Palidine, you have made some very good points, but clarifications are necessary.
Quote:Original post by Palidine
Quote:Original post by apollodude217
However, does anyone know of any RTS's that actually use this scheme? As far as I can tell, all the RTS's I've tried don't even have a good bottom-up AI design--that is, there isn't even variation in AI from unit to another based on that unit's tactical function. That is, a tank won't be any more willing to advance against the enemy automatically than an arty unit. The units blindly do _literally_ what you tell them, requiring the player to micromanage, effectively doing away with MANY desireable strategies that would otherwise be employed. Are there any major commercial games using the top-down AI designed discussed in this article? |
Couple things. There's no desire to add the kind of unit AI that you're talking about (tanks more willing to execute orders, etc). We try to add that all the time and it is universally despised by designers and testers alike. The point of an RTS is that your units do what you tell them no matter how much it puts them at risk. You want 2 snipers to defend a fort against an army of 100 enemies? No problem, they just do it. The "smarter" unit AI generally just gives the impression of "why the hell won't my units do what I tell them to do??!?!". |
Here, you are talking about one of at least 4 prominent understandings of what AI is: Human-like behavior. Such behavior is, in fact, implemented in some "realistic" games, such as the Close Combat series, and has its pros and cons--On the plus side, they lack the cheesiness of obvious realism and add the dimension of morale to the game's strategy, but on the other hand this behavior is annoying to many players. But I digress.
I am advocating greater rational behavior, which is what micromanagement is all about anyway. Such rational behavior should be implemented only for the obvious things; e.g., in the case of pathfinding for non-combat (e.g. resource-gathering) vehicles, the risk of dying in a given area can be added to the path costs, thus causing such units to take a safe route to the resources, even if it implies taking a longer route. If the cost calculation is calibrated well, then such units will _generally_ take the route that the player would want them to take anyway.
Quote:
The problem with smarter units (on the player's team anyway) is that by being smarter you are restricting the player's options. Sure it's "smarter" and "more realistic" but it removes a lot of bizarre suicide strategies that the player may wish to employ. The RTS genre is all about micro-management and "dumb" units.
|
Not necessarily. First of all, there sometimes exists one and only one "best" thing for all circumstances, and moreover, there are many, many things that will _almost_ always be the best, or nearly the best, in most games. For example, in a typical fighting game, you never ever want to just stand there; you always want to either parry, block, or attack. The first strategy most players learn when playing such games is always do _something_; do the best move for the situation if you can think that fast, but always do _something_. This leads to button mashing. Good "AI" (or even good UI) would cause the fighter to block automatically when no input is given. Of course, there exist games where this is not true, but in many fighting games, this seems to be the case; I think you get my point. If a default move (such as blocking) were given automatically, then the player's mind would be freed up to think through what buttons to push when, as a little less brain power is being put into pushing buttons quickly.
Likewise, in Total Annihilation, you almost always want your power plants and factories spread out by 1 or 2 spaces. You almost always want to build metal mines on all metal deposits within the terrain you have captured (and captured terrain could be reasonably deduced by the computer, or manually indicated by the player for added precision). In the classic Command and Conquer games, you almost never want your ore trucks / harvesters to go through the enemy base on its way to and from the enemy base; you instead want it to go around, taking a path known (or at least projected) to be safe because the player has scouted that area and found no enemies.
None of the above applies to top-down AI, but there do exist examples of top-down AI that do what the player nearly _always_ wants to do. In particular, I'm thinking of tanks leading the way in the attack, speed units rushing forward when the tanks start getting shot in order to attack the enemy's resource structures, arties and other support units to hit the enemies once they've been found (but without actually going to the front lines), medics / repair units to heal damaged units within a certain group or area but without actually going to the front line, and all attack units actually retreating and _going to_ the repair unit once damaged. A grouping sort of AI is a very easy way of implementing all the above.
Quote:
Now such a system could be readily employed to make strategy games possible on consoles which, because of their more limited input devices need streamlining in how the combat system works. However, you need to work out a game design that works with smarter units. You can't just plug them into the existing RTS genre and have it work; the entire core of the genre is micro-management. You'd have to alter the design of the game such that the compulsion for micro-management no longer exists. |
"RTS" is supposed to stand for "Real-time strategy", but most rts's are actually real-time tactics games. The distinction between strategy and tactics is discussed extensively on other threads on many message boards, so no need to discuss that here. Tactics are fine. I love tactics. But I also love strategy, and I'm annoyed when RTT's are built on a grand strategy scale where rational behavior in the form of things like formations, flank attacks, pathfinding that avoids unnecessary enemy confrontation, etc. are simply impossible because the player can't point-and-click quickly enough. RTS's are too often a race to see who can point-and-click the fastest. We might as well invent a game called GUI Speed Challenge where the player must copy and past files, rename files, navigate weppages, etc. in a race against the clock! How fun would that be? Unfortuneately, playing an RTS / RTT is much like getting those 3rd quarter analysis spreadsheets done and turned in to the boss. Regretably, it is too often too difficult to employ very many strategies or tactics in large-scale RTS's, and it's all due to "dumb" AI and limited UI.
Also, offering features of macromanagement need not in any way remove the ability to micromanage. The two are not opposed to each other, except that keybindings might become suboptimal as the definition of point-and-click-on-that-there-soldier can mean only one of two things: select one soldier, or select his group (for example).
|
Posted - 3/13/2007 8:31:55 AM
