# Tip on turn based game AI

I'm quite new in video game AI, and I'm trying to design an AI for a turn based game.

The rules are simple:

• The battle field is represented by an NxM board.
• The game is between two team, each team has X character.
• Each character, based on his type, can move within C cells and can attack within A cells.
• On each turn a player can move and then attack (not attack and then move ) with each one of the caracter in his team (time limit).

I'm thinking on how design the AI for this kind of game.

One solution I have thought is a Multi Agent System composed by two kind of agents: The first one charachert Agent calculate on each step a function SCORE for each of the cells reachable by the character, and the second one, team Agent calculate the new position for each character based on the score calculated by the former agent. Each character can then attack an opponent if there is one reachable.

Is it a good solution?

Can someone suggest a different solution or give me some tips?

• How are you going to calculate the score of each cell? And what does the team agent do what the character agent can't do? – Philipp Dec 31 '15 at 12:48
• stackoverflow.com/questions/4549986/… Not a straight copy of your question but the answer there is really complete, you might get some infos from that ! – Alexandre Beaudet Dec 31 '15 at 13:21
• @Philipp I was thinking on how to do it. The basic idea is to calculate a function influenced by some factors as: enemy position (number reachable opponent characters), character health, opponent reachable character health, characher and opponent weapon power and so on. So mixing some of those values I should have a "score" for each cells. It is only a basic idea I'm still studing how to do it. – GGC Jan 5 '16 at 16:05
• @AlexandreBeaudet it is a really good answer. Anyway I was trying to avoid algorithm as minimax or A* because I think the generated tree (even if I cut) could be really huge (each team can move or attach with all the character each turn). And this is basically one of my questions...is this assumption (huge decision tree) wrong? – GGC Jan 5 '16 at 16:20