EggPlant AI ramblings

[Berg]

I have been watching this thread and have a question
Are you leaving the pathfinding to the game or are your going to build it into your AI

[aubergine]

I'm leaving it to the game for now, but eventually I plan to augment it with my Map/Sector AIs.

Each sector will have descriptive "safety" data - eg. is a sector unsafe for land units, air units or both? Does it have a repair facility/droid? Are there VTOL reaming pads there? Does the sector have ECM, CB, etc?

My plan is to have the AI, given chance to do so, start fortifying sectors so that I can make my groups try different "safe sector paths" before handing over to the WZ pathfinding. There are lots of stumbling blocks however, especially for land units. For example, the JS API doesn't expose continent data so it's difficult for me to work out how continents fit together or where continents overlap. I'd need that sort of info to make AI pathfinding more feasible.

[Berg]

Well I ask this because that is one thing the AI is faulty ingame with keep up the good work

[Rman Virgil]

.

I was going through my archive of papers and A.I. software related to Artificial War and while re-reading a paper called "SIMULATING SMALL UNIT MILITARY OPERATIONS WITH AGENT-BASED MODELS OF COMPLEX ADAPTIVE SYSTEMS", I came to the following and thought of EggPlant A.I.:

......The distillation models and their more detailed cousins are based on goal-directed agents, sometimes referred to as intelligent agents, that perceive the condition of their environment and select their behaviors to effect changes to that environment. Such agents, also referred to as intelligent autonomous agents, are:

* goal-oriented - able to build courses of action by taking the initiative to change elements of the
world state to desired objectives,

* perceptive - able to receive data from their environment, including knowledge of their own state and that of other entities of interest to them,

* active - able to perform actions affecting their environment, and

* autonomous - able to initiate behavior sequences based on internal logic to determine what is appropriate given the perceived environment.

Agents representing combat forces must also generally be:

* mobile - able to move around in their simulated environment,

* capable of inferring the intentions of others, the desires and plans of other agents, and

* social - able to share goals, cooperate with or coerce other agents.

.......................... Complete Source Paper HERE

This single paper alone is also a fine snap-shot of an entire field you may find of interest. It also contrasts top-down and bottom-up modus operandi.

- Regards, RV.
.

[Rman Virgil]

.

Ditto.... (As in connections to your EP a.i. design goals.) Situation Awareness

(In addition, there are essential correlates here to exactly How WZ Creators, Pumpkin, originally envisioned - & initially developed too - indispensable Commander use in MP game play and, pragmatically, it's far greater feasibility NOW than in 1997-1998. But those details and discussion are for another topic, another day.)

-

.

[cumandgetit]

really interesting concepts even if my brain can't wrap around it all. lol cool of you to post. tanks.

[aubergine]

Currently, this topic is giving me a lot of food for thought: viewtopic.php?f=42&t=8878

[cumandgetit]

Currently, this topic is giving me a lot of food for thought: viewtopic.php?f=42&t=8878

yes it's interesting but in our lan club we like no unit previews.

we also play with simple mod that adds to game start pre-reqs for ea player like this: a CC & 2 trucks plus 2 uavs made of bug/sen vtol scouts.

we reason that not having uavs is really really dumb for future post nuke appoc hi-tech game world.

so far gameplay is very good. happy end of story. k.i.s.s. (not lips - design. lol)

[Per]

we also play with simple mod that adds to game start pre-reqs for ea player like this: a CC & 2 trucks plus 2 uavs made of bug/sen vtol scouts.

Sounds interesting. Is that mod available anywhere?

[stiv]

plus 2 uavs made of bug/sen vtol scouts.

Seems to me, given the importance of air surveillance in modern war ( things like AWACS and JSTARS), VTOL sensors should be a default part of the game.

[aubergine]

IMHO VTOL sensors could be too much of an advantage even for a newbie like me. I could just have them patrolling and not need to worry about building radars.

I do like the idea of a dedicated scout unit though.

In terms of that post I linked to, it was more the insight I'm getting in to how the pro MP gamers adapt strategy to their opponents, something I want to try and make EggPlant do. It's one of the reasons I'm spending quite a bit of time thinking about the research tree and how I might achieve an AI that adapts it's research path based on what the enemy does. All other AIs I've seen tend to just use a single, predefined research path and I think that's one of the reasons why they are less effective opponents.

[cumandgetit]

@ Per: not up for dl. we see if l8r. gotta ask 1st.

@ stiv: yippy-ky-yea !

IMHO VTOL sensors could be too much of an advantage even for a newbie like me. I could just have them patrolling and not need to worry about building radars.

how you figure an advantage if all start the same - newbie to "pro" ?

and Tom Cadwell:

"Just because a designer expects something to work a particular way, doesn't mean it necessarily does in the game..."

speculating without reality testing is not good enough here. the reality plays not at all like you suppose.

btw, where i sign up to make money playing WZ mp ? lol. if no money then "expert" is the right word.

an expert mp player's research choices and build order when they are adaptions to opponent's play - and not rote rush tacs, of course - are based on scouting (or bunk unit preview) so we are back to square one - scouting.

i get the concept of dynamic, reactive, research you are seeing in that thread for eggplant a.i. & think it links back to scouting / situation awareness & again, uavs.

sometimes small simple change works and complex analysis & designs miss the mark wide.

[Emdek]

Maybe those UAVS could be available after some additional research (after some sensor upgrades) and could be kind of specific unit like transports, but not VTOL propulsion but something more primitive (so it won't depend on VTOL research - if we have them already we mostly don't need special unit then), even maybe as slow kite or balloon.

[Rman Virgil]

............

speculating without reality testing is not good enough here. the reality plays not at all like you suppose.

..........

You can only appreciate that if you embrace actual RL experiments, perform them in the spirit of the Scientific Method, play extensively with changes in MP games and after all that, speak with assurance the results. That's not the default protocol here so, again, your whistling Dixie in Yank town - getit !

.

[cumandgetit]

speculating without reality testing is not good enough here. the reality plays not at all like you suppose.

You can only appreciate that if you embrace actual RL experiments, perform them in the spirit of the Scientific Method, play extensively with changes in MP games and after all that, speak with assurance the results. That's not the default protocol here so, again, your whistling Dixie in Yank town - getit !

meh. yu know me. i pass experience some see only idea, nothing more i can say, right ? we play our games, have fun as they are. changing minds, not my goal. what can be done, we know, we do. happy ending. fighting about ideas, bah. just do it & find out, i say. yu know it's only hard for lazy guys. lol