Commander Command Improvements -- What ? Details ?

Discuss the future of Warzone 2100 with us.
User avatar
whippersnapper
Regular
Regular
Posts: 1183
Joined: 21 Feb 2007, 15:46

Re: Commander Command Improvements -- What ? Details ?

Post by whippersnapper » 26 Jun 2009, 23:48

andron wrote:
whippersnapper wrote: "What I have been in the habit of doing is assigning 2-4 Borg Repair Mechanics to each Commander led Combat group so that units are repaired "in the field, continuously" and never retreat to Repair Facilities, thus side-stepping the short-comings raised."

the problem with every solutions ist that:

The Computer normally focus its fire on one unit, and when he focus on the commander it will either quickly die because the repair bots are not quick enough or when the repair bots are extreme quick he will catch all the fire, or he will be able to retreat quick enough, in that case it not only need repair skill but also unearthly healt points on the commander turrent then it will work that you will mostly never loose a commander.
I see what you're saying. What chassis do you use for your commanders ? What game mode do you usually play - Cam, Ski or MP ?

regards whipper :cool:
Last edited by whippersnapper on 27 Jun 2009, 00:12, edited 1 time in total.

User avatar
whippersnapper
Regular
Regular
Posts: 1183
Joined: 21 Feb 2007, 15:46

Re: Commander Command Improvements -- What ? Details ?

Post by whippersnapper » 27 Jun 2009, 00:11

whippersnapper wrote:................................

We'll come back to this. It merits closer examination, all the way around. Brings to mind Peter Molineux's game A.I. work introduced in the first "Black & White" game back in 2001 and further evolved in "B & W 2". I'm thinking that on-the-fly, in-game, "Reinforcement Learning" would be better than the "Off-line Training" of the PSO algorithm {Swarm A.I.}.... but I need to cogitate on it some more in either case.

Regards, whipper :cool:

NEXT: I will share the results of a survey I conducted in the WZ community back in 2004 on the subject of applying Reinforcement Learning A.I. Techniques to Commander A.I. Far from being strictly scientific the survey nonetheless yielded some interesting results that informed my approach to the GCI in MP AND individual Commander A.I. in MP as well. Again these results reflected, starkly, the differences that must be kept in mind between the 3 Modes of game play - Campaign, Skirmish & MP - and their distinct audience tastes-appeals.
.
.
Informal survey results on applying "Reinforcement Learning" A.I. to WZ Commanders:

* Campaign A.I...... absolutely !

* Skirmish A.I........ yup !

* MP Commanders..... no way !!!

Six months ago I read some commentary at SlashDot on the "State of the Art in Game A.I." and this one stuck in particular:
Most humans aren't that smart (Score:5, Insightful) by TheLink (130905) on Tuesday November 04 2008, @03:59AM (#25623377) Journal


But does increasing the I of the AI actually make games fun?

The Problem that AI is supposed to solve in most Games is not "how to beat the human".

The Problem is "how to make it fun for the human"

Creating an AI that can consistently beat humans is not hard. Making it fun for most humans might not be so easy.

Fact is humans aren't that good at most games (amongst other things). You don't have to be very intelligent to be good at most games. How many of you can beat a computer at chess at high difficulty? How many people actually _lose_ in tic-tac-toe - I've seen more than a few :).

It's often not hard to make a computer extremely good at a game, at least good enough to beat most people. But does that make it fun?

In most FPS games, stupid humans want to be able to mow down _thousands_ of stupider computer controlled enemies - "against the odds". That's what makes it fun for them.

That's just not possible if the enemies start having a lot more brains. Then most players might have difficulty getting past the first 3 enemies :).

It's not that difficult to make an enemy FPS "bot" have superb tactics, coordination, timing etc. Especially if the map is pre-known (which is usually the case). You can code the tactics and heuristics in. If you hear the player in position X, group A enemies head to position Y and group B head to position Z, and bye bye player.

Imagine if enemies that are low in health kept running away and hiding, and then snipe at you from far away when they see that you are busy doing something else. While that might be more realistic, it might not be so fun eh? Who really wants realism in games?

At that rate the player can never pretend to be the hero he wants to be. He'll just be dead. And your game won't sell.

Same goes for RTS games, believe me, you don't have to make a computer cheat to beat humans - a computer can micromanage better than most humans.

Just ensure that basic stuff like navigation is better. Stuff doesn't have to be that smart, but at least they shouldn't be totally stupid - they should be able to walk around stuff without getting stuck - even a "dumb" animal can navigate open spaces better than many computer controlled stuff in games.
HERE is the source page if you wanna read the whole of it...

NEXT: I'll bring my thoughts together on how all the foregoing relates specifically to Swarming A.I., Reinforcement Learning A.I., Fun Decision-making, WZ MP, Asymmetric GPMs, Commander A.I. and the GCI...........

Regards, whipper :ninja:
.
.
.
"I need no warrant for being, and no word of sanction upon my being. I am the warrant and the sanction." Anthem

"Art is the selective recreation of reality according to the artist's metaphysical value judgments." A. Rand
.

User avatar
whippersnapper
Regular
Regular
Posts: 1183
Joined: 21 Feb 2007, 15:46

Re: Commander Command Improvements -- What ? Details ?

Post by whippersnapper » 27 Jun 2009, 23:12

whippersnapper wrote:
NEXT: I'll bring my thoughts together on how all the foregoing relates specifically to Swarming A.I., Reinforcement Learning A.I., Fun Decision-making, WZ MP, Asymmetric GPMs, Commander A.I. and the GCI........
.
Oh my, that's quite a handful and, every bit, worth attending, IMHO. But though my natural enthusiasm in word and deed is undiminished it is still somewhat effected by inequity aversion. So before that latter state encroaches in its fullest measure (akin to taking a swim in a cement over-coat) I'll wrap this up with clock-work concision to honor my word and save most of the discursive expression on those interrelated topics for my developer journal-blog in another cyberspace.

Variety of Field Command decision-making is at the heart of this. The GCI attempts to buy-time, in real-time, by making for an intelligent but limited autonomy in commander led combat groups within a battle theater. This last allows the MP player to handle supportive non-combat GPMs adequately while effectively managing (through an engrossing fun, rather than frustrating, switch-tasking) multiple, coordinated, vector maneuver (mixed arms by ground, air and water) which in itself is essential to that variety of RL military maneuver (currently totally lacking) including the viability of asymmetric strats and tacs of 21st century, 4G, velocity doctrine.... (aka, beating the odds or clever, nimble, maneuver out-witting massed, sheer force strength).

Let me leave it there, whipper :cool:
.
.
"I need no warrant for being, and no word of sanction upon my being. I am the warrant and the sanction." Anthem

"Art is the selective recreation of reality according to the artist's metaphysical value judgments." A. Rand
.

User avatar
psychopompos
Trained
Trained
Posts: 470
Joined: 08 Nov 2007, 09:18
Location: UK

Re: Commander Command Improvements -- What ? Details ?

Post by psychopompos » 28 Jun 2009, 01:16

andron wrote:the problem with every solutions ist that:
The Computer normally focus its fire on one unit, and when he focus on the commander it will either quickly die because the repair bots are not quick enough or when the repair bots are extreme quick he will catch all the fire, or he will be able to retreat quick enough, in that case it not only need repair skill but also unearthly healt points on the commander turrent then it will work that you will mostly never loose a commander.
that is why in the past i have suggested that the commander should ensure it moves a unit into place to shield itself from being targeted.

i would like to see commanders being rather self contained.
ordering their own replacement units, organising their own units/formations, such as moving artillery/systems turrets away from flanking opponents & covering with combat units themselves.
moving damaged units away from fire & to assigned repair turrets.

perhaps pulling off some basic manoeuvring techniques, having assigned units move continuously when under fire for instance.

hell, i wouldnt mind seeing them throw up a repair station & some bunkers if a truck is assigned.

this way i can leave them to cover area X, Y or Z while i go attack a base

User avatar
whippersnapper
Regular
Regular
Posts: 1183
Joined: 21 Feb 2007, 15:46

Re: Commander Command Improvements -- What ? Details ?

Post by whippersnapper » 29 Jun 2009, 18:23

.

I'm gonna close-out my input by sharing some resources I mentioned elsewhere but merit reiterating here..

Lots of experimentation with EW, WZCK, 2 WZ A.I. Editors & and play-testing within WZ was supplemented by work done with the following software resulting in data / experience which in turn informed my PoV here on Commander Command Improvements, my work on "War School" and "RTS Design" in the round (both these have additional resource lists).

A.I. Experimentation Software Tools and environments I have found helpful over the years...

Regards, whipper :cool:
.
.
"I need no warrant for being, and no word of sanction upon my being. I am the warrant and the sanction." Anthem

"Art is the selective recreation of reality according to the artist's metaphysical value judgments." A. Rand
.

User avatar
whippersnapper
Regular
Regular
Posts: 1183
Joined: 21 Feb 2007, 15:46

Re: Commander Command Improvements -- What ? Details ?

Post by whippersnapper » 04 Jul 2009, 00:13

.
Part Two of my concise personal summing-up on this topic... (My last post being "Part One: A.I. Experimentation Software Tools and environments I have found helpful over the years...")

- All my Commander proposals herein, including the "GCI" (2004-09) and everything associated with it, are based on, and intended for, the following:
  • 1.) These are for WZ 2120 as I've conceived it.

    2.) They are the cornerstone of T-4 as I see it.

    3.) This may come off like I'm anointing myself but really it is only based on the same things in-force around the neighborhood: the GPL and the wherewithal to act on what I see as a fun direction for my WZ "tinkering" passion.

    4.) So, in essence, this represents a clearly focused continuity from '99 to the present and what I foresee precisely as my continued active interest in WZ (a hobby pastime going into its second decade, which is hard to believe).

    5.) WZ 2120 will begin at T-4 and will presume WZ 2100's continued development in the areas of Re-Balance (by Zarel), ECM-Stealth implementation (by Per), full implementation of Beta-Widget-LUA (by, Evil Guru, Gerard, Devurandum, ?), full implementation of "Pilot-Mode" (?) and Elio's proposed new UIs off of Betawidget.
- T-4 (Tech & UIs) for WZ 2120 will place Commander GPMs squarely in the thick of Combat (essential to much more varied, & satisfying, winning game play Tacs & Strats), Borgs viable as DBS Spec Ops and will be informed by the following.
  • - (I.) RL 21st Century Military Doctrine such as (but not limited to):

    a.) Army After Next,

    b.) 4G Warfare,

    c.) NetCentric-Info Warfare

    d.) Asymmetric Conflict

    e.) C3-C41 Logistics and Data Fusion

    f.) Dominant Battlespace Knowledge (aka DBS)

    g.) Air Force Doctrine for Irregular Warfare

    h.) Combined Arms

    - (II.) RL 21st Century Military Tech sponsored by various D.A.R.P.A. initiatives over the last 2 decades such as (but not limited to):

    a.) Collaborative Human-Computer Decision Making for Command and Control

    b.) MIT's Institute for Soldier Nanotechnologies and work on Intelligent Aircraft

    c.) Honeywell's AugCog Soldier Technologies

    d.) Deep Green

    e.) U.A.V.s and M.U.A.V.s - 21CT Air Force Tech

    f.) General Dynamics Littoral Naval Combat Ships

    h.) Self-Repairing Aircraft materials

    i.) Objective Force Warrior (aka, Future Force Warrior, Land Warrior Technologies)

    j.) Programmable Matter

    k.) SOAR Technologies Human Behavior Representation (HBR) / Computer Generated Forces (CGF) and "C4ISR / Command and Control"

    - (III.) Work in the Cognitive Sciences of the last decade will also inform 2120 and is drawn from the following identified areas of explosive and seminal research:

    a.) Cognitive Task Analysis

    b.) Deep Fun

    c.) Body-Mapping

    d.) Flow

    e.) Cognitive Augmentation

    - (IV.) The "wasteland scenario" of 2120 will be informed by the latest science and not rehashing the artistic fictive vision of movies, stories or other games. I have a slew of specific resources for this but you'll have to do your own "leg-work" in this instance because my divulging would fall under the category of "spoilers", IMO.

    - (V.) I've also drawn from the work being done in the field dubbed "Serious Games" (esp. the Delta3D Community).
My last words here:

* Again - you will find many specific references, resources, detailed explanations & illustration graphics to all the foregoing in the The Future of RTS thread.

* My work on "War School: Map-Mod WIP" is also deeply informed by all the proceeding as much as is possible without introducing essential new UIs to improving GPM Switch-Taskings.

* I am very much aware that working from the very same source materials referenced above MANY quite different but still viable implementations are possible..... and lots I cannot even perceive or imagine right now - thus, the skys the limit for any willing to do the work of making something they see uniquely, real.

Regards, whip :cool:
.
EDIT: The following really belongs in my previous post but due to a Paternalism I've rarely encountered since ARPnet days I must put my simple oversight here... oh well...

ADDENDUM to..... Part One: A.I. Experimentation Software Tools and environments I have found helpful over the years...(continued from above post)
.
Last edited by whippersnapper on 04 Jul 2009, 22:20, edited 3 times in total.
.
"I need no warrant for being, and no word of sanction upon my being. I am the warrant and the sanction." Anthem

"Art is the selective recreation of reality according to the artist's metaphysical value judgments." A. Rand
.

User avatar
whippersnapper
Regular
Regular
Posts: 1183
Joined: 21 Feb 2007, 15:46

Re: Commander Command Improvements -- What ? Details ?

Post by whippersnapper » 04 Jul 2009, 15:57

.
Couldn't resist after just reading Alex Lee's (Pumpkin Team, WZ Creators, A.I. Guru) answer to a question posed by Per....
(The italics and bolding of the text are mine...)

alex wrote:
• Was there many things that you were planning to do with commanders that you never got around to ? The source code certainly gave that impression

Yes, commanders did a bunch of extra things during development, but they got scaled back, both the control mechanism and ui got very complicated. We wanted them to order units realistically, not just become a grouping with a bonus and a common goal.
And in answer to this question:
whippersnapper wrote:.

Hello Alex... One thing that would be cool, I think, is if you could speak to Pumpkin's design idea behind WZ 2100 being
a "living product", to quote Nick Cooke (I recall Jim B. and A. McLean referring to it as such as well). With the original
Pumpkin comments lost along with the original bbs this was a key generative commitment that merits re-exposure, from the
"horses mouth", you could say. Thanks.

Regards, whipper (aka, Rman) :)
.
* Again, the italics and bolding of the text are mine...
alex wrote:I don’t remember if we ever used those words, but it was definitely something we were aware of. We played a lot of quake (and later unreal tournament) in the office and one of the things that really struck us about it was the vibrant community. Maps, mods and genuine improvements were everywhere. When working on warzone as engineers we made decisions to leave openings so that people could make changes where they saw fit. Examples including using a scripting language (the slo,vlo’s) that wasn’t compiled into the game so users could edit them, or choosing zip as our packing format instead of something proprietary. I still think it’s a positive thing to do on pc titles, am I’m busy still doing it today!

This ended up benefiting us too. While we always intended to do multiplayer, skirmish started as a bit of a hobby for me. It grew into something fun because the flexibility of the scripting language really opened up the possibilities (thanks to John Elliot)

By demonstrating our commitment to the users through the various rapid patches after release we were trying to build a community of people who would be able to keep making changes to the product. There was an intention to make warzone2120, so it was the feedback from you guys that would have really driven that title.
You can read Alex Lee's postings in their entirety HERE..

Regards, whip :ninja:
.
.
"I need no warrant for being, and no word of sanction upon my being. I am the warrant and the sanction." Anthem

"Art is the selective recreation of reality according to the artist's metaphysical value judgments." A. Rand
.

User avatar
whippersnapper
Regular
Regular
Posts: 1183
Joined: 21 Feb 2007, 15:46

Re: Commander Command Improvements -- What ? Details ?

Post by whippersnapper » 07 Jul 2009, 17:35

.

When I read Alex Lee's recent post about what he did after Pumpkin was disbanded - worked at SEGA USA on their Basketball & Football Sims for 6 years - I thought how cool because I played those games to death and they were also part of my inspiration for improving Commander functioning in the heat of battle. I have a ton of questions I would love to ask Alex about the A.I. he worked on for those franchises but he probably couldn't answer because of signing an NDA.

Last time I shared the connection between Football Sims and RTS was a couple years ago now in this BB - though we had talked about it in the NEWST & Pumpkin-2 dev BBs way back in the day...

In the post quoted below you can readily draw the connection between the A.I.s.... and there is a lovely symmetry (& synchronicity as well) now knowing what Pumpkin's A.I. guru went on to do at SEGA after Pumpkin was disolved.. Stuff like this I find facinating in itself... but I'm easily amused.

Rman Virgil wrote:Re: Any analysis of the Wz CAM Script Triggers ? posted» Sun Sep 09, 2007 8:40 am HERE ------------------>


* This might sound a little weird but here goes...

* Been a fan of Sport Sims for ages: baseball, basketball, soccer, American football.

* My all-time fav is "Madden Football" which as a franchise is appoaching 20 years of a perenially popular existence - multi-millions of unit sales on just about every console & PC platform over that span..

* About 3-4 years ago I had an epiphany: there was much in common between Madden FB and a 3D RTS (maybe that's the root of my passion).

* Then I got to thinking about parallels: create a player / design a unit, franchise mode / campaign, player moves / pathing, player-coach / field-general, player-specialist positions / diff unit types... and so on.

* But the most telling parallels involved the customizable playbooks (offense & defense) that "Madden Football" introduced in 2003 if I recall...

* Remember a couple of customized offensive playbooks with a total of 11 formations - each with 81 Plays... do the math.

* Then this ? occurred to me: "What is a a play..... what is a playbook ?"

* My answer: a play is pre-scripted a.i. you can modify in RT / on-the-fly in game. A Play-book is modular a.i. - a collection of plays (pre-scripted a.i.) that can also be called-up in RT & assigned on-the-fly. And because the gamer can assume control of any player/unit on the field at any time you can create countless variants of play-action..

* Now all this is housed in an iconic menu driven gui that the totally a.i. illiterate can manipulate handily (the likely inspiration for the postulated WZ GCI aka "Global Command Interface").

* Interesting, huh ?

Cheers, RV :)
I would only add one clarification - think Football Sim "play-action" as synonymous with RTS Commander led Combat Group's "field-maneuver"....

Regards, whip :cool:
.

User avatar
Crushy
Trained
Trained
Posts: 58
Joined: 01 Aug 2009, 05:40
Location: Somewhere in Portugal, land of the Cod

Re: Commander Command Improvements -- What ? Details ?

Post by Crushy » 05 Aug 2009, 06:15

I would just like to make one simple suggestion: Usually when you order a commander he will move to where you clicked and bring his whole swarm with him. Sometimes he's even the first unit there, which almost certainly kills him

How about we make it sort of a reversed sensor vehicle? Instead of artillery following a sensor vehicle around, people would order their comm to move assigned units to where we clicked. The commander should use some algorithm to pick the most "central" unit in the group and follow it around. Logicaly, lone commanders would work like regular units.

There, commanders aren't as fragile any more :)

Another idea: Unify sensor and commander assignments. If you have an attached sensor unit and press shift (for example, any other key would work) while left-clicking an enemy unit, that sensor unit will move up and target the enemy. Likewise, artillery units in your swarm would automatically make use of these spotters. Should the spotter die, a commander would also automatically request one (maybe make this a check-mark on his options and also a check-mark on the design screen for designating the spotter a commander would order from a factory). The idea is making this work with dynamic factory reinforcements someone suggested.

This has many advantages: It makes commanders feel more like they're making your job easier and therefore immerses you more in the game, and makes mobile artillery units not run the risk of becoming "orphans" without their designated spotter.

We should also change sensor assignments so it doesn't work based on a fixed "safe" distance. Assigned units should keep further back according to their firing range.

Other minor suggestions: Shift could be use for artillery strikes on a target, Ctrl could be used to direct VTOL strikes (again, using a dedicated VTOL strike vehicle) Shift and/or control dragging would designate an area for strikes. Holding both keys would make a combined artillery and air attack. Spotters are independent of other group orders while carrying out these missions.

Example: You hold Control and Shift to drag a rectangle around small outpost for bombardment and VTOL attacks. You left click in the middle of the outpost afterwards. Your direct combat units swarm the outpost while your artillery and spotters keep bombarding the areas your marked until there are no more targets or you hit a special "Abort all operations" button.


All of these are rather simple to implement compared to all your awesome ideas, but would make commanders useful that much quicker. Perhaps not enough, but it could be a start.


----------------------------------------------------

In another note, just to add to someone else's article:
Has anyone ever played the first Perimeter? The AI is absolutely amazing and as soon as it's completely turned on (you'll notice when you see it) it will start kicking your ass. It would attack your least defended positions, it would use the most effective units it could, it would deploy artillery to support itself, it would surround you... The only way you could win was to "cheat" by capturing parts of it's base. In fact, a main part of perimeter was disabling parts of the enemy base and capturing (think power-lines and taking them out to replace them with your own). That sort of AI is no fun in a conventional RTS because you know you will never outwit it. Perimeter was fun because it gave you a way to outsmart it in another area.
Current status: Sick and Learning to model in blender. Failing at both.

Post Reply