The RTS game I've been working on has the concept of "Lieutenants". The Lieutenant role (there's only one unit type in the game, and a given unit will have the skills for one or more roles) is one of subordinate control: other units can be assigned to it, and it will delegate them to various high-level tasks (repair, navigation, etc.) that are under its control. Each high-level task has a panel of sliders that the player can mess with to change the lieutenant's priorities.

In order to make this system workable, I've had to formulate the gameplay such that there is generally a very obvious "right thing to do" for a given unit at a given time. Of course, this needs to be balanced against the need for actual strategy. Additionally, lieutenants--especially inexperienced ones, or those handling too many tasks at once--will be somewhat slow at times to delegate jobs to the units under their control. Additionally, players can countermand the lieutenant at any point WRT a single assignment or multiple assignments, and the lieutenant will honor that countermanding without the rest of its assignments getting screwed up.

While I haven't done much playtesting yet, what I'm hoping is that players will leave lieutenants in charge of roles which are not of immediate importance, while handling those that need to be done especially well at a given time themselves. The goal is to reduce the tedium of micromanagement, while giving an advantage to those who can from time to time.

One of the majo goals of my RTS is to add AI to the human player. Some of it sounds a little similar to Sneftel's ideas, other stuff I'm not aware of being done before.

Racing games use this already - as well as things like auto-gears and auto-braking, I know of one game where the steering is silently adjusted by the AI to make it seem easier. For isntance you get into a spin and the AI automatically pulls you out - making you think you've done a stylish 360 with your 1337 skillz.

I agree with most of the posts that your idea isn't quite suited for a typical action game. You could use it as a real time "conscience" to help the player along, i.e. think for him, but I do not believe this is what you are after.

Along with your original though line from Riddick; it might be best to use it to do things that the player himself cannot do. The whole two players at once fighting thing would be a good example of this. In that case I would probly use a learning AI to figure out the players style, just so they don't get pissed when it does something realy dumb that they themselves would never have done.

Quote:Original post by Sneftel
The RTS game I've been working on has the concept of "Lieutenants". The Lieutenant role (there's only one unit type in the game, and a given unit will have the skills for one or more roles) is one of subordinate control: other units can be assigned to it, and it will delegate them to various high-level tasks (repair, navigation, etc.) that are under its control. Each high-level task has a panel of sliders that the player can mess with to change the lieutenant's priorities.

I'm working on a similar system for my space sim/RTS. Since the plan is to make control of the player's ship fairly complex, I decided to delegate most of that control to "officers", leaving the player to focus on whatever task is currently most pressing or interesting. In particular, the three major areas of importance during combat are maneuvering, firing control, and damage control. Since the game deals with large, complex ships all of those tasks end up being pretty involved.

The player can delegate maneuvering, for instance, to an officer (by issuing a command along the lines of "stay within 20km of that ship"). The game takes over flying the ship, while the player can focus on using the weapons and dealing with damage. Alternately, the player can automate weapons and damage control and just focus on flying. Masochistic players can just ignore the automation completely do everything themselves, but I'm trying to focus the gameplay heavily on interacting with and delegating to the AI "officers".

Of course, the downside to this is that the playability of the game heavily depends on the quality of the AI. If I screw up in implementing the AI that drives the way officers work the whole game ends up falling apart.

Still, I'd agree with everyone who says that this kind of stuff doesn't belong in an action game. The whole idea of an action game is that the player is forced to quickly react to events as they happen. You take away a big part of the fun if you remove the fast paced nature of the game, especially since most action games feature a very limited set of actions (move around, aim, shoot, switch weapons, etc.). The strategy in these kind of games comes from the complete freedom of movement combined with the player's skill at reacting to input as quickly as possible.

Well, AI-assistance can be a great thing. For example - a first-person RPG game. Your hero can shoot, but it is dependent on his skill how can he shoot. Then, for player, you have to options -
1) automatic aim - computer automatically sets center of fire on the target
2) manual aim - you manually set center of fire, which is harder, but that way some 3D-action game lovers can be added to the audience of the game ;)

EDIT: Skill controls how WELL our hero can shoot.

Everyone keeps talking about using AI assists in action games. What about in strategy? I believe that strategy can have things that are extremely complex, and that this can be a good thing. Ifyou increase the complexity yto the point where you HAVE to use AI to deal with it, you've got something completely new. I indend having a 2d space strategy game, but I intend that other than the number of demensions, it will be a hard sci-fi game. I want to make it so that the default is that you fight exactly as well as the AI ship, and as the commander, you deactivate certain ai and take control of that system. You give the AI parameters ingame. You give the weapons groups and give the groups targets and they fire on abjects 1, 2, 3, and whatever they can see, in that order.

Effectively what I'm looking at is a high degrwee of automation, not an actual AI. WHat I want is for a halfway decent player to be able to whoop the AI, simply because he knows how to use this automation effectively. Of course, because of the nature odf the game, the AI will only be used against the player where the player is heavily outmatched, in some way or another, thus resulting in about equal battles.

Is it wise to give the player control over the AI-script? How many players will use it? ANd how much help can I expect if I make that part more accessible to the general gaming audience? How much will they take advantage of it if I do this?

I think a decent AI could assist you well in a strategy game. It gets edgy when the AI goofs up and causes you to lose the match.

Nowadays, strategy games already offer quite a bit of assistance, Civilization IV being a good example with their automated cities. But also Pathfinding is, in my eyes, AI assistance. and some games, such as Total annihilation, have really user-friendly path finders, with units that scoot out of the way to let less agile units through the shorter way, etc pp.