game_dqn

DQN 游戏AI

代码： """ 物品： 空地：0 自己：1 敌人：2 墙：3 剑：4 盾牌：5 已经走过：6

动作： 'u':0 'd':1 'l':2 'r':3 'u_l':4 'u_r':5 'd_l':6 'd_r':7 'stop':8

反馈定义： 越界：-100 走到墙：-100 走到敌人：-100 走到剑：+100 走到盾牌：+100 走到已经走过的：-100 走到敌人上下左右的格子：+100 走到空地：0 当停下来时，是起始点：-100 当停下来时，不是起始点： 空地：0 周围有一扇墙：+10 周围有两扇墙：+20 周围有三扇墙：+30 周围有四扇墙：+40

终止条件定义： 撞到墙，已经走过的格子，敌人的格子 自己选择停止

"""

import numpy as np np.random.seed(1)

class Game(object): def init(self): self.action_space = ['u', 'd', 'l', 'r', 'u_l', 'u_r', 'd_l', 'd_r', 'stop'] self.n_actions = len(self.action_space) self.map_h = 12 # grid height self.map_w = 12 # grid width self.map = np.zeros([12, 12])

# 随机初始化地图,wall_num为墙的数量
def init_map(self, wall_num=2):
    # 设置我的位置
    my_pos_x, my_pos_y = self.get_random_pos()
    self.map[my_pos_x,my_pos_y] = 1

    # 设置敌人位置
    enemy_pos_x, enemy_pos_y = self.get_random_pos()
    while self.map[enemy_pos_x, enemy_pos_y] != 0:
        enemy_pos_x, enemy_pos_y = self.get_random_pos()
    self.map[enemy_pos_x, enemy_pos_y] = 2

    # 设置墙的位置
    for i in range(wall_num):
        wall_pos_x, wall_pos_y = self.get_random_pos()
        while self.map[wall_pos_x, wall_pos_y] != 0:
            wall_pos_x, wall_pos_y = self.get_random_pos()
        self.map[wall_pos_x, wall_pos_y] = 3

    # 设置剑的位置
    sword_pos_x, sword_pos_y = self.get_random_pos()
    while self.map[sword_pos_x, sword_pos_y] != 0:
        sword_pos_x, sword_pos_y = self.get_random_pos()
    self.map[sword_pos_x, sword_pos_y] = 4

    # 设置盾的位置
    shield_pos_x, shield_pos_y = self.get_random_pos()
    while self.map[shield_pos_x, shield_pos_y] != 0:
        shield_pos_x, shield_pos_y = self.get_random_pos()
    self.map[shield_pos_x, shield_pos_y] = 5

    # 设置已经走过的格子
    walked_pos_x, walked_pos_y = self.get_random_pos()
    while self.map[walked_pos_x, walked_pos_y] != 0:
        walked_pos_x, walked_pos_y = self.get_random_pos()
    self.map[walked_pos_x, walked_pos_y] = 6

def set_map_pos(self, pos_x, pos_y, goods):
    self.map[pos_x, pos_y] = goods

def get_random_pos(self):
    pos_x = np.random.randint(0, self.map_h)
    pos_y = np.random.randint(0, self.map_w)
    return pos_x, pos_y

# 获取我的位置
def get_my_pos(self):
    for i in range(self.map_h):
        for j in range(self.map_w):
            if self.map[i][j] == 1:
                return i, j
    return None

# 获取敌人位置
def get_enemy_pos(self):
    for i in range(self.map_h):
        for j in range(self.map_w):
            if self.map[i][j] == 2:
                return i, j
    return None

def step(self, action):
    my_pos_x, my_pos_y = self.get_my_pos()

game = Game() game.init_map(5) print(game.map) 五子棋github： https://github.com/junxiaosong/AlphaZero_Gomoku