Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Game AI
skipeth
Member Posts: 66
... continued from Possible Class Options thread.
Except for heavily scripted encounters in an MMO, the AI is at the individual level. Adding varied intelligence is definitely cool and will probably be quite distinct in fighting a zombie horde versus fighting a group of rocket scientists.
Groups of mobs can be chained together so that if one creature's state goes to aggro, they all go to aggro. Actions like heals and buffs that they perform are simply part of their FSM or tree of actions. They don't actually work together. So even though some enemies may perform actions to help the group, they are still acting as an individual.
I'd guess that AoC is using some rudimentary form of squad level scripting with their battle formations, but I haven't bothered to try it out.
Besides being the cool AI kind of stuff that I like, I think squad AI can add a lot to a game such as...
The seemingly same group of enemies can show different tactical behaviors based on the squad goals. This makes it more difficult for players to develop a standardized method for dealing with a specific group.
Taking out the squad leader removes the squad AI and the group resorts to individual AI tasks. This gives the players the added goal of trying to take out the enemy that is directing the fight.
Reduce dumb behavior like running from a position with tactical advantange to your death.
Jatar previously mentioned the Hold the Line ability. This is a perfect example of the kind of tactic that would be applied at a squad level not an individual.
Comments
A few things already in the design for Citadel of Sorcery...
We mix mobs with opponents that have different roles, just like the player group they will face. Therefore, the opponents will have varied intelligence within the mob you are fighting.
Based on those roles, they have different strategic actions in a battle.
In cases where they are intelligent enough, or have a leader who is intelligent, they will act with squad based tactics. In the case of a leader situation, if he dies, their tactics revert to individual, in the case of intelligent peons, they may still act with squad based intelligence, though at a lesser level.
These squad based tactics will take advantage of terrain as well. The whole idea of the combat AI in Citadel of Sorcery is to make player groups deal with actual varied tactics by their opponents, forcing the players to create their own squad based tactics... or die.
That being said, we will not be throwing extremely intelligent monsters and high end tactics at players early in the game. It would be too hard, and players would leave the game because they were being wiped out by the monsters. Over time you will meet smarter and smarter opponents, ramping the game play up the deeper you get into the world, and forcing you to constantly adapt your game play tactics to the new types of opponents and their better squad based tactics.
A few things already in the design for Citadel of Sorcery...
We mix mobs with opponents that have different roles, just like the player group they will face. Therefore, the opponents will have varied intelligence within the mob you are fighting.
Based on those roles, they have different strategic actions in a battle.
In cases where they are intelligent enough, or have a leader who is intelligent, they will act with squad based tactics. In the case of a leader situation, if he dies, their tactics revert to individual, in the case of intelligent peons, they may still act with squad based intelligence, though at a lesser level.
These squad based tactics will take advantage of terrain as well. The whole idea of the combat AI in Citadel of Sorcery is to make player groups deal with actual varied tactics by their opponents, forcing the players to create their own squad based tactics... or die.
That being said, we will not be throwing extremely intelligent monsters and high end tactics at players early in the game. It would be too hard, and players would leave the game because they were being wiped out by the monsters. Over time you will meet smarter and smarter opponents, ramping the game play up the deeper you get into the world, and forcing you to constantly adapt your game play tactics to the new types of opponents and their better squad based tactics.
How do you plan on stopping CoS from having the same problem AC2 had? Ican't find the article, but it was talking about how the AI was so intelligent that they were impossible to kill (the monsters kept calling for help until the play ended up dying, in the end the devs just turned them into regular, mindless mobs).
Also, wouldn't lag be an effect from having good AI?
Just a few questions.
__________________________________________________
In memory of Laura "Taera" Genender. Passed away on Aug/13/08 - Rest In Peace; you will not be forgotten
Well, I can't speak for the programming in AC2. We are writing our own routines for the A.I., so we won't have the same problems (we'll have our own unique problems, thank you). As for the A.I. being so smart they kept calling for help... well, that's not very smart at all, in game design. How on earth can there be endless opponents to call?
The fact is, in CoS an opponent can call for help (assuming you don't use a silencing spell on them first). But... only the opponents that are close enough to hear their call can come in CoS. So, they can scream their heads off all day and no more will show up unless some happen to wander into hearing range. Monsters do not pop into existance in CoS, so there is no endless supply.
If a general alarm is sounded (some monster rings a massive bell or something to signal an attack on their domain) well, it's up to you as a player to deal with that. Run, hide, or get them all in one room and use an Uber Abilitiy to get as many of them as possible. Whatever, you have strategies, use them.
Lag has no serious issues for the A.I. A half to one second delay between new strategy and movement is not any big deal. The problems with lag has more to do with things like rubber banding (when a player follows another live player and sees them enter a dangerous room, so they follow them inside. In truth the player they were following stopped, but game prediction showed them continuing because the stop information is a second behind.)
A.I. decisions won't seem to have any lag at all (though of course they have the same lag as anything else). But to your perception as a player, when did that opponent decide to run and hide behind that rock? Now, or one second ago? When did they decide to call for help? Now, or one second ago? It's not a problem.
Sounds like you've got a game that's going to be challenging not grinding. I'm looking forward to this so much that this will be the first game I sign up for Beta!
I didn't play AC2, but yeah, that's not smart at all.
As for speed, Game AI generally uses a finite state machine (FSM) to determine individual actions. It is VERY FAST. Genetic algorithms and neural networks are better suited to higher-order learning. Using a neural network to determine the course of action is simple and fast. The training algorithm is what is slower. But you don't have to be constantly training your neural net during combat.
FPS games make good use of neural nets, but the AI will generally go against the same player(s) until they get bored with the game. I'd be interested to see the development of neural net/genetic AI on an MMORPG with an exposure to a LOT of players and techniques over a long period of time.
Speaking of A.I. Will there be large mobs of monsters? I know there are limitations to how many monsters can come at a player. Logistically it's hard for players to keep changing targets, there is the poly count problem, multiple animations running and most difficult, that much A.I. for the game to calculate, especially since you mention there will be NPC group tactics (I read that somewhere, right?)