TensorFlow5: 训练一个简单的游戏AI（Deep Q Network）

Deep Q Network是DeepMind最早(2013年)提出来的，是深度强化学习方法。

最开始AI什么也不会，通过给它提供游戏界面像素和分数，慢慢把它训练成游戏高手。

Github上有不少DQN实现，在本帖中，我使用TensorFlow训练一个简单的游戏AI。

1. 使用pygame写一个简单的小游戏
2. 使用强化学习训练游戏AI

pygame小游戏

import pygamefrom pygame.locals import *import sys BLACK     = (0  ,0  ,0  )WHITE     = (255,255,255) SCREEN_SIZE = [320,400]BAR_SIZE = [20, 5]BALL_SIZE = [15, 15] class Game(object):def __init__(self):pygame.init()self.clock = pygame.time.Clock()self.screen = pygame.display.set_mode(SCREEN_SIZE)pygame.display.set_caption('Simple Game') self.ball_pos_x = SCREEN_SIZE[0]//2 - BALL_SIZE[0]/2self.ball_pos_y = SCREEN_SIZE[1]//2 - BALL_SIZE[1]/2# ball移动方向self.ball_dir_x = -1 # -1 = left 1 = right  self.ball_dir_y = -1 # -1 = up   1 = downself.ball_pos = pygame.Rect(self.ball_pos_x, self.ball_pos_y, BALL_SIZE[0], BALL_SIZE[1]) self.score = 0self.bar_pos_x = SCREEN_SIZE[0]//2-BAR_SIZE[0]//2self.bar_pos = pygame.Rect(self.bar_pos_x, SCREEN_SIZE[1]-BAR_SIZE[1], BAR_SIZE[0], BAR_SIZE[1]) def bar_move_left(self):self.bar_pos_x = self.bar_pos_x - 2def bar_move_right(self):self.bar_pos_x = self.bar_pos_x + 2 def run(self):pygame.mouse.set_visible(0) # make cursor invisible bar_move_left = Falsebar_move_right = Falsewhile True:for event in pygame.event.get():if event.type == QUIT:pygame.quit()sys.exit()elif event.type == pygame.MOUSEBUTTONDOWN and event.button == 1:  # 鼠标左键按下(左移)bar_move_left = Trueelif event.type == pygame.MOUSEBUTTONUP and event.button == 1: # 鼠标左键释放bar_move_left = Falseelif event.type == pygame.MOUSEBUTTONDOWN and event.button == 3: #右键bar_move_right = Trueelif event.type == pygame.MOUSEBUTTONUP and event.button == 3:bar_move_right = False if bar_move_left == True and bar_move_right == False:self.bar_move_left()if bar_move_left == False and bar_move_right == True:self.bar_move_right() self.screen.fill(BLACK)self.bar_pos.left = self.bar_pos_xpygame.draw.rect(self.screen, WHITE, self.bar_pos) self.ball_pos.left += self.ball_dir_x * 2self.ball_pos.bottom += self.ball_dir_y * 3pygame.draw.rect(self.screen, WHITE, self.ball_pos) if self.ball_pos.top <= 0 or self.ball_pos.bottom >= (SCREEN_SIZE[1] - BAR_SIZE[1]+1):self.ball_dir_y = self.ball_dir_y * -1if self.ball_pos.left <= 0 or self.ball_pos.right >= (SCREEN_SIZE[0]):self.ball_dir_x = self.ball_dir_x * -1  if self.bar_pos.top <= self.ball_pos.bottom and (self.bar_pos.left < self.ball_pos.right and self.bar_pos.right > self.ball_pos.left):self.score += 1print("Score: ", self.score, end='\r')elif self.bar_pos.top <= self.ball_pos.bottom and (self.bar_pos.left > self.ball_pos.right or self.bar_pos.right < self.ball_pos.left):print("Game Over: ", self.score)return self.score pygame.display.update()self.clock.tick(60) game = Game()game.run()

自制的垃圾游戏。

操作：按住鼠标左键左移棒子，按住鼠标右键右移棒子。每次接住小方块得一分。

把棒子调短，提高游戏难度，看看训练出来的游戏AI有多强

ps. 其实我想做一个俄罗斯方块，呵呵呵呵，留着以后再做。

基于强化学习的AI（TensorFlow）

import pygameimport randomfrom pygame.locals import *import numpy as npfrom collections import dequeimport tensorflow as tf  # http://blog.topspeedsnail.com/archives/10116import cv2               # http://blog.topspeedsnail.com/archives/4755 BLACK     = (0  ,0  ,0  )WHITE     = (255,255,255) SCREEN_SIZE = [320,400]BAR_SIZE = [50, 5]BALL_SIZE = [15, 15] # 神经网络的输出MOVE_STAY = [1, 0, 0]MOVE_LEFT = [0, 1, 0]MOVE_RIGHT = [0, 0, 1] class Game(object):def __init__(self):pygame.init()self.clock = pygame.time.Clock()self.screen = pygame.display.set_mode(SCREEN_SIZE)pygame.display.set_caption('Simple Game') self.ball_pos_x = SCREEN_SIZE[0]//2 - BALL_SIZE[0]/2self.ball_pos_y = SCREEN_SIZE[1]//2 - BALL_SIZE[1]/2 self.ball_dir_x = -1 # -1 = left 1 = right  self.ball_dir_y = -1 # -1 = up   1 = downself.ball_pos = pygame.Rect(self.ball_pos_x, self.ball_pos_y, BALL_SIZE[0], BALL_SIZE[1]) self.bar_pos_x = SCREEN_SIZE[0]//2-BAR_SIZE[0]//2self.bar_pos = pygame.Rect(self.bar_pos_x, SCREEN_SIZE[1]-BAR_SIZE[1], BAR_SIZE[0], BAR_SIZE[1]) # action是MOVE_STAY、MOVE_LEFT、MOVE_RIGHT# ai控制棒子左右移动；返回游戏界面像素数和对应的奖励。(像素->奖励->强化棒子往奖励高的方向移动)def step(self, action): if action == MOVE_LEFT:self.bar_pos_x = self.bar_pos_x - 2elif action == MOVE_RIGHT:self.bar_pos_x = self.bar_pos_x + 2else:passif self.bar_pos_x < 0:self.bar_pos_x = 0if self.bar_pos_x > SCREEN_SIZE[0] - BAR_SIZE[0]:self.bar_pos_x = SCREEN_SIZE[0] - BAR_SIZE[0]self.screen.fill(BLACK)self.bar_pos.left = self.bar_pos_xpygame.draw.rect(self.screen, WHITE, self.bar_pos) self.ball_pos.left += self.ball_dir_x * 2self.ball_pos.bottom += self.ball_dir_y * 3pygame.draw.rect(self.screen, WHITE, self.ball_pos) if self.ball_pos.top <= 0 or self.ball_pos.bottom >= (SCREEN_SIZE[1] - BAR_SIZE[1]+1):self.ball_dir_y = self.ball_dir_y * -1if self.ball_pos.left <= 0 or self.ball_pos.right >= (SCREEN_SIZE[0]):self.ball_dir_x = self.ball_dir_x * -1 reward = 0if self.bar_pos.top <= self.ball_pos.bottom and (self.bar_pos.left < self.ball_pos.right and self.bar_pos.right > self.ball_pos.left):reward = 1    # 击中奖励elif self.bar_pos.top <= self.ball_pos.bottom and (self.bar_pos.left > self.ball_pos.right or self.bar_pos.right < self.ball_pos.left):reward = -1   # 没击中惩罚 # 获得游戏界面像素screen_image = pygame.surfarray.array3d(pygame.display.get_surface())pygame.display.update()# 返回游戏界面像素和对应的奖励return reward, screen_image # learning_rateLEARNING_RATE = 0.99# 更新梯度INITIAL_EPSILON = 1.0FINAL_EPSILON = 0.05# 测试观测次数EXPLORE = 500000 OBSERVE = 50000# 存储过往经验大小REPLAY_MEMORY = 500000 BATCH = 100 output = 3  # 输出层神经元数。代表3种操作-MOVE_STAY:[1, 0, 0]  MOVE_LEFT:[0, 1, 0]  MOVE_RIGHT:[0, 0, 1]input_image = tf.placeholder("float", [None, 80, 100, 4])  # 游戏像素action = tf.placeholder("float", [None, output])     # 操作 # 定义CNN-卷积神经网络 参考:http://blog.topspeedsnail.com/archives/10451def convolutional_neural_network(input_image):weights = {'w_conv1':tf.Variable(tf.zeros([8, 8, 4, 32])),               'w_conv2':tf.Variable(tf.zeros([4, 4, 32, 64])),               'w_conv3':tf.Variable(tf.zeros([3, 3, 64, 64])),               'w_fc4':tf.Variable(tf.zeros([3456, 784])),               'w_out':tf.Variable(tf.zeros([784, output]))} biases = {'b_conv1':tf.Variable(tf.zeros([32])),              'b_conv2':tf.Variable(tf.zeros([64])),              'b_conv3':tf.Variable(tf.zeros([64])),              'b_fc4':tf.Variable(tf.zeros([784])),              'b_out':tf.Variable(tf.zeros([output]))} conv1 = tf.nn.relu(tf.nn.conv2d(input_image, weights['w_conv1'], strides = [1, 4, 4, 1], padding = "VALID") + biases['b_conv1'])conv2 = tf.nn.relu(tf.nn.conv2d(conv1, weights['w_conv2'], strides = [1, 2, 2, 1], padding = "VALID") + biases['b_conv2'])conv3 = tf.nn.relu(tf.nn.conv2d(conv2, weights['w_conv3'], strides = [1, 1, 1, 1], padding = "VALID") + biases['b_conv3'])conv3_flat = tf.reshape(conv3, [-1, 3456])fc4 = tf.nn.relu(tf.matmul(conv3_flat, weights['w_fc4']) + biases['b_fc4']) output_layer = tf.matmul(fc4, weights['w_out']) + biases['b_out']return output_layer # 深度强化学习入门: https://www.nervanasys.com/demystifying-deep-reinforcement-learning/# 训练神经网络def train_neural_network(input_image):predict_action = convolutional_neural_network(input_image) argmax = tf.placeholder("float", [None, output])gt = tf.placeholder("float", [None]) action = tf.reduce_sum(tf.mul(predict_action, argmax), reduction_indices = 1)cost = tf.reduce_mean(tf.square(action - gt))optimizer = tf.train.AdamOptimizer(1e-6).minimize(cost) game = Game()D = deque() _, image = game.step(MOVE_STAY)# 转换为灰度值image = cv2.cvtColor(cv2.resize(image, (100, 80)), cv2.COLOR_BGR2GRAY)# 转换为二值ret, image = cv2.threshold(image, 1, 255, cv2.THRESH_BINARY)input_image_data = np.stack((image, image, image, image), axis = 2)with tf.Session() as sess:sess.run(tf.initialize_all_variables())saver = tf.train.Saver()n = 0epsilon = INITIAL_EPSILONwhile True:action_t = predict_action.eval(feed_dict = {input_image : [input_image_data]})[0] argmax_t = np.zeros([output], dtype=np.int)if(random.random() <= INITIAL_EPSILON):maxIndex = random.randrange(output)else:maxIndex = np.argmax(action_t)argmax_t[maxIndex] = 1if epsilon > FINAL_EPSILON:epsilon -= (INITIAL_EPSILON - FINAL_EPSILON) / EXPLORE #for event in pygame.event.get():  macOS需要事件循环，否则白屏#if event.type == QUIT:#pygame.quit()#sys.exit()reward, image = game.step(list(argmax_t)) image = cv2.cvtColor(cv2.resize(image, (100, 80)), cv2.COLOR_BGR2GRAY)ret, image = cv2.threshold(image, 1, 255, cv2.THRESH_BINARY)image = np.reshape(image, (80, 100, 1))input_image_data1 = np.append(image, input_image_data[:, :, 0:3], axis = 2) D.append((input_image_data, argmax_t, reward, input_image_data1)) if len(D) > REPLAY_MEMORY:D.popleft() if n > OBSERVE:minibatch = random.sample(D, BATCH)input_image_data_batch = [d[0] for d in minibatch]argmax_batch = [d[1] for d in minibatch]reward_batch = [d[2] for d in minibatch]input_image_data1_batch = [d[3] for d in minibatch] gt_batch = [] out_batch = predict_action.eval(feed_dict = {input_image : input_image_data1_batch}) for i in range(0, len(minibatch)):gt_batch.append(reward_batch[i] + LEARNING_RATE * np.max(out_batch[i])) optimizer.run(feed_dict = {gt : gt_batch, argmax : argmax_batch, input_image : input_image_data_batch}) input_image_data = input_image_data1n = n+1 if n % 10000 == 0:saver.save(sess, 'game.cpk', global_step = n)  # 保存模型 print(n, "epsilon:", epsilon, " " ,"action:", maxIndex, " " ,"reward:", reward)  train_neural_network(input_image)

训练中：

如果你使用Linux，你可以使用htop监控内存使用情况。

刚开始，AI傻傻的，只会控制棒子来回瞎晃，通过try-error，它会慢慢掌握这个游戏。等我一觉醒来，这货已经玩的不亦乐乎了。

ps.准备换一个顶级显卡，CPU玩tensorflow太费劲，看来非游戏玩家也有必要买好显卡。

使用训练出来AI玩游戏

这步要做的就是加载使用前面保存的模型。

上面是自己手动实现的强化学习算法，其实有一个特别好的专门为开发测试AI而设计的库openai gym。OpenAI Gym是一个为比较、构建强化学习Ai的一个Python库，它包含很多测试游戏。

参考：https://www.nervanasys.com/openai/

• OpenAI文档：https://gym.openai.com/docs
• OpenAI源代码：https://github.com/openai/gym

强化学习模型

安装Gym

$ git clone https://github.com/openai/gym$ cd gym# 安装依赖#$ brew install cmake boost boost-python sdl2 swig wget  # macOS python2# brew install boost-python --with-python3 # python3#$ sudo apt-get install -y python-numpy python-dev cmake zlib1g-dev libjpeg-dev xvfb libav-tools xorg-dev python-opengl libboost-all-dev libsdl2-dev swig  # Ubuntu$ pip install gym[all]

• DeepLearningFlappyBird

Share the post "TensorFlow练习5: 训练一个简单的游戏AI（Deep Q Network）"