Idea: Mini-AI to automate player's commanders

Ideas and suggestions for how to improve the Warzone 2100 base game only. Ideas for mods go in Mapping/Modding instead. Read sticky posts first!
User avatar
aubergine
Professional
Professional
Posts: 3459
Joined: 10 Oct 2010, 00:58
Contact:

Idea: Mini-AI to automate player's commanders

Post by aubergine »

I've been pondering how to make commanders more useful in skirmish games (and possibly MP too, but I don't play MP much so no idea on that really)...

What if a mini-AI was written in to rules.js (the script associated with human players) so that when a human makes commanders the mini-AI controls them for you. If the CRC gets destroyed, the mini-AI would de-activate (you could still manually control the commanders) until the CRC has been rebuilt.

The human could interact with their commanders (well, the mini-AI controlling them) via console messages. Some orders would require the human to select a commander to let the mini-AI know the order is just for that commander, others would apply to all commanders.

Example orders could be:

@get oil = go get more oil resources
@defend oil = build defences around oil resources
@attack <player> = attack a specific player (specified by player number or colour)

Commanders could choose which troops they need to carry out a specific task, etc.
"Dedicated to discovering Warzone artefacts, and sharing them freely for the benefit of the community."
-- https://warzone.atlassian.net/wiki/display/GO
User avatar
Corporal Punishment
Trained
Trained
Posts: 291
Joined: 28 Aug 2009, 12:29

Re: Idea: Mini-AI to automate player's commanders

Post by Corporal Punishment »

Somehow, I doubt a AI capable of automating commanders in the way you describe will be exactly mini... Making them capable to seize or defend resources is easy. But letting a commander choose the troops it will need for a given task, possibly from troops currently available is another matter. It would need to analyze the tactical information available to it's side, like what enemy troops were observed over the last X seconds, where were they going, what were they doing, derived from this it will have to be decided what the enemy's current objective likely is and ultimately what resistance must be expected at a specific oil well. It must also consider the distance to the next repair station and the likeliness of damaged units reaching that station and the time it will take them to return to battle. A planning cycle like this depends on good intelligence in the first place, because if no enemy units were spotted, the AI will necessarily assume few or no resistance must be anticipated and take a weak force along. So there must be a minimum complement it always takes to avoid frustration on the player's part or the script must be deactivated unless a certain percentage of the map area is covered by own sensors. Another problem is that a novice commander might not be capable of leading the number of troops it calculates as necessary to get the job done, how should it react? Refuse the order, inform player about it's incapacity and ask for further orders or go on a suicide mission?
If this AI is supposed to be smarter then a lobotomized cockroach it needs some serious predictive modeling capacity. I'm no expert on AI but I know a fair bit about predictive modeling. I have no doubt such an AI is possible, but I wonder if the gain is worth the cost in terms of developer effort. It might just be easier to modify a existing skirmish AI to the purpose. Somebody would have to teach it the concept of retreat and repair first, though.
Qui desiderat pacem bellum praeparat
Flavius Vegetius Renatus, De re militari
Per
Warzone 2100 Team Member
Warzone 2100 Team Member
Posts: 3780
Joined: 03 Aug 2006, 19:39

Re: Idea: Mini-AI to automate player's commanders

Post by Per »

All examples of games I can think of that have gone this route have shown that this is a singularly bad idea. Case in point - Master of Orion 3.

This is not Sim City.
User avatar
Rman Virgil
Professional
Professional
Posts: 3812
Joined: 25 Sep 2006, 01:06
Location: USA

Re: Idea: Mini-AI to automate player's commanders

Post by Rman Virgil »

.

If this was implemented by way of the player being able to assign a basic maneuver script (from a drop-down menu of several basic, different ones on the unit's Command & Control UI) under specific battle conditions (like when to retreat and repair and then re-engage, call for air support, fire 'n fade, provide suppression fire, etc.), then it could facilitate effectively controlling multiple commanders in coordinated velocity offenses which to me could very well become the grand pay-off of all the non-combat decisions a player makes.

I think this would be complemented by multiple commander combat group velocity maneuver facilitated through an enhanced mini-map that shows your combat groups by Commander #s for you to move the group then there or jump to the actual deployment.

In short, I think the player should still be responsible for the bulk of the situational awarness velocity coordination and that these selected basic a.i. scripts would just augment the players switch-tasking capacity (esp. between multiple fielded commanders) but not replace the players own priority cognitive task analysis in theater, as in center of gravity creation & assessment (or aforementioned situational awareness). Must be careful to not automate too much of the decision-making that is most satisfying. Facilitating the player's switch-tasking (as opposed to crux decision-making), provides the fundamental GP guideline, I believe.

In effect, you become a General and your multiple Commanders become your teammates (aka Field Lieutenants). Play that is possible currently only in Team MP becomes viable in MP FFA !

I'm reminded of the very effective & popular Harvester a.i. scripting mods that were done by the community for "Earth 2150".

There were so many things that contributed to the failure of MOO 3 that its hard to single out any one thing, IMHO. It's overall GPM complexity combined with equally complex and inelegant layers of UI, killed it right outta of the gate for many. Wending your way throught all of that just leeched the fun well before you could even hope to get close to an effective battle engagement pay-off. Anyway that's how I experienced finally playing it after looking foward to it through its MANY years of deving and being a huge fan of the original MOO.
.
.

Impact = C x (R + E + A + T + E)

Contrast
Reach
Exposure
Articulation
Trust
Echo
.
User avatar
dak180
Trained
Trained
Posts: 288
Joined: 01 Nov 2009, 23:58
Location: Keeper of the Mac Builds

Re: Idea: Mini-AI to automate player's commanders

Post by dak180 »

While a good idea in theory, this is one of those where the execution can really make or break it, particularly the GUI execution.

I would file this as something to be revisited when the in-game GUI has been rewritten.
User:dak180
Keeper of the Mac Builds
User avatar
Iluvalar
Regular
Regular
Posts: 1828
Joined: 02 Oct 2010, 18:44

Re: Idea: Mini-AI to automate player's commanders

Post by Iluvalar »

other suggestion that would greatly improve the commanders without proprer "AI" :

*Have the factory act as a commander when he die, holding the units for the new one.
*Have the factory replace the old commander by recycling the highest level unit.
*Have the factory queue behave slighlty differently when attached to a commander, and distribute the units proportionnaly instead of in a queue. For exemple if i as 1 heavy gunner and 2 tmg ht, there will always be 1 heavy gunner for 2 tmg ht in the commander squad. Whatever who dies or not.
*Have the commander resume automaticaly the action of his predecessor when replaced or repaired.
*Have the commander retreat at low units just as it retreat at medium damage. So it dont stay on the front alone or with only one unit left.
*Have the commander drop a "blue" beacon when facing egal or higher cost squad. Or when retreating from a defended spot. Warning the player instantly that he need help.
*Have the command relay act like a unit pool. Sending a commander attached to any blue beacon received automaticaly.

With those changes, a commander would be pretty much "send and forget". If you send one commander in an enemy base, you know that you will have a commander attacking the enemy base until one of you dies.
Heretic 2.3 improver and proud of it.
User avatar
aubergine
Professional
Professional
Posts: 3459
Joined: 10 Oct 2010, 00:58
Contact:

Re: Idea: Mini-AI to automate player's commanders

Post by aubergine »

I love the idea about factories acting different when attached to commander, ensuring the commander has the same configuration of droids at all times.
"Dedicated to discovering Warzone artefacts, and sharing them freely for the benefit of the community."
-- https://warzone.atlassian.net/wiki/display/GO
User avatar
Rman Virgil
Professional
Professional
Posts: 3812
Joined: 25 Sep 2006, 01:06
Location: USA

Re: Idea: Mini-AI to automate player's commanders

Post by Rman Virgil »

.

Essentially, tasking the factory (& CR) with basic command decisions already made that the player wants to iterate thus saving the player cognitive task load (and follow-through switch tasking time) that can be reapplied to making new command decisions, and executing them, in the heat of battle. Would also seem to not require UI additions. Good paradigm for change. :)
.
.

Impact = C x (R + E + A + T + E)

Contrast
Reach
Exposure
Articulation
Trust
Echo
.
Lord Apocalypse
Regular
Regular
Posts: 678
Joined: 29 Jul 2009, 18:01

Re: Idea: Mini-AI to automate player's commanders

Post by Lord Apocalypse »

Per wrote:All examples of games I can think of that have gone this route have shown that this is a singularly bad idea. Case in point - Master of Orion 3.
:lol2: :augh: :roll: :stare: :annoyed:

As much as I love playing and modding Moo3 I must say that that is the worst example to use as the whole game is screwed up (at least w/o the various player patches). The Viceroy AI was a dud.. but it you read over all the tech data and see what was originally intended for the game you will find out that ol roy was probably a late stage addition. There is a lot of leftover code that was tossed or never finished before release. Besides.. Moo3 is a 4x game not an RTS.

@Rman - Did you play Moo3 with any of the player patches or try any of the popular mods?

Now, an AI for the commander isn't too bad of an idea. Its more of an RPG style element than a typical RTS/TBS (Civ) AI. Think of NWN and how each npc party member can be given a few commands and no more worries, if its within their code it happens.

My own take on this is more of a given set of command options. Scout, harass, search n destroy, or fire base (with commander as arty spotter). Though thinking about it now.. an explore option could also be beneficial. At any rate, the AI in most RTS games seems a bit dumb.. The most interesting would be to have a unit/group (commander)/faction AI to handle all operations.

If unit A is acting alone and is either a scout or engineer they should know better and retreat if they get attacked, not sit still and get blasted. You could even give units/groups a morale score (modified for experience). Would give a more real world feel and seems to work well for the Total War games. But, the TW games also track a lot of other unit related data.

Anyway, I am getting pulled away by my 2yo. Will try to finish this a little later.
User avatar
Nameless
Trained
Trained
Posts: 176
Joined: 03 May 2010, 08:25
Location: Space, the Final Frontier

Re: Idea: Mini-AI to automate player's commanders

Post by Nameless »

Is it just me or is this just another way to get out of learning to micro better?? :P

In any case; I agree with Per and the Corporal, but I'd like to add that while adding such a feature would make it seem that the Commanders are being used more (to their original purpose maybe?) but in reality, we're just making another excuse NOT to use them at all?? (Not bring up *that* subject once again haha)
If you're reading this; you're awesome.
User avatar
Rman Virgil
Professional
Professional
Posts: 3812
Joined: 25 Sep 2006, 01:06
Location: USA

Re: Idea: Mini-AI to automate player's commanders

Post by Rman Virgil »

.
@Rman - Did you play Moo3 with any of the player patches or try any of the popular mods?
Never did, LA. Just off the shelf retail. Perhaps I should revisit with those community patches / mods.

The following thread speaks to Pumpkin's original design intent for Commanders, thier 14 year aborted state, as well simple concrete modifications that for lack of MP play testing support there has been no progress or any WZ evidence-based verdict:

viewtopic.php?f=30&t=9119

.
.

Impact = C x (R + E + A + T + E)

Contrast
Reach
Exposure
Articulation
Trust
Echo
.
User avatar
Rman Virgil
Professional
Professional
Posts: 3812
Joined: 25 Sep 2006, 01:06
Location: USA

Re: Idea: Mini-AI to automate player's commanders

Post by Rman Virgil »

....simple concrete modifications that for lack of MP play testing support there has been no progress or any WZ evidence-based verdict....
That said, there was a perfect SP test bed to clearly isolate all the basic flaws of Commanders. Had zero to do with microing efficacy. Which is to say, not all microing is created with equal fun value even with an absolute dedication to proficiency.

The test bed I speak of was Black Project's "Hard Core Campaign Mode" for v. 2.3.x (I forget at which version it became incompatible).

The original campaign even at highest difficulty was a cake walk compared to this one.

While much was changed, Commanders remained stock. But here's the interesting thing - you pretty much had to use Commanders to have any chance of success in many of the missions. And in using them all their basic flaws became obvious in combat. Being the supreme micro gamer in the WZ universe would only make you concious of these flaws and frustrate you at the same time. These same flaws would also manifest in MP and prove disastrous.

This was a valuable evidence-based experience in understanding Commanders - warts and potential. I would recommend this experience over pure speculation or untested hypotheticals.

Pumpkin is on record for what they wanted Commanders to be and why they aborted. The reasons for the abortion of '97-'98 donot hold in 2012.

.
.

Impact = C x (R + E + A + T + E)

Contrast
Reach
Exposure
Articulation
Trust
Echo
.
iap
Trained
Trained
Posts: 244
Joined: 26 Sep 2009, 16:08

Re: Idea: Mini-AI to automate player's commanders

Post by iap »

I don't like the idea that the computer will play instead of me.
However, some additions to the commander would make them really effective like:
* The ability to plan a strategy (Somethink similar to queue actions, but in generalized manner, like queue goals)
* The ability to target an area and not just a single target
* The option to wait for a specific conditions
* The ability to give autonomy to it's attached utints without them being unattached (The player can override the commander when idle, or even they will shoot other things then the target when they can't shoot the target)
* The ability to set the commander to retreat on number of nearby units drops below a value.

Annnnd (I have to mention) Just fixing the basic commander's bugs wiull make a huge difference :)
User avatar
Rman Virgil
Professional
Professional
Posts: 3812
Joined: 25 Sep 2006, 01:06
Location: USA

Re: Idea: Mini-AI to automate player's commanders

Post by Rman Virgil »

iap wrote:I don't like the idea that the computer will play instead of me.
Very important, and fundamental, guideline to any changes.
However, some additions to the commander would make them really effective like:

* The ability to plan a strategy (Somethink similar to queue actions, but in generalized manner, like queue goals)
Player situational awareness enhanced through expanded cognitve load and switch-tasking facility, in the heat of battle.
* The ability to target an area and not just a single target
Smarter excercise of power command control as opposed to stupid, suicidal and GP frustrating, anti-fun.
* The option to wait for a specific conditions
Smarter excercise of power command control as opposed to stupid, suicidal and GP frustrating, anti-fun.
* The ability to give autonomy to it's attached utints without them being unattached (The player can override the commander when idle, or even they will shoot other things than the target when they can't shoot the target)
Smarter excercise of power command control as opposed to stupid, suicidal and GP frustrating, anti-fun.

Player situational awareness enhanced through expanded cognitve load and switch-tasking facility, in the heat of battle.

* The ability to set the commander to retreat on number of nearby units drops below a value.
Smarter excercise of power command control as opposed to stupid, suicidal and GP frustrating, anti-fun.

Player situational awareness enhanced through expanded cognitve load and switch-tasking facility, in the heat of battle.
Annnnd (I have to mention) Just fixing the basic commander's bugs will make a huge difference :)
+ 1

"Huge difference" - indeed. Enough to qualify, put plainly, as a thoroughly fresh and better WZ while remaining true to the game creator's original design intent. Put the game on the RTS map, like nothing else ever has or, frankly, ever will without this as part of the total package (unit design, extensive tech tree, etc.).

But, achieving such goals as these can inspire trepidation. Daring an emergence from obscurity. This excercise of combat power with common sense, finer-tuned, command control. The audacity of gaining greater attention outside the little isolated pond... :hmm:
.
.

Impact = C x (R + E + A + T + E)

Contrast
Reach
Exposure
Articulation
Trust
Echo
.
User avatar
Rman Virgil
Professional
Professional
Posts: 3812
Joined: 25 Sep 2006, 01:06
Location: USA

Re: Idea: Mini-AI to automate player's commanders

Post by Rman Virgil »

.

Just to make explicit what I have been doing in my posts here and elsewhere, for some years now.

I'm using a conceptial vocabulary derived from true 21st century military doctrine (the War Tech Machine NOT divorced from multi-vector, velocity, maneuver and augmented situational awareness) and the latest cognitive science discoveries of the last decade plus. WZ is not Sim City but nor should it be WW 1, or Steam Punk, Victorian, combat. I'm not talking anything remotely like a war gaming simulation, just Hard Sci Fi on WZ's own narrative substrate that expands fun within traditional RTS GPMs.

The cognitive science discoveries themselves are currently being applied to fundamental, cutting-edge, game design as in GPMs and in their intimate linkage to UIs and CTA / Switch-Tasking (multi-tasking is a myth and thus its correlation to micro-management has to be revised along with its inherent bias & unquestioned assumptions - a different mindset, like a paradigm shift, can be very difficult and protracted stumbling block). There is actual, serious, overlap in RL combat preperations with this type gameing.

Pumpkin, WZ Creators, were way ahead of their time with their original Commander functionality design vision.

.
Last edited by Rman Virgil on 15 Dec 2012, 08:41, edited 1 time in total.
.

Impact = C x (R + E + A + T + E)

Contrast
Reach
Exposure
Articulation
Trust
Echo
.
Post Reply