of/working on this together

[Psychosis]

You are ALL working to hard on AI commanding, its very simple, here is a basic logic tree

1) what faction is the other team? what is a good counter?
2) check settings, is there a clear faction advantage from settings?
3) does this faction have cheesetech? if so, buy it!
4) what techpath is the enemy? if so, counter or steal it!

you could program a stupid AI easily, It would be predictable and boring, but If you give it some solid cheese/ strategies to run off, that might work. Do not attempt to program randomness or anything, just program solid boring reliable commanding

[HSharp]

Psych if an interface can be set up for an AI commander to issue orders and receive data then anyone can make their own AI commanding bot which can be simple or gloriously complicated.

[Psychosis]

QUOTE (HSharp @ Aug 21 2010, 12:23 PM) Psych if an interface can be set up for an AI commander to issue orders and receive data then anyone can make their own AI commanding bot which can be simple or gloriously complicated.

I can't wait for the release of HAROLD3bot

[raumvogel]

QUOTE (Psychosis @ Aug 21 2010, 01:10 PM) You are ALL working to hard on AI commanding, its very simple, here is a basic logic tree

1) what faction is the other team? what is a good counter?
2) check settings, is there a clear faction advantage from settings?
3) does this faction have cheesetech? if so, buy it!
4) what techpath is the enemy? if so, counter or steal it!

you could program a stupid AI easily, It would be predictable and boring, but If you give it some solid cheese/ strategies to run off, that might work. Do not attempt to program randomness or anything, just program solid boring reliable commanding

This will help newer commanders to learn the tech tree in different situations.

Or we could wait for aptest's decision.

[Andon]

There would have to be some randomization involved, as there are always situations where there are multiple "Decent" options. Plus, in an AI vs AI game, they'd both be waiting for the other to do something.

[AaronMoore]

QUOTE (HSharp @ Aug 21 2010, 07:45 AM) sounds like a nice project, I'm interested in AI (did my dissertation on it) so sounds fun, if you can get the interface sorted I wouldn't mind helping out with the bots.

Thanks HSharp, I'd appreciate your help. I did a couple of subjects on AI but nothing major. I'm reading up again and hope to put some proof of concepts together over the next few weeks

QUOTE (Phantom032 @ Aug 21 2010, 11:42 PM) I actually planned on beginning this after my last exam on september 6th, starting with a (client sided) DM-bot for beta as 'AI-programming practice'.
Drop me a PM or tell me on IRC if youre interested in what i thought of/working on this together

Thanks Phantom, I'm definitely interested in working together...
I probably wont have anything substantial to show that soon, but I'll drop you a PM to run some ideas by you

QUOTE (BackTrak @ Aug 21 2010, 08:56 AM) I would think that an AI commander mode game that would have AI commanders on both sides would be a game creation option at the lobby level. Then you don't have to worry about the GC issues. You just set it all up from the lobby and let fly.

Thanks, that's a good idea!

QUOTE (Adaven @ Aug 21 2010, 02:02 PM) improving the AI would be a useful step towards better training missions or maybe even a single player 'mini' campaign. Humans v Humans will always be the main draw, but I think this couid be used to help lessen the learning curve and encourage more people to try the game.

Yeah, more game-like training missions would be a real possibility with this. A single player 'mini' campaign would also be awesome!
But you're right, Human v Human will be unmatched in the come-back glory/frustrating decay that we love in Allegiance.

QUOTE (DasSmiter @ Aug 21 2010, 02:30 PM) The interaction between commanders on different teams and between commander/team is so central to the game that if it was reduced to a series of if/thens I think I might leave if either team had one.

Besides the biggest problem with AI is that you have a commander that does the same thing all the time. The biggest issue for any single player RTS experience is that it's skirmish can be reduced to showing the ai a series of stimuli to force it into a certain direction and then simply counteracting that.

Yes, I think it will never be a substitute for a real, experienced commander, and you're right, the interaction just won't be the same. But I hope that this will allow us who have to play in off-peak times, the option to have a classic conquest game of Allegiance, without the pressure to command a team, and to have more than 3 players a side.

QUOTE (DasSmiter @ Aug 21 2010, 02:30 PM) In the end I think the commander AI would be a nice side project for someone interested in AI and whatnot, but given the choice between having a dev work on that and having a dev work on solving Trac tickets I'd straight up choose the tickets. Of course you're your own person and can do what you want, but know that I think your efforts are futile and wasteful

thanks Das, I also think it will be an interesting project.

QUOTE (pkk @ Aug 21 2010, 06:09 PM) Hint: PIGs

THanks pkk
KGJV pointed me at these a little while ago. I've read through the code, and it seems like it will be a good starting point, but I wasn't able to get them running on my local server.
Have you had any luck getting them running?

QUOTE (Psychosis @ Aug 22 2010, 03:10 AM) You are ALL working to hard on AI commanding, its very simple, here is a basic logic tree

1) what faction is the other team? what is a good counter?
2) check settings, is there a clear faction advantage from settings?
3) does this faction have cheesetech? if so, buy it!
4) what techpath is the enemy? if so, counter or steal it!

you could program a stupid AI easily, It would be predictable and boring, but If you give it some solid cheese/ strategies to run off, that might work. Do not attempt to program randomness or anything, just program solid boring reliable commanding

Thanks Psychosis, that's right. Once we have the interface avalaible to script up a Commander AI, we should definitely start with something simple.

Thanks!

[Psychosis]

I still want to program the Harold3 bot, and then force an unwitting team to fly for it.

I think that while the idea is good, before hard code is committed ALL the details are thought out and hammered over

[WolfDeSol]

you might have to be careful, if you have task that an AI pilot would do (possible better than the average voob), then some players on the team might see that a AI player is on the team and think "oh an AI, well that means i don't have to probe anything, the AI will do that for me"

[VenCain]

I remember doing pigs once. It's been way too long ago but I think you'd get a bot each time you issued the command to bring one in. I think I had to hack the server to accept the command. I also seem to recall sabretooth getting up in arms on IGN thinking we were doing something underhanded to stack stats (I was an FZer but this was when there were free AZ days) while we were playing around with them. Long, long time ago.

-VenCain