Archived

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

T-man

Player vs. Super AI Game

Recommended Posts

Hi, I posted this in the help wanted forum, but got no luck so far. What I'm wondering is if any of you have an idea of how to do a game with one player against one AI. NO board games, fighting games (Mortal Kombat, etc.), or otherwise totally overdone games. I'm looking for something fairly original. Here's an example I've come up with, but personally I don't like the style, it's just an example of the kind of thing I'm talking about. You are a bounty hunter in the year 2020(?). You accept a mission to kill/capture someone. The employer transports you to the location. Once there it's like a two player search and destroy (or capture, as the mission dictates) game, except that one of the players is an AI. Any comments, etc. would be greatly appreciated. P.S., if you don't want to post your ideas here email me with them and we can talk. Edited by - T-man on July 23, 2001 11:47:34 PM

Share this post


Link to post
Share on other sites
It sounds like you want to concentrate on an easy to implement game. I''d go with a topdown 2D arcade if I were you, as it will be both fun and feasible.

You can throw a lot cool things into the mix: Powerups, different weapons, and maybe cool special moves.

You''ll prolly spend a lot of energy doing AI if you want to make it a good match for the player. I''d think a lot about making the AI player and human player equal, and figuring out first excellent pathfinding, then next smart evasion and pursuit strategies.

Unless you''ve got great skills, I''d steer clear of FPS or complicated 3D games. If you want a real challenge, figure out how to do your game text-based w/o losing any of the excitement. That would show some real design skills!

--------------------
Just waiting for the mothership...

Share this post


Link to post
Share on other sites
No, I'm not looking for an easy to implement game. It's just that I'm really into doing AI coding. It's the part of game programming that is the most fun for me. I want to have good graphics by today's standards, but with a great AI, again by today's standards (which is not hard to exceed btw). I basically just want to work on a game where I can be free to develop the AI as much as I can without sacrificing CPU speed for graphics, because the "gameplay" part, no matter how complicated doesn't take all that much CPU time, it's the graphics that do.

FPS, not my style (they just plain suck)... same with most text-based games.

BTW, Wavinator, what excatly made you think I was looking for an easy to implement game?

Edited by - T-man on July 23, 2001 12:20:41 AM

Share this post


Link to post
Share on other sites
Why do you want your AI to compete against the player? What is your AI doing that makes you want to compete against a lve opponent?

Aren''t you concerned that players of your game will think the AI can cheat and look at how the game is working - where the player will not be able to? How do you convince them the AI isn''t cheating?

Dash Zero
Credits: Fast Attack - Software Sorcery - Published by Sierra 1996

Share this post


Link to post
Share on other sites
Well, have you thought about simulating some sort of behavior? Say, a gerbil or something that the player can interact with?

You could have it learn about the enviroment and the player''s ations, and store it somewhere, causing it to learn over time to expect the player''s actions?

Share this post


Link to post
Share on other sites
Kind of like that story I think it was "The Ultimate Game" where some guy goes to an island to play a game with some rich hunter. The game is that whoever wins gets the fortune. To play each is given a gun and the one who kills the other wins. The visitor gets a 10 minute head start to hide/learn the terrain/plan the attack.

Powerups and what not would be unnceccary in such a game. Single shot to kill would be more fun and challenging.

I would start with 2D to work on AI for the computer hunter and then switch to 3D tiles. Or just implement field of view for the player. Top down but the computer can duck behind trees or whatnot and disappear from view. They would also have only a field of view to work with. The player can also hide behind things out of view.

The computer would have to learn the terrain. You may also beable to implement the ability to see where the human player is looking by watching their head.

I''d work in a split screen for testing purposes. So you can see exactly what each is looking at.

Ben
http://therabbithole.redback.inficad.com

Share this post


Link to post
Share on other sites
Well, I suggest you do some type of pre-compilation of situation calculus into a partitioning of graphs and then apply resolution refutation using the graph partitions as a way to constrain the search. Given queries to the knowledge base, the agent should be able to accept or reject new incoming knowledge from percepts and communication channels and build a model of its environment. This, combined with inference and unification via a set of axioms modeled in situation calculus will enable the agent to plan ahead using what it believes about the world. Simultaneously, given constant perception, these plans can adapt or be canceled given new incoming information about the agent''s world. Also, if you model the percepts at a low enough level, you should be able to model the behavior which arises from distraction. If one of your percepts are vision, and the agent must scan its enviroment left and right, you will get the simulated and necessary behavior which arises from limited vision and distraction from peripheral vision. The beauty of such a model is the synergies which arise out of both a reactive and rational agent.

Yeah, others are as interested in (or more interested in) AI as you are.

Share this post


Link to post
Share on other sites
Finally some decent replies.

DashZero,
"Aren''t you concerned that players of your game will think the AI can cheat and look at how the game is working - where the player will not be able to? How do you convince them the AI isn''t cheating?"

-I guess the player will just have to believe...

ThoughtBubble,
But how could you make a behaviour simulation fun to play? I can''t think up any good scenerios to do. Besides, I''m not sure if I''d want a smarter digital version of Furby loose in this world.

KalvinB,
I remember seeing that in a movie. In the movie there were several hunters after the one guy (I think Ice-T played the prey), but in the game there could be just one hunter. The game could have a choice as to whether the AI is the hunter or the prey, as well as several terrain choices. Hey, I think we have a winner here! Thanks KalvinB.

Share this post


Link to post
Share on other sites
Ok. If you had wanted to keep the behavior simulator, add some danger. Now, give the player an avatar, and instead of fuzzy creatures, we have big pack animals with pointy teeth. Give them some basic patterns (dominance, breeding, territory, feeding, etc)and the player some tools that will interact with them (mating calls, things to hide behind, scent, big sides of meat, traps). Then finally determine a few goals (possibly on a per stage basis) for example, catching (without hurting) the alpha male, or simply surviving through their territory.

Then, if you were really crazy, find a way to parameterize and randomize the relationships to different items. So that one time the creatures may be agressive and dangerous, another time they may be flighty, and so on.

Another spin could be that you have an area to defend, and the creatures are trying to eat the people therin (sorta like the ailiens movies).

Hope that helps.

Share this post


Link to post
Share on other sites
There was a 2D action game that had quite a few missions similar to the situation that you described. The name is Hunter/Hunted and it was created by the k.a.a. division of Sierra. (now the division is dead) It was an excellent game, and the missions where it was just you against one other opponent were most excellent.

I am a fan of 3D action games and I would eagerly await one where the entire scope of the game was hunting an equal or percievably (not sure if that is a real word) greater enemy.

P.S.
I believe the short story KalvinB refered to was _The Most Dangerous Game_, written by Richard Connel.

For all those that tried, and haven''t gone and died. I salute you.

Share this post


Link to post
Share on other sites
quote:
Original post by T-man
FPS, not my style (they just plain suck)... same with most text-based games.


Uh, okay. (?)

quote:

BTW, Wavinator, what excatly made you think I was looking for an easy to implement game?



Pardon, T-man, you were a new (to me) poster here. I probably would have had a better idea if you''d maybe given a bit more detail (technical especially). That''s the trouble with the forums, no one knows what skill level everyone''s at.

As for AI suggestions, what about either 2D or 3D isometric in a variety of hostile environments. Think Prince of Persia, only iso, or Lode Runner, with traps, ropes, cover, etc. This should be a lot less CPU intensive, and won''t make the motion sick hurl.

The fugitives could be in this environment, trying to escape, and you''ve got to outwit the land and them in order to stop them.

The juicy AI elements I see are actually figuring out first the game rules for this hostile area, then how the AI should play. If the environment is significant, then discovering optimal strategies will be difficult. If players can create content, then you get into real-time map analysis and lots of interesting predictive challenges.

If you want to make things even tougher on yourself, make the entire environment maleable. If the player can dig a bambo trap, or wind back tree branches to cause damage, or roll boulders downhill, then the AI should be able to do this, too. Now you''re into figuring out how to bait and lure players, something which hasn''t been done before.


--------------------
Just waiting for the mothership...

Share this post


Link to post
Share on other sites
Player vs AI?

Okay... let''s see... I think AI is not developed enough yet to be an even match to any player (without cheating).

So... you''d have to make the AI more powerful from the get-go.

Also, I think that the intelligence of the AI side, no matter how smart you make it, will never feel real. So I''d always opt for a robot.

Think Terminator.

You as the player play the human part, the AI takes control of the more powerful robotic Terminator.

To make it interesting for the player, you''d have to give the AI unpredictable predictability. What I mean by that is that the AI should behave in a way that can be expected... but sometimes will do unexpected things.

Simple example:
If you throw a rock to a location, it''ll make a sound. The predictable behaviour should be for the AI to hear the sound and go and investigate. The unpredictable part of it, could be that he suddenly turns around.

"Unpredictable predictability" I like that

Woohoo... I''m on day 4 on my C++ in 21 days course. %Another two weeks and I''ll be a master programmer!%

Share this post


Link to post
Share on other sites
Player against AI? Well, I guess my game will feature that, though not in the way you expect it

I have been designing a neural net to ''control'' the NPC''s, and more important, to generate good quests.

So, you are more or less strugling against my AI :p

Share this post


Link to post
Share on other sites