//Permet de simuler le GS d'un etat donne jusqu'a un etat final
public float SimulateGameState(ref Dictionary <TicTacToe.GameState, Vector2Int> policy,
ref List <TicTacToe.GameState> exploredState,
ref TicTacToe.GameState gs)
{
int playerTurn = 0;
bool gameEnd = false;
// IA = Rond, donc playerWinner = 1
int playerWinner = -1;
while (!gameEnd)
{
//Je prend l'action possible
var c = gs.GetAvailableCells();
var selectedCell = c[Random.Range(0, c.Count)];
var clone = gs.Clone();
if (playerTurn == 1)
{
var test = policy.Keys.ToList();
// Epsilon greedy = range entre 0 - 10, et on tire RDM, si en dessous de Epsi action rdm, sinon policy
// a Utiliser quand on fait la simulation d'un episode
var gridToTest =
GetIndexOf(ref test,
ref clone.Grid); //policy.Any(x => x.Key.Grid == clone.Grid);//policy.ToList().FirstOrDefault(x => x.Key.Grid == clone.Grid).Key;
if (gridToTest >= 0)
{
if (Random.Range(0.0f, 1.0f) > epsilonGreedy)
{
selectedCell = (policy[policy.ElementAt(gridToTest).Key].x,
policy[policy.ElementAt(gridToTest).Key].y);
}
}
else
{
policy.Add(clone, new Vector2Int(selectedCell.Item1, selectedCell.Item2));
}
exploredState.Add(clone);
}
else
{
exploredState.Add(clone); //new Vector2Int(selectedCell.Item1, selectedCell.Item2));
}
// Grid, N, R, ...
if (ticTacToe.SetCellWithoutChangeGraphics(playerTurn, selectedCell.Item1, selectedCell.Item2, ref gs))
{
var victoryState = ticTacToe.CheckVictory(ref gs);
if (!victoryState.Item1)
{
if (!ticTacToe.CheckNullMatch(ref gs))
{
ticTacToe.NextTurn(ref playerTurn);
}
else
{
gameEnd = true;
playerWinner = victoryState.Item2;
}
}
else
{
gameEnd = true;
playerWinner = victoryState.Item2;
}
}
}
return(playerWinner == 1 ? winR : playerWinner == 0 ? nulR : loseR);
}