def update(self, state, action):
print("ENVIRONMENT UPDATE: ", state, action)
# given the current state, determine reward and next state
next_state = board.add_move(action['player'], action['cell'], state)
reward = 0.0
if play.is_winner(next_state, action['player']):
reward = 1.0
self.is_complete = True
self.outcome = action['player'] + "_WINNER"
elif play.is_cat_game(next_state):
self.is_complete = True
self.outcome = "CAT_GAME"
if not self.is_complete:
# OTHERWISE, let the OTHER player play
opponent_id, opponent_cell = self.opponent.pick_next_move(
next_state)
next_state = board.add_move(opponent_id, opponent_cell, next_state)
if play.is_winner(next_state, opponent_id):
reward = -1.0
self.is_complete = True
self.outcome = opponent_id + "_WINNER"
elif play.is_cat_game(next_state):
self.is_complete = True
self.outcome = "CAT_GAME"
return next_state, reward
def model_environment(opponent, state, action):
game_complete = False
initial_board = state
file.write("AGENT MAKING MOVE: " + str(action) + str(board.to_state(action)) + "\n")
current_board = p.add_move('X',action,initial_board)
print("AFTER AGENT MOVE:")
print(game.to_display_string(current_board))
file.write("AFTER AGENT MOVE:\n")
file.write(game.to_display_string(current_board))
reward = 0.0
if p.is_winner(current_board,'X'):
game_complete = True
reward = 1.0
elif p.is_cat_game(current_board):
game_complete = True
reward = 0.0
if not game_complete:
# let the opponent make a move ...
(opponent_id, opponent_move) = opponent.pick_next_move(current_board)
current_board = p.add_move(opponent_id, opponent_move, current_board)
print("AFTER OPPONENT MOVE")
print(game.to_display_string(current_board))
file.write("AFTER OPPONENT MOVE\n")
file.write(game.to_display_string(current_board))
if p.is_winner(current_board,opponent_id):
game_complete = True
reward = -1.0
elif p.is_cat_game(current_board):
game_complete = True
reward = 0
return current_board, reward, game_complete
def play_game(p1, p2, file=None):
board = p.empty_board()
players = [p1, p2]
current_player_index = 0
winner = None
move_count = 0
while (True):
print("Current move is for player: ", players[current_player_index].player)
if (file is not None):
file.write("PRIOR TO MOVE " + str(move_count) + " ------------\n")
file.write(to_display_string(board))
if p.is_cat_game(board):
if (file is not None):
file.write("RESULT IS CAT GAME")
break
m = players[current_player_index].pick_next_move(board)
board = p.add_move(m[0], m[1], board)
p.display_board(board)
move_count += 1
if p.is_winner(board, players[current_player_index].player):
winner = players[current_player_index]
if (file is not None):
file.write("FINAL BOARD AFTER MOVE " + str(move_count) + " WINNER IS: " + winner.player + "\n")
file.write(to_display_string(board))
break
# alternate players
if (current_player_index == 0):
print("Switching to player 1...")
current_player_index = 1
else:
print("Switching to player 0...")
current_player_index = 0
if (winner is None):
print("CAT GAME")
else:
print("WINNER IS PLAYER: ", winner.player)
return winner
def get_reward(self, board, action):
next_board = action['board']
print("In REWARD given next board -----")
p.display_board(next_board)
if p.is_winner(next_board, self.player):
return 1.0
elif p.is_potential_loser_on_next_move(next_board, self.player):
print("POTENTIAL LOSER BOARD: ")
self.log.write("POTENTIAL LOSING BOARD -----\n")
self.log.write(self.to_string(next_board))
p.display_board(next_board)
return -0.5
else:
return 0.0
def pick_next_move(self, current_board):
# identify all available cells
available_cells = []
for row in p.get_rows(current_board):
for col in p.get_cols(current_board):
cell = (row, col)
if (not p.is_occupied(current_board, cell)):
available_cells.append(cell)
for cell in available_cells:
tmp_board = board.add_move(self.player, cell, current_board)
if p.is_winner(tmp_board, self.player):
return self.player, cell
# otherwise, pick move at random
n = len(available_cells)
inx_select = np.random.randint(n)
return (self.player, available_cells[inx_select])