Archived

This topic is now archived and is closed to further replies.

Secondary Player vs AI

This topic is 5110 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

So I am working on my vehicle building program, and I come to a realization. A computer program cannot be designed to pickup on unique uses of a custom unit. If a designer creates a unit that looks exactly like enemy units down to the color a computer wouldn’t figure out that the unit would be well used as a spy or for reconnaissance. It is impossible for the computer, no the program, no the programmer, to handle the vast array of possibilities. I think that is some of the argument behind paper and pencil versus MMORPG, correct? The best option is for the designer to explain its uses to a player and let the player implement it. So that said… let me see what you guys think about this idea. Have there be two types of players for some RTS game (hell even fps… later) Type 1: Player – same as any RTS player now. Except that before he plays online, he has the option to pick a type 2 person to work with him. Type 2: Designer – designs units midgame. Can save a certain number of well thought out, well tested designs. When the game started the player could build either standard (read el crapo) units or he could build the custom units made by the designer. Certain components would only be available from certain buildings so you would still be required to do the normal nation building. Or--- (oh this is good) the components would be available on a global market with market economics active - so designers who made good use of little used components would be sought after by players who understood the game. The best designers would build useful units from cheap parts. What do you guys think? Dauntless, I am interested in your view especially. [edited by - Thermodynamics on December 16, 2003 5:26:18 PM]

Share this post


Link to post
Share on other sites
Just a small nitpick: your comment about the computer being unable to use the identical unit for reconnaissance might not be valid; if this was a game with a minimap where you could identify your units and enemy units, the enemy would immediately realize that the unit is not her own. They might not be paying attention for some time, but AI that makes units auto-attack targets would immediately recognize it.

Unless you''re intending to not have computer players, or to program this into your game.

Share this post


Link to post
Share on other sites
One of the components could be a threat recognizer. Basically program it to work so "if it looks like a duck and it sounds like a duck it must be a duck." So you either build your spies at a distance from other units or you put something akin to a transponder signal in the spy unit. It all depends on the components available.

Share this post


Link to post
Share on other sites
Maybe your camo AI could work like this...


line of | enemy | colors | sighted?
sight | moving | like tile |
--------+--------+-----------+-------------
blocked | * | * | no
--------+--------+-----------+-------------
clear | no | none | yes
| | some | high chance
| | yes | low chance
--------+--------+-----------+-------------
clear | yes | none | yes
| | some | yes
| | yes | high chance
--------+--------+-----------+-------------
or if enemy is firing... yes!

Share this post


Link to post
Share on other sites
I don''t think I''m getting the same message from your post that previous responders are getting. What I see is a concern that in a game where the player customizes vehicles or loads out weapons packages, the AI operator of that vehicle will be unable to determine the strengths and weaknesses of the design. Since the AI won''t know how to use it, even a terrific design could fail. After all, a super-stealthy craft with a surgical weapon set will fare badly in toe-to-toe combat, and a heavy bomber with a thousand nukes won''t get through the sensor array to reach its target.

My solution is human designation of purpose. There are different ways to do it.

One is to define different "behavior patterns" for the AI to adhere to, as if the AI "pilots" have received specialized training. So you build your stealth vehicle and then pilot it with a "spy" class pilot. He''s more inclined to use cameras and cloaking devices, and will only bust out the disintegration beam when his butt is already in the stew. Or build a juggernaut craft and put a "cowboy" pilot in it, who will shoot first and mail the questionaire to next-of-kin. A handful of pilot types, from "spy" to "cowboy" to "recon" to "transport" to "slingshot" to "courier" to "surveyor" to "support" to "escort" to "assault" to anything else you want to include would be sufficient. In fact, this would allow you to use the same craft for multiple purposes. Build three of that turbo gunship (reducing production costs), and you can put a "cowboy" in one, a "recon" in another, and an "escort" in the third. Like the Joint Strike Fighter plane used by the Air Force, Marines and Navy, with slightly different features. A real money saver.

Another possibility is to customize a pilot for each craft. This is of course more time consuming and might not jive with your overall intent. I don''t like the idea myself, so I won''t pursue it.

A third possibility is some kind of AI ability to assess the strengths of a craft, or, if it''s possible, a behavior routine that actually reassesses the craft''s functionality every time it reassesses the situation in which it finds itself.

The first and third, to varying degrees, put rigid boundaries on the way the craft can be used, and so limit the innovation that is possible. After all, if you develop a ship that''s just right for running the Death Star''s equatorial trench and planting a proton torpedo in the exhaust port and you don''t have a "Luke Skywalker" AI type and the fluid AI can''t "imagine" that as a feasible course of action, you''re boned. I doubt you could put an "assault" pilot in there and then just explain to him in a "planning phase" that he has to get into the trench, stay there, and when he gets to the end put a glowing ball of fusion down the little hole. Or maybe you can. I''m no programmer.

Share this post


Link to post
Share on other sites
Iron CC, you did it again! That is what I am talking about.

The only way you can create a game that would allow you Skywalker scenario is if the ship was created especially for that purpose and you had a living person controlling the ship. An AI could not be built broad enough to handle either situation no matter how many details about it were described by the player.

The effect I am going for is to allow custom unit to be used in custom ways. Simply having AI modifiers like "sharpshooter" or "balls to the wall carnage" is too limiting. And yet there will be too much detail work for one person in real-time.

unless..... You build your units between battles. No! that would eliminate the market system for components.

Is there a team based game on the market already? I don''t mean a game like starcraft where groups can play on a team*. The closest game I can think of is AOE2 where two people can be the same Player. not team, player. at the same time. Is the another game that does this?

* As a side note, I have played starcraft in a team fashion where the Toss were defense and Zerg was offence. But thats not the point.

So I am back to my original topic. Would a team game be fun to play? Personally I feel better knowing I have someone at my back to help in a crunch rather than a computer that keeps chopping wood or whatever in the middle of an attack.

Share this post


Link to post
Share on other sites
Yes, team play with specialization is good. You see this in RPG parties where tasks are divided among the classes. I''ve played AOE too where players team up on the same side and one runs production and the other manages the combat and it''s fun for both.

Also I''ve been on text-based muds an economic system where some players played an enchanter class that did little else but hone their skills of enhancing items for other players. Other''s played a tinker class that forged the best armour. They could mark the items. Players paid high prices for items marked by certain grandmaster''s names on them. They would also pay high prices for summoners and necromancers to provide them with guardians, or tinkers to craft traps, or enchanters to set up protective runes to protect their castles from raids.

Share this post


Link to post
Share on other sites
The same was true in UO. Grandmaster Blacksmiths in UO were highly sought after. I want to create the same effect with the designers.

Each designer could store units that do particularly well or have unique uses. When a player teamed with a well known designer he could expect to get some decent units. In truth design is my ultimate goal but I feel the best way to weed out the good from the bad is with other people playing.

Share this post


Link to post
Share on other sites
Have you ever played warzone 2100? It was an admittedly mediocre RTS, but it used this "build your own units out of components" idea. It had obvious flaws, in that 99% of the time there was an all-out-bestest combination of parts. Your market for parts idea could help fix that I'd say. Please add a wombat-array-artillery piece to your game. Thx.


edit: also starcraft DID have a mode where players all controlled the same base. Team Melee i think it was called? You could fight over control of units and it would swerve back and forth in a confused manner.

[edited by - CombatWombat on December 18, 2003 10:01:43 PM]

Share this post


Link to post
Share on other sites
You''re welcome to use any of my ideas. (notice the correct use of "You''re")
I realize that I am not going to make the next blockbuster game by myself. I would rather see a game made using an idea of mine, just so I could play it. I have one condition for using my ideas though... You gotta let me be a beta tester

Share this post


Link to post
Share on other sites
It''s quite interesting what you are saying about an AI making the best use of a player designed unit. It is infact what I was thinking about as I clicked the thread. I am currently making a game (slowly but surely where the player designs units with nearly infinite variation and then the AI battles it out with the units. The player can then watch the battles and tweak the behaviours. I''m providing some default AI scripts/behaviours like shoot everything in sight, go for the flagship etc but I need a good way to allow the user to define highly custom behaviours. For instance the best way for a player to win might be to fight a retreating war due to their design so in the AI script they will say something along the lines of "Move away from nearest enemy" but if his oponent has "Surround Enemy" then it''s clearly not going to work and the script requires more specialisation. Through the battle viewer interface the player can tweak the AI behavious though by adding cases such as if(this is occuring) then do special behaviour X but this leads to so many special cases that things get to be a bit of a headache. This game uses no random elements so the idea is the player makes the best use of what he has and with the massive number of variations tactics are almost endless and I''m hoping this wins out over potentially silly AI making it look stupid. If it looks like the player won because of a concientous design decision rather than better parts etc then I''ll be happy.

To clarify this is a simultaneous turn based game where the player designs stuff and gives orders and acts on the results when he gets his next turn.

Share this post


Link to post
Share on other sites