fnerg
01/18/04 04:12 AM
216.254.19.50
|
Hey, I'm working on my programming portfolio, and I've decided to try to make a basic Battletech game, in the vein of Ralph Reed's Mechforce, and I'm wondering if anyone knows how the AI worked for that game, or failing that, if anyone has any tips for a good AI. If I wanted to be lazy, I could probably just make it run towards the nearest enemy and fire off it's most damaging weapons first, but it'd be so much more fun to have it look for good terrain, and make decisions about which weapons to use in which situations. Any help's appreciated!
|
tgsofgc
01/18/04 10:54 AM
67.4.202.159
|
While I know nothing of how previous ai worked here is a suggestion, though it might not be worth much.
One good way would be to set up a layered selection and have the ai decide which "type" of action to take based on how much damage it thinks it can do and will recieve. For instance 1 action type might be a full alpha strike charge, another walk toward most advantagous terrain (such that its to hit modifier is as high as possible), etc. If you have enough time and the code can handle it with out loopt time outs you could set up between 20-60 of these and then set one instance at the beginning that compares:
AI's Mech's mass and Target mechs mass (+- some guess error, so that the computer messes up sometimes)
AI 'Mechs total possible damage and target mechs total possible damage (+- some guess error)
Ai 'mechs heat generated and the ai mechs heat it can dissipate (+- some guess error)
This would also allow you to put in a handy danady difficulty setting for ai controlled 'mechs that is proportionate to the guess error. Namely the smarter the ai the more likely they will make the right guesses to make the best possible decissions (so the smaller the guess error), or the opposite for dumb ai.
just my take, hope you can understand me and sorry if thats no help
I find that 'pinpoint' accuracy during a bombing run increases proportionally with the amount of munitions used.
-Commander Nathaniel Klepper,
Avanti's Angels, 3058
|
fnerg
01/18/04 05:37 PM
216.254.19.50
|
That actually makes some good sense. One of the things I distinctly remember Mechforce not having, was the concept of fire support. It'd just send everything you gave it willy-nilly towards the front, but I think if I use what you were talking about, I could start to see some emergent sort of fire support thing happen. I really like the idea of adding a bit of error to get a sort of difficulty level going.
On of the other ideas I had was to assign each weapon a class, like LRMs as long range, Medium Lasers as close, etc. and see how much damage a mech could do within those classifications, then have the mech try to keep to that range. Reading over what you're saying, it might be a better idea to give each range a weight (like a like/dislike value) and modify those weights according to the things you were talking about, so we get a paper-rock-scissors sort of thing going, then within those ranges, have the computer search for the most advantageous terrain.
|
tgsofgc
01/19/04 02:33 PM
67.4.193.169
|
Well theoretically each type of 'mech could have its own set of questions, then mechs are sorted into a class based on things like damage at range class, armor, movement, etc. But I quickly see this escalating, hope you have a while to invest in this.
I find that 'pinpoint' accuracy during a bombing run increases proportionally with the amount of munitions used.
-Commander Nathaniel Klepper,
Avanti's Angels, 3058
|
Iceman
01/19/04 03:35 PM
217.129.240.212
|
Quote:
On of the other ideas I had was to assign each weapon a class, like LRMs as long range, Medium Lasers as close, etc. and see how much damage a mech could do within those classifications, then have the mech try to keep to that range.
Define range classes like Close (1~3), Short (4~6), Medium (7~18) and Long (19+). Weapons with min range shouldn't count towards Close range, IS LRMs to Close and Short. You'll have to make some judgement calls for some of the weapons' ranges to "fit" them into these categories, probably "rounding" to the nearest limit.
Or make a full damage table for each hex up to 30 and determine at what distance(s) potential damage is maxed (counting min ranges).
|
fnerg
01/19/04 06:55 PM
64.122.126.73
|
I'm giving myself a couple of months to do this. (I'm between jobs right now) It's not going to be graphically spiffy, no fancy 3d effects, just hot mech-on-mech action, and maybe an event system sort of like Dynasty Warriors on the Playstation. (F'rinstance, in a scenario, if you kill a certain NPC, you get some dialogue, and part of the map changes, like, "Argh, I'm dead, but you can never have the secret plans!", then a building turns to rubble, and anything in the vincinity takes x points of damage to each hit location) This is why having decent AI is important to me, as I'm actually more interested in the single player experience than multiplayer, (though they way I'm thinking of doing this should make multiplayer relatively easy), or networked play.
My pie in the sky ideal would be to have a seperate map/event edtor, and a simple way to set up a centralized database/repository for scenarios, but I'm not sure if I can get all that done in a few months, so I'll be happy with just a game that allows you to pound on a few decent AI mechs.
|
|
|
Extra information
|
0 registered and 84 anonymous users are browsing this forum.
Moderator: Nic Jansma, Cray, Frabby, BobTheZombie
Print Topic
|
Forum Permissions
You cannot start new topics
You cannot reply to topics
HTML is disabled
UBBCode is enabled
|
Topic views: 4852
|
|
|
|
|
|
|