视频讲解:
DQN 玩 2048 实战|第一期!搭建游戏环境(附 PyGame 可视化源码)
代码仓库:GitHub - LitchiCheng/DRL-learning: 深度强化学习
2048游戏介绍,引用维基百科
《2048》在4×4的网格上进行。玩家可以使用上、下、左、右四个方向键移动所有方块。[1]但在部分情形下,不可向某些方向移动。[2]:66游戏开始时,网格上会出现两个数值为2或4的方块。每次移动后,另一个数值为2或4的新方块会随机出现在空方格上。[2]:66方块会沿着指定的方向滑动,直到被其它方块或网格边缘阻挡。如果两个相同数值的方块碰撞,它们将合并成一个方块,其数值等于两个方块的数值之和。[3][4]如果三个数值相同的方块碰撞,则只会合并靠近终点方向的两个方块,距起点最近的方块的数值不变。若一行或一列中的方块数值均相同,则沿着该行或该列滑动会合并前两个和后两个方块。[5]在同一移动过程中,新生成的方块不能再与其他方块合并。[6]数值较高的方块会发出柔和的光芒;但随着得分增加,光芒会不断变暗。[6]方块数值都是2的幂,最大为131072。[7]界面右上方的记分牌会记录玩家的分数。玩家的初始分数为零,每当两个方块合并时,分数会增加,得分取决于合并后方块的数值。[8]
玩家获取胜利的条件是获得数值为2048的方块。达到这一目标后,玩家可以继续游戏,以获得更高的分数。[9][10][11]当玩家没有合法的移动方法时,即出现了网格上没有空方格,且相邻方块的数值均不相同的情况,游戏就直接结束。[6][12]
使用pygame创建一个交互环境,实际作为env使用的时候可以不用pygame,换成matplotlib只做一个简单的显示即可,安装pygame
pip3 install pygame -i https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
创建一个Env2048Gui.py
import numpy as np
import random
import pygame# 2048 游戏环境类
class Game2048:def __init__(self):self.board = np.zeros((4, 4), dtype=int)self.add_random_tile()self.add_random_tile()def add_random_tile(self):empty_cells = np.argwhere(self.board == 0)if len(empty_cells) > 0:index = random.choice(empty_cells)self.board[index[0], index[1]] = 2 if random.random() < 0.9 else 4def move_left(self):reward = 0new_board = np.copy(self.board)for row in range(4):line = new_board[row]non_zero = line[line != 0]merged = []i = 0while i < len(non_zero):if i + 1 < len(non_zero) and non_zero[i] == non_zero[i + 1]:merged.append(2 * non_zero[i])reward += 2 * non_zero[i]i += 2else:merged.append(non_zero[i])i += 1new_board[row] = np.pad(merged, (0, 4 - len(merged)), 'constant')if not np.array_equal(new_board, self.board):self.board = new_boardself.add_random_tile()return rewarddef move_right(self):self.board = np.fliplr(self.board)reward = self.move_left()self.board = np.fliplr(self.board)return rewarddef move_up(self):self.board = self.board.Treward = self.move_left()self.board = self.board.Treturn rewarddef move_down(self):self.board = self.board.Treward = self.move_right()self.board = self.board.Treturn rewarddef step(self, action):if action == 0:reward = self.move_left()elif action == 1:reward = self.move_right()elif action == 2:reward = self.move_up()elif action == 3:reward = self.move_down()done = not np.any(self.board == 0) and all([np.all(self.board[:, i] != self.board[:, i + 1]) for i in range(3)]) and all([np.all(self.board[i, :] != self.board[i + 1, :]) for i in range(3)])state = self.board.flatten()return state, reward, donedef reset(self):self.board = np.zeros((4, 4), dtype=int)self.add_random_tile()self.add_random_tile()return self.board.flatten()# 颜色定义
COLORS = {0: (205, 193, 180),2: (238, 228, 218),4: (237, 224, 200),8: (242, 177, 121),16: (245, 149, 99),32: (246, 124, 95),64: (246, 94, 59),128: (237, 207, 114),256: (237, 204, 97),512: (237, 200, 80),1024: (237, 197, 63),2048: (237, 194, 46)
}# 绘制游戏板
def draw_board(screen, board, tile_size, margin):for i in range(4):for j in range(4):value = board[i][j]color = COLORS.get(value, (0, 0, 0))pygame.draw.rect(screen, color,(j * (tile_size + margin) + margin,i * (tile_size + margin) + margin,tile_size, tile_size))if value != 0:font = pygame.font.Font(None, 36)text = font.render(str(value), True, (0, 0, 0))text_rect = text.get_rect(center=(j * (tile_size + margin) + margin + tile_size // 2,i * (tile_size + margin) + margin + tile_size // 2))screen.blit(text, text_rect)# 主函数
def main():pygame.init()tile_size = 100margin = 10width = height = 4 * (tile_size + margin) + marginscreen = pygame.display.set_mode((width, height))pygame.display.set_caption("2048 Game")game = Game2048()done = Falseclock = pygame.time.Clock()while not done:for event in pygame.event.get():if event.type == pygame.QUIT:done = Trueelif event.type == pygame.KEYDOWN:if event.key == pygame.K_LEFT:_, _, done = game.step(0)elif event.key == pygame.K_RIGHT:_, _, done = game.step(1)elif event.key == pygame.K_UP:_, _, done = game.step(2)elif event.key == pygame.K_DOWN:_, _, done = game.step(3)screen.fill((187, 173, 160))draw_board(screen, game.board, tile_size, margin)pygame.display.flip()clock.tick(60)pygame.quit()if __name__ == "__main__":main()
运行后,可以通过上下左右键进行控制
python3 Env2048Gui.py