def __init__(self, board, parent, color, action):
self.parent = parent # 父节点
self.children = [] # 子节点列表
self.visit_times = 0 # 访问次数
self.board = board # 游戏选择这个Node的时的棋盘
self.color = color # 当前玩家
self.prevAction = action # 到达这个节点的action
self.unvisitActions = list(board.get_legal_actions(color)) # 未访问过的actions
self.isover = self.gameover(board) # 是否结束了
if (self.isover == False) and (len(self.unvisitActions) == 0): # 没得走了但游戏还没结束
self.unvisitActions.append("n")
self.reward = {'X': 0, 'O': 0}
self.bestVal = {'X': 0, 'O': 0}
def tree_search(self, board, color):
"""
实现UCT方法顶层
"""
# 只有一种选择
actions = list(board.get_legal_actions(color))
if len(actions) == 1:
return list(actions)[0]
# 创建根节点
curBoard = deepcopy(board)
root = Node(curBoard, None, color, None)
# 时间限制60s以内
try:
func_timeout(59, self.iterate_function, args=[root])
except FunctionTimedOut:
pass
return self.best_child(root, math.sqrt(2), color).curAction
def search(self, board, color):
# board: 当前棋局
# color: 当前玩家
# 特殊情况：只有一种选择
actions = list(board.get_legal_actions(color))
if len(actions) == 1:
return list(actions)[0]
# 创建根节点
newboard = deepcopy(board)
root = Node(newboard, None, color, None)
# 考虑时间限制
try:
# 测试程序规定每一步在60s以内
func_timeout(59, self.whileFunc, args=[root])
except FunctionTimedOut:
pass
return self.best_child(root, math.sqrt(2), color).prevAction
def game_over(board):
l2 = list(board.get_legal_actions('O'))
l1 = list(board.get_legal_actions('X'))
return len(l2) == 0 and len(l1) == 0
def gameover(self, board):
l1 = list(board.get_legal_actions('X'))
l2 = list(board.get_legal_actions('O'))
return (len(l1) == 0 and len(l2) == 0)