This is discussion archived from a time before the current discussion method was installed.
Cidolfas: In keeping with the brevity needed to turn this page into an index, most of the text describing games has been moved to the corresponding game pages. The exception are the Paradox games, which don't have game pages associated with them, and rather than add a whole bunch of pages for games I haven't played, I've deleted the reference to it for now.
Seven Seals: "Most games of this type suffer from units with bad AI, as trying to make the massive number of individual elements involved intelligent would overwhelm any CPU."
This should probably read "would overwhelm any team of developers". It's probably not the CPU power that's the problem. It's just very hard to instill anything resembling intelligence into computer algorithms (and yes, sufficiently advanced algorithms probably would need a lot of processing power, but I don't think they've even been developed yet, so it's not really an issue).
Kendra Kirai: Actually, it can and does overwhelm the CPU. While the AI of one unit is inconsequential, if there's three hundred of them, all having to deal with each other, the terrain, any dynamic additions and subtractions to the map (destruction of terrain features, placement of buildings, etc), and deal with the graphics...
Seven Seals: ...it still sounds like a CPU should be able to handle this. The bulk of the graphics can be handled by the GPU, assuming we don't waste CPU cycles on those fancy-schmancy desynchronized individual unit animations, of course. Destruction of terrain is a fairly recent addition to RT Ses, and although pathfinding has never been easy, it's not rocket science either, and not where the most gain in "intelligence" would be expected.
As for having to deal with each other, you obviously wouldn't want 300 individual pieces of game state interdepending on each other (it's not just a performance hit, it's hard to get it to work well too), so you'd give them a "hive mind" to coordinate the lot of them, or at least those grouped into squadrons. Keep in mind that "decisions" don't need to be taken on a frame-by-frame basis either, at least not most of them. Since there will only be so many unit types, there's a lot of potential for vectorizing the calculations too.
But, of course, I'm just speculating, it's not like I've actually written an RTS. I'm not convinced all the RT Ses today are failing to display intelligent units because it would just be too much CPU power, though. Even if that's the case, I'm not convinced it's not the developers not being smart enough to deal with that. (That's three negatives in one sentence; how elegant.)
Much of the fun of present RT Ses comes from the player controlling large numbers of units at breakneck speeds; I'm not sure the developers would *want* to spend a lot of time making units smart, because it would change the whole outlook of the game (making it much more strategical than tactical).
Gus: Encapsulation of the "AI" per-element is fairly straightforward — the design pattern is called Pixie, after the style used in early graphics programs, so the programmer confusion is no more than usual. The trouble comes from the fact that an AI-like algorithm takes more than few cycles, and if you have too many pixies stacked up waiting for their turn at the processor, you "slice" things into a deadlock. So... I am agreeing with KK. <AOL!>
Kendra Kirai: Exactly. You can't give them a 'hive mind', because they aren't all acting as one unit. And since there's only so many cycles the computer has..and if too many units need the cycles at once, the result is a laggy AI, which quickly becomes useless. If you limit the possible actions the unit can take, you limit the cycles needed to decide which action to take...resulting in more units possible without a laggy AI. With only six to 20 (To pull numbers out of my butt) AI routines to deal with, it can easily be incredible (Look at F.E.A.R., or Gears of War), but more than that and it gets very messy, very quikckly.
Seven Seals: Well, I'll have to defer to the people who appear to know more about it than I do. That said, the only way I'll really be convinced is trying to do it myself and failing; seeing is believing, after all. <pulls out a humongous to-do list and scribbles> I'll let you know how it fares...
Gus: Put it near the top. The results would be very cool. Actually, the hive-mind idea is kind of interesting. Right now, all hive-like "thinking" we see are just the result of "flocking" among the little automata. There might be a ecological niche for multiple "hive-minds" taking turns on the processor, giving the little automata instructions to execute limited individual-discretion actions...
Seth: The Generals instalment of C&C did have slightly better AI for all its units and as a result totally decimated P Cs with a less powerful CPU go to the wiki on the zero hour expansion and look at the system reqs.
Kendra Kirai: For that matter, just use a trainer or something to boost the amount of units you can build in something like Warhammer 40,000: Dawn of War. Even with no units on the screen, when the system doesn't really have to worry about animations, if you get to more than about double the normal population cap on both sides (Or much higher than the max NORMAL cap in an 8 way game), even fairly beefy systems will start to wheeze.
Eric DVH: F.E.A.R. and GoW huh? Here's a challenge: Get ten entities and put them on one side of a wall with a door that they can only fit through single-file, but said door is right next to the start of an alternate route which would take far longer than the door if they were patient enough to all file through, tell them to get to the other side of the wall. Now, find a game, ANY game where they all take the door (the stupid ones turning around for the door after it unclogs doesn't count, extra penalty if another group passing through the door makes them turn around again and again). For bonus points, if the game passes the first test, make the alternate path short enough for some (not all) of them splitting up to be quicker, and see if the correct number take the alternate path this time using the exact same AI as the first time. This ain't rocket science.
BT The P: What are these "other" 3D space RTS games, besides Homeworld and its franchise. I'd like to try them. As far as I know, all RTS games that have made the claim so far have been tied to an arbitrary 2D plane.
Kendra Kirai: O.R.B., Offworld Resource Base. It's a far, far, far slower type of RTS than Homeworld. It's by the same folks who did GalCiv.
Madacaek: Whoops, sorry about that, I must've been sleepy or something. :(
MarqFJA: Shouldn't Faction Calculus and A Commander Is You be included in the trope list, since they are pretty much specific to RTS?
- Looks like someone did. ~Slvstr Chung
Slvstr Chung: I made the So You Want To Write A Real-Time Strategy page. It's canted heavily towards the actual-game-mechanics side of things right now, and even then, a lot of it might be pure guff, since my playing experience is limited and I'm a biased son-of-a-bitch anyhow. Help! =)
