基于DQN和TensorFlow的LunarLander实现（全代码）

使用深度Q网络（Deep Q-Network, DQN）来训练一个在openai-gym的LunarLander-v2环境中的强化学习agent，让小火箭成功着陆。
下面代码直接扔到jupyter notebook或CoLab上就能跑起来。

在这里插入图片描述

安装和导入所需的库和环境

安装和设置所需的库和环境，使其能够在Jupyter Notebook中运行。

!pip install gym
!apt-get install xvfb -y
!pip install pyvirtualdisplay   #用于在没有显示器的环境中创建虚拟显示
!pip install Pillow             #一个图像处理库
!pip install swig
!pip install "gym[box2d]"

创建并启动一个虚拟显示，在没有图形界面的服务器上运行强化学习环境：

from pyvirtualdisplay import Display
display = Display(visible=0, size=(1400, 900))
display.start()

引入所需库：

import gym
import time
import tqdm
import numpy as np
from IPython import display as ipydisplay
from PIL import Image

创建一个LunarLander-v2环境的DQN代理：

agent = DQNAgent('LunarLander-v2')total_score, records = agent.simulate(visualize=True)
print(f'Total score {total_score:.2f}')
record_list = []
for i in tqdm.tqdm(range(100)):total_score, _ = agent.simulate(visualize=False)record_list.append(total_score)print(f'Average score in 100 episode {np.mean(record_list):.2f}')

Q网络搭建

import tensorflow as tfL = tf.keras.layersdef create_network_model(input_shape: np.ndarray,action_space: np.ndarray,learning_rate=0.001) -> tf.keras.Sequential:model = tf.keras.Sequential([L.Dense(512, input_shape=input_shape, activation="relu"),L.Dense(256, input_shape=input_shape, activation="relu"),L.Dense(action_space)])model.compile(loss="mse",optimizer=tf.optimizers.Adam(lr=learning_rate))return model

经验回放实现

经验回放是一种在深度强化学习中常用的技术，主要用于打破数据的相关性和减少过拟合。
在强化学习中，代理通常会在训练过程中与环境进行大量交互，经验回放允许代理存储这些经验，并在后续的训练中反复利用这些数据。这种机制有助于改善学习效率，减少数据样本间的时间相关性，提高训练过程的稳定性。

import random
import numpy as np
from collections import namedtuple# 代表每一个样本的 namedtuple，方便存储和读取数据
Experience = namedtuple('Experience', ('state', 'action', 'reward', 'next_state', 'done'))class ReplayMemory:def __init__(self, max_size):self.max_size = max_sizeself.memory = []def append(self, state, action, reward, next_state, done):"""记录一个新的样本"""sample = Experience(state, action, reward, next_state, done)self.memory.append(sample)# 只留下最新记录的 self.max_size 个样本self.memory = self.memory[-self.max_size:]def sample(self, batch_size):"""按照给定批次大小取样"""samples = random.sample(self.memory, batch_size)batch = Experience(*zip(*samples))# 转换数据为 numpy 张量返回states = np.array(batch.state)actions = np.array(batch.action)rewards = np.array(batch.reward)states_next = np.array(batch.next_state)dones = np.array(batch.done)return states, actions, rewards, states_next, donesdef __len__(self):return len(self.memory)

DQNAgent实现

DQNAgent类是DQN算法的核心实现。它包含以下关键部分：
1、初始化：初始化环境、神经网络模型和经验回放缓存。
2、行为选择（choose_action）：根据当前状态和ε-greedy策略选择行为。
3、经验回放（replay）：从记忆中随机抽取小批量经验进行学习。
4、训练（train）：进行多个episode的训练。

from IPython import display
from PIL import Image# 定义超参数
LEARNING_RATE = 0.001
GAMMA = 0.99
EPSILON_DECAY = 0.995
EPSILON_MIN = 0.01class DQNAgent:def __init__(self, env_name):self.env = gym.make(env_name)self.observation_shape = self.env.observation_space.shapeself.action_count = self.env.action_space.nself.model = create_network_model(self.observation_shape, self.action_count)self.memory = ReplayMemory(500000)self.epsilon = 1.0self.batch_size = 64def choose_action(self, state, epsilon=None):"""根据给定状态选择行为- epsilon == 0 完全使用模型选择行为- epsilon == 1 完全随机选择行为"""if epsilon is None:epsilon = self.epsilonif np.random.rand() < epsilon:return np.random.randint(self.action_count)else:q_values = self.model.predict(np.expand_dims(state, axis=0))return np.argmax(q_values[0])def replay(self):"""进行经验回放学习"""# 如果当前经验池经验数量少于批次大小，则跳过if len(self.memory) < self.batch_size:returnstates, actions, rewards, states_next, dones = self.memory.sample(self.batch_size)q_pred = self.model.predict(states)q_next = self.model.predict(states_next).max(axis=1)q_next = q_next * (1 - dones)q_update = rewards + GAMMA * q_nextindices = np.arange(self.batch_size)q_pred[[indices], [actions]] = q_updateself.model.train_on_batch(states, q_pred)def simulate(self, epsilon=None, visualize=True):records = []state = self.env.reset()is_done = Falsetotal_score = 0total_step  = 0while not is_done:action = self.choose_action(state, epsilon)state, reward, is_done, info = self.env.step(action)total_score += rewardtotal_step += 1rgb_array = self.env.render(mode='rgb_array')records.append((rgb_array, action, reward, total_score))if visualize:display.clear_output(wait=True)img = Image.fromarray(rgb_array)# 当前 Cell 中展示图片display.display(img)print(f'Action {action} Action reward {reward:.2f} | Total score {total_score:.2f} | Step {total_step}')time.sleep(0.01)self.env.close()return total_score, recordsdef train(self, episode_count: int, log_dir: str):"""训练方法，按照给定 episode 数量进行训练，并记录训练过程关键参数到 TensorBoard"""# 初始化一个 TensorBoard 记录器file_writer = tf.summary.create_file_writer(log_dir)file_writer.set_as_default()score_list = []best_avg_score = -np.inffor episode_index in range(episode_count):state = self.env.reset()score, step = 0, 0is_done = Falsewhile not is_done:# 根据状态选择一个行为action = self.choose_action(state)# 执行行为，记录行为和结果到经验池state_next, reward, is_done, info = self.env.step(action)self.memory.append(state, action, reward, state_next, is_done)score += rewardstate = state_next# 每 6 步进行一次回放训练# 此处也可以选择每一步回放训练，但会降低训练速度，这个是一个经验技巧if step % 1 == 0:self.replay()step += 1# 记录当前 Episode 的得分，计算最后 100 Episode 的平均得分score_list.append(score)avg_score = np.mean(score_list[-100:])# 记录当前 Episode 得分，epsilon 和最后 100 Episode 的平均得分到 TensorBoardtf.summary.scalar('score', data=score, step=episode_index)tf.summary.scalar('average score', data=avg_score, step=episode_index)tf.summary.scalar('epsilon', data=self.epsilon, step=episode_index)# 终端输出训练进度print(f'Episode: {episode_index} Reward: {score:03.2f} 'f'Average Reward: {avg_score:03.2f} Epsilon: {self.epsilon:.3f}')# 调整 epsilon 值，逐渐减少随机探索比例if self.epsilon > EPSILON_MIN:self.epsilon *= EPSILON_DECAY# 如果当前平均得分比之前有改善，保存模型# 确保提前创建目录 outputs/chapter_15if avg_score > best_avg_score:best_avg_score = avg_scoreself.model.save(f'outputs/chapter_15/dqn_best_{episode_index}.h5')

训练

# 使用 LunarLander 初始化 Agent
agent = DQNAgent('LunarLander-v2')
import glob
# 读取现在已经记录的日志数量，避免日志重复记录
tf_log_index = len(glob.glob('tf_dir/lunar_lander/run_*'))
log_dir = f'tf_dir/lunar_lander/run_{tf_log_index}'# 训练 2000 个 Episode
agent.train(20, log_dir)agent.model.summary()