public void DebugPrint(State _state)
{
TicTacState state = (TicTacState)_state;
Console.WriteLine("-----");
for (int i_row = 0; i_row < 3; i_row++)
{
Console.Write(" ");
for (int i_col = 0; i_col < 3; i_col++)
{
if (state.grid[i_row, i_col] == -1)
{
Console.Write(debug_print_toggle?"O":"X");
}
else if (state.grid[i_row, i_col] == 1)
{
Console.Write(debug_print_toggle?"X":"O");
}
else
{
Console.Write(" ");
}
}
Console.Write("\n");
}
Console.WriteLine("-----");
debug_print_toggle = !debug_print_toggle;
}
public override bool IsEndState(State _state)
{
TicTacState state = (TicTacState)_state;
// Game ends if the grid is full...
bool found_empty = false;
for (int i_row = 0; i_row < 3; i_row++)
{
for (int i_col = 0; i_col < 3; i_col++)
{
if (state.grid[i_row, i_col] == 0)
{
found_empty = true;
break;
}
}
if (found_empty)
{
break;
}
}
if (!found_empty)
{
return(true);
}
// ...or the reward is large, so other player has won
if (Math.Abs(Reward(_state)) > 10)
{
return(true);
}
return(false);
}
public override State TogglePlayer(State _state)
{
TicTacState old_state = (TicTacState)_state;
TicTacState new_state = new TicTacState();
// Change the sign of each element, i.e. swap X and O
for (int i_row = 0; i_row < 3; i_row++)
{
for (int i_col = 0; i_col < 3; i_col++)
{
new_state.grid[i_row, i_col] = -old_state.grid[i_row, i_col];
}
}
return(new_state);
}
public override bool IsSameState(State _state1, State _state2)
{
TicTacState state1 = (TicTacState)_state1;
TicTacState state2 = (TicTacState)_state2;
for (int i_row = 0; i_row < 3; i_row++)
{
for (int i_col = 0; i_col < 3; i_col++)
{
if (state1.grid[i_row, i_col] != state2.grid[i_row, i_col])
{
return(false);
}
}
}
return(true);
}
public override List <Action> PossibleActions(State _state)
{
TicTacState state = (TicTacState)_state;
List <Action> actions = new List <Action>();
// Go through the grid and list empty ones as possible actions
for (int i_row = 0; i_row < 3; i_row++)
{
for (int i_col = 0; i_col < 3; i_col++)
{
if (state.grid[i_row, i_col] == 0)
{
actions.Add(new TicTacAction(i_row, i_col));
}
}
}
return(actions);
}
public override State CreateNextState(State _state, Action _action)
{
TicTacState state = (TicTacState)_state;
TicTacAction action = (TicTacAction)_action;
// Copy existing state...
TicTacState new_state = new TicTacState();
for (int i_row = 0; i_row < 3; i_row++)
{
for (int i_col = 0; i_col < 3; i_col++)
{
new_state.grid[i_row, i_col] = state.grid[i_row, i_col];
}
}
// ...and set one square according to the action
new_state.grid[action.row, action.col] = 1;
return(new_state);
}
public override State CreateFirstState()
{
TicTacState state = new TicTacState();
// Empty grid
for (int i_row = 0; i_row < 3; i_row++)
{
for (int i_col = 0; i_col < 3; i_col++)
{
state.grid[i_row, i_col] = 0;
}
}
/*
* // Build example situation for testing
* state.grid[0, 0] = -1;
* state.grid[0, 1] = -1;
* state.grid[0, 2] = 1;
* state.grid[1, 0] = 1;
* state.grid[2, 0] = -1;
*/
return(state);
}
public override double Reward(State _state)
{
TicTacState state = (TicTacState)_state;
int i_sum = 0;
// Each row
for (int i_row = 0; i_row < 3; i_row++)
{
i_sum = 0;
for (int i_col = 0; i_col < 3; i_col++)
{
i_sum += state.grid[i_row, i_col];
}
if (i_sum == 3)
{
return(REWARD_WIN);
}
if (i_sum == -3)
{
return(REWARD_LOSE);
}
}
// Each column
for (int i_col = 0; i_col < 3; i_col++)
{
i_sum = 0;
for (int i_row = 0; i_row < 3; i_row++)
{
i_sum += state.grid[i_row, i_col];
}
if (i_sum == 3)
{
return(REWARD_WIN);
}
if (i_sum == -3)
{
return(REWARD_LOSE);
}
}
// Diagonals
i_sum = 0;
for (int i = 0; i < 3; i++)
{
i_sum += state.grid[i, i];
}
if (i_sum == 3)
{
return(REWARD_WIN);
}
if (i_sum == -3)
{
return(REWARD_LOSE);
}
i_sum = 0;
for (int i = 0; i < 3; i++)
{
i_sum += state.grid[i, 2 - i];
}
if (i_sum == 3)
{
return(REWARD_WIN);
}
if (i_sum == -3)
{
return(REWARD_LOSE);
}
// In case neither won, the immediate reward is small and negative.
// That guides the strategy to use minimum amount of steps to win.
return(-0.1);
}