Request for Advanced Formation AI

[PicassoCT]

The more epic Spring gets - if 100 Units are moved , the more the real strategy disappears, and the Units send over the Map (even when moving at speed of the slowest) just form a huge bulk - easy to attack, self blockading and not controllable. With simple Formation you archieve a sort of controll, but i still saw AntiAir & Con Units in Front while the Warriors stood far behind.

Therefore i like to ask if one of the Programmers is up to the Task of Producing a Advanced Formation AI.

The Idea: The Units form a Convoy at the Speed of the Slowest. The AI also lets the whole convoy select the way of the most movement limited Unit. (Hovercraft for example) If Units can`t go that Way, they are left behind.
Units that can take in some Shots form the Outer Hull (Tankers)
Following are Direct firing LongRangeUnits. (Annihilators, Snipers
The FrontCore is made of indirect firing Units (Diplomats, Luger, Mortys you know them.)
In the Safe Area the ConUnits (if they are with the Convoy) repairing Units by Request.
Last not Least the Speedcounters - send out to Short Sideway Patrolls (not further then the Directlongrangers can shoot. (Zippers, Weasels) - they Autoscout for Attacks from the Side. (Now something to Provoke some AI Designers And that is all we need...

[AF]

I'm sorry but you've staked a request at the worst possible time.

The active contributing AI community in spring atm is comprised of me hughperkins lindir and submarine. And we're all busy on skirmish AIs and university etc Infact hughperkins was on holiday in china when he released the first CSAI, and I ahve 3 courseworks to finish before I can start NTai work again properly.

[PicassoCT]

Sorry AF, didn`t knew that.

I will try to bring that Thing Up again, later if somebody has free time, okay?
Better to have good skirmish AIs to Battle against, instead of Good Formations to Battle against yourself !

But it could be working, or - no logical mistakes in it ?


Pic

[Kloot]

Creating a formation from a group of sufficient size and variety, sure, that could work. Maintaining it while moving might too. Keeping one intact in battle very likely wouldn't, though.

e: just to clarify, I say "might" in the sense that the issues surrounding general formation management are practical in nature and so could be overcome given enough CPU resources, not theoretical.

[PicassoCT]

I know that it would fall apart, simply because every Unit trys to get a shot.
It is one of the Sad things about Real Time Strategy - you loose Controll over the Battle the more Units take Part and you gain it back when half of the Combat is over.
Ideal wouldbe if you could tell some AI your Intention for Grouped Units before the Battle beginns (like keep the Mavs together, no matter what and move into that direction) blasting your way throug searching for the com - and get them into Battle with that already "in mind"
But that is impossible. Another Sollution would be Auto- MatrixTime in Great Combats, so that everybody who takes Part can give detailed Orders. But what with those not taking Part - waiting till the Clock ticks again? No Way.

[AF]

This is a prblem because this sort of groupAI would need a pathfinder and terrain analysis to be really effective, and thats something that would require a lot of cpu usage for this sort of thing.

Remember that this would require a bigger pathfinding system resource usage than the internal spring pathfinder, and the two would be running at the same time.

[PicassoCT]

Is there no Dirty Work Around ? For example calculating it as a single huge Unit in Pathfinding ? Or at least a Group of Great Units and that way deliver the Work to the standard pathfinding ?

[Neddie]

I have a friend who may be interested in getting in on this AI orgy - I'll see if he wants this role. No promises.

[PicassoCT]

Convoy for the Go - .. Hurray

[Triaxx2]

I don't know about a massive formation, but a single column wouldn't be difficult, or even two.

The path finding is simple enough, because I've proposed it numerous times. A simple hexagon, which detects obstructions in the path. Instead of that convoy though, you'd be better off with a five column set, with the outer two comprised of the 'tankers', and moving inwards until it reached the 'con units' in the chewy center.

[AF]

Dirty quick hack columns are easy but they'll fall apart immediatly when confronted with anything but flat terrain.

As for pathfinding, yes you could represent them as a whole unit, but then you have to keep the units in that formation anyway, and most importantly you'd requrie a second pathfinder running which is the root cause of lack of progress in this area as it's computationally expensive.

The KAI pathfinder ahd these issues and Tournesol would reduce its detail untill you had rather large pathing squares which werent particularly useful for finding movement paths like that but rather for using influence maps for generalizations etc such as which path will take me around andaway from threats?

Thus I'm not sure this would be a fair groupAI as it's only really feasable in skirmish AI's, since low end users couldnt run it many times and expect no lag giving high end PC users and advantage.

Then there's the adage of mod compatability, what position would you place a unit in? Its fine all and all saying this should go there but we're humans and we're thinking in high abstract terms.

So I think instead it would be best placing effort and emphasis elsewhere but related problems so that at some point in the future we can come back and tackle this using stuff we've already coded.

For example, a routine to use basic spring line formations where appropriate in skirmish AI's, and basic line of fire calculations.

[silv]

in my opinion such an ai would take away a lot of fun from the game if i used it myself. And additionally i would not want to play anyone using such an ai, as it is almost cheating - having a perfect formation all the time.
you are supposed to control your units, not some cpu-ressource-dependent script.

if you want to see your pc battle start an AI vs. AI game

[PicassoCT]

You can controll your Units.. but if you watch GAmes with Great Numbers, the are just moving bulks. I see this as essential to controll big Battles. Most People then no longer use Wapoints & Ques - instead they spent most Minutes Micromanaging there Units.

[silv]

i still think it's part of the game / of the skill and shouldn't be managed by an AI that puts my units into a perfect formation.

[AF]

Picasso, the system resources used rises greatly as the number of units rises. What you are suggesting would put this groupAI out of reach of the vast majority of spring players as soon as you pass the 30 unit mark. I dont think you realize just how intensive pathfinding can be.

Also most maps dont have spaces big enough for such a formation to exist.

[PicassoCT]

Well, i will store that idea somewhere in the Backyard - and in ten years, when everybody has forgotten it, i pull it out and scare the masses :)

[krogothe]

It is feasible cpu-wise, its not really all that computationally expensive or that hard to work in most maps (a lot of games have this implemented by default), though the idea itself is pretty useless.
A formation as such is expensive, slow, vulnerable to AOE, predictable and ultimately useless due to the wide difference between unit uses, ranges, speeds and tactics.

[Peet]

Who needs groupAIs when theres methamphetamines. Speed + spring = uber micro of dewm!

[Triaxx2]

Actually, using the main pathfinder, and having the units move from one waypoint to the next in straight lines, as if they were a single unit would be more than sufficient.

As group AI certainly not, but as a part of a Skirmish AI, yes.

[zimbob]

It sounds like a super idea am I right in thinking it would be like the pre-set formations you can choose in Warcraft?