Announcement: Pylon Wars - Page 2

Announcement: Pylon Wars

WolfeGames and projects headed by Argh.

Moderators: Moderators, Content Developer

User avatar
Argh
Posts: 10920
Joined: 21 Feb 2005, 03:38

Re: Announcement: Pylon Wars

Post by Argh » 22 Nov 2009, 23:21

So, for scouting purposes, what algorithm design might be fast and flexible? An area search per scouting Unit to determine density, filter by target type, arrive at a weight for a given sector?

I'm not sure how I'd like to determine weighting, etc. for that algorithm, basically. I've never designed that kind of algorithm, I just kinda assume that godlike AI for lolz games is more fun for players- the "smarter" the AI behaves, the more likely it'll do something really stupid, like sit around doing nothing because it can't find a valid target.

More importantly, I don't want the Pylon People sitting around during initial play, that would be a Bad Thing, imo.
0 x

Google_Frog
Moderator
Posts: 2442
Joined: 12 Oct 2007, 09:24

Re: Announcement: Pylon Wars

Post by Google_Frog » 23 Nov 2009, 02:14

It doesn't have to be a great scouting algorithm, you're not trying to make an AI that plays a playable side realistically. Just make a few extra units that are scouts. Give them a move order to a random position on the map, if they reach their destination then give them another random order. If they see an enemy then all (or some for ffa?) of the enemies attack that unit.

The inefficient unit useage does not matter as none of their units really cost anything, they are spawned.
0 x

User avatar
AF
AI Developer
Posts: 20669
Joined: 14 Sep 2004, 11:32

Re: Announcement: Pylon Wars

Post by AF » 23 Nov 2009, 02:36

Sometimes attack units themselves are useful for scouting. Early game some players use peewees and flashes as scouts even though that's not their role. I would regularly send peewees in XTA games mid and late game as probes to see how far they got before they got blown up if at all.

Different scouting algorithms have different strengths and weaknesses. For example an influence map where the unit moved to the nearest square with the highest value, and every frame squares in LOS were set to 0, squares outside LOS were incremented. That works well for aircraft.
0 x

User avatar
Argh
Posts: 10920
Joined: 21 Feb 2005, 03:38

Re: Announcement: Pylon Wars

Post by Argh » 23 Nov 2009, 04:03

I will take a look at this, and certainly test some concepts out. I am still fairly dubious that it's worthwhile in the context of the game design.

The current algorithm's main advantage is that it can't screw up something as basic and important as targetting. It will attack something important, every time.

If I redirect all the active Units other than scouts every time the threat changes, it seems to me that they'd do a lot of semi-random flailing about, and it wouldn't really serve the game design. I mean, this isn't meant to be a tactical extravaganza, other than the very complex tactics of defense, especially in P.U.R.E.- it's about survival, then assault, then trying to survive the counter-assault and hopefully being able to take down another Pylon, and so forth.

It wouldn't be good to screw up such a simple concept for a feature that isn't totally necessary, basically.

However, I'll still work on an algorithm. It may be something that's more useful for another type of mini-game, like a "hunt the hero" type game, where the hero needs to be able to hide and recuperate, and that's not a bad idea for a future thing, once the other thing is done.
0 x

User avatar
AF
AI Developer
Posts: 20669
Joined: 14 Sep 2004, 11:32

Re: Announcement: Pylon Wars

Post by AF » 23 Nov 2009, 06:03

In practice it doesnt result in units changing their minds and goign all over the place every few seconds.

All the AIs in spring relied on scouting for years.
0 x

Locked

Return to “Argh's Projects”