Deep Reinforcement Learning (DRL) 算法在 PyTorch 中的实现与应用

深度强化学习算法在PyTorch中的实现与应用

近年来,深度强化学习(Deep Reinforcement Learning, DRL)在人工智能领域取得了巨大的进展,成为了解决复杂决策问题的有力工具。本文将详细介绍多种流行的DRL算法在PyTorch中的实现,包括Q-learning、DQN(Deep Q-Network)、PPO(Proximal Policy Optimization)、DDPG(Deep Deterministic Policy Gradient)、TD3(Twin Delayed Deep Deterministic Policy Gradient)和SAC(Soft Actor-Critic)等。我们将探讨这些算法的原理、优缺点以及在实际问题中的应用。

DRL算法概述

深度强化学习结合了深度学习和强化学习的优点,能够在复杂的环境中学习最优策略。以下是一些主要的DRL算法:

1. Q-learning: 经典的值迭代算法,用于离散动作空间。
2. DQN: 将深度神经网络与Q-learning结合,能够处理高维状态空间。
3. PPO: 基于策略梯度的算法,通过限制新旧策略的差异来稳定训练。
4. DDPG: 适用于连续动作空间的确定性策略梯度算法。
5. TD3: DDPG的改进版本,通过多个技巧提高了性能和稳定性。
6. SAC: 结合了最大熵强化学习的思想,在探索与利用之间取得良好平衡。

PyTorch实现

PyTorch作为一个灵活而强大的深度学习框架,非常适合实现各种DRL算法。以下是这些算法在PyTorch中实现的关键点:

Q-learning

Q-learning是最基础的强化学习算法之一,它通过迭代更新Q值表来学习最优策略。虽然Q-learning本身不需要深度学习,但了解它有助于理解更复杂的DRL算法。

import torch

class QLearning:
    def __init__(self, state_dim, action_dim, learning_rate=0.1, gamma=0.99):
        self.Q = torch.zeros((state_dim, action_dim))
        self.lr = learning_rate
        self.gamma = gamma

    def update(self, state, action, reward, next_state):
        target = reward + self.gamma * torch.max(self.Q[next_state])
        self.Q[state, action] += self.lr * (target - self.Q[state, action])

    def get_action(self, state):
        return torch.argmax(self.Q[state]).item()

DQN (Deep Q-Network)

DQN通过使用深度神经网络来近似Q函数,极大地扩展了Q-learning处理复杂环境的能力。

import torch
import torch.nn as nn

class DQN(nn.Module):
    def __init__(self, state_dim, action_dim):
        super(DQN, self).__init__()
        self.fc1 = nn.Linear(state_dim, 64)
        self.fc2 = nn.Linear(64, 64)
        self.fc3 = nn.Linear(64, action_dim)

    def forward(self, x):
        x = torch.relu(self.fc1(x))
        x = torch.relu(self.fc2(x))
        return self.fc3(x)

class DQNAgent:
    def __init__(self, state_dim, action_dim, learning_rate=1e-3, gamma=0.99):
        self.q_network = DQN(state_dim, action_dim)
        self.target_network = DQN(state_dim, action_dim)
        self.target_network.load_state_dict(self.q_network.state_dict())
        self.optimizer = torch.optim.Adam(self.q_network.parameters(), lr=learning_rate)
        self.gamma = gamma

    def update(self, state, action, reward, next_state, done):
        state = torch.FloatTensor(state)
        next_state = torch.FloatTensor(next_state)
        q_values = self.q_network(state)
        next_q_values = self.target_network(next_state).detach()
        
        target = q_values.clone()
        target[action] = reward + (1 - done) * self.gamma * next_q_values.max()
        
        loss = nn.MSELoss()(q_values, target)
        self.optimizer.zero_grad()
        loss.backward()
        self.optimizer.step()

    def get_action(self, state):
        state = torch.FloatTensor(state)
        q_values = self.q_network(state)
        return torch.argmax(q_values).item()

PPO (Proximal Policy Optimization)

PPO是一种基于策略梯度的算法,通过引入信任区域约束来提高训练的稳定性。

import torch
import torch.nn as nn
import torch.optim as optim

class PPO(nn.Module):
    def __init__(self, state_dim, action_dim):
        super(PPO, self).__init__()
        self.actor = nn.Sequential(
            nn.Linear(state_dim, 64),
            nn.Tanh(),
            nn.Linear(64, 64),
            nn.Tanh(),
            nn.Linear(64, action_dim),
            nn.Softmax(dim=-1)
        )
        self.critic = nn.Sequential(
            nn.Linear(state_dim, 64),
            nn.Tanh(),
            nn.Linear(64, 64),
            nn.Tanh(),
            nn.Linear(64, 1)
        )
        
    def forward(self, state):
        return self.actor(state), self.critic(state)

class PPOAgent:
    def __init__(self, state_dim, action_dim, lr=3e-4, gamma=0.99, epsilon=0.2):
        self.ppo = PPO(state_dim, action_dim)
        self.optimizer = optim.Adam(self.ppo.parameters(), lr=lr)
        self.gamma = gamma
        self.epsilon = epsilon

    def update(self, states, actions, rewards, next_states, dones, old_probs):
        states = torch.FloatTensor(states)
        actions = torch.LongTensor(actions)
        rewards = torch.FloatTensor(rewards)
        next_states = torch.FloatTensor(next_states)
        old_probs = torch.FloatTensor(old_probs)

        for _ in range(10):  # 多次更新
            new_probs, state_values = self.ppo(states)
            new_probs = new_probs.gather(1, actions.unsqueeze(1)).squeeze(1)
            
            ratio = new_probs / old_probs
            surr1 = ratio * rewards
            surr2 = torch.clamp(ratio, 1-self.epsilon, 1+self.epsilon) * rewards
            actor_loss = -torch.min(surr1, surr2).mean()
            
            critic_loss = nn.MSELoss()(state_values, rewards)
            
            loss = actor_loss + 0.5 * critic_loss
            
            self.optimizer.zero_grad()
            loss.backward()
            self.optimizer.step()

    def get_action(self, state):
        state = torch.FloatTensor(state)
        probs, _ = self.ppo(state)
        return torch.multinomial(probs, 1).item()

DDPG (Deep Deterministic Policy Gradient)

DDPG是一种适用于连续动作空间的算法,它结合了DQN和确定性策略梯度的思想。

import torch
import torch.nn as nn
import torch.optim as optim

class Actor(nn.Module):
    def __init__(self, state_dim, action_dim, max_action):
        super(Actor, self).__init__()
        self.fc1 = nn.Linear(state_dim, 400)
        self.fc2 = nn.Linear(400, 300)
        self.fc3 = nn.Linear(300, action_dim)
        self.max_action = max_action
        
    def forward(self, state):
        a = torch.relu(self.fc1(state))
        a = torch.relu(self.fc2(a))
        return self.max_action * torch.tanh(self.fc3(a))

class Critic(nn.Module):
    def __init__(self, state_dim, action_dim):
        super(Critic, self).__init__()
        self.fc1 = nn.Linear(state_dim + action_dim, 400)
        self.fc2 = nn.Linear(400, 300)
        self.fc3 = nn.Linear(300, 1)
        
    def forward(self, state, action):
        q = torch.cat([state, action], 1)
        q = torch.relu(self.fc1(q))
        q = torch.relu(self.fc2(q))
        return self.fc3(q)

class DDPGAgent:
    def __init__(self, state_dim, action_dim, max_action, lr=1e-4, gamma=0.99, tau=0.001):
        self.actor = Actor(state_dim, action_dim, max_action)
        self.actor_target = Actor(state_dim, action_dim, max_action)
        self.actor_target.load_state_dict(self.actor.state_dict())
        self.actor_optimizer = optim.Adam(self.actor.parameters(), lr=lr)
        
        self.critic = Critic(state_dim, action_dim)
        self.critic_target = Critic(state_dim, action_dim)
        self.critic_target.load_state_dict(self.critic.state_dict())
        self.critic_optimizer = optim.Adam(self.critic.parameters(), lr=lr)
        
        self.gamma = gamma
        self.tau = tau

    def select_action(self, state):
        state = torch.FloatTensor(state.reshape(1, -1))
        return self.actor(state).cpu().data.numpy().flatten()

    def update(self, replay_buffer, batch_size=100):
        # 从经验回放中采样
        state, action, next_state, reward, done = replay_buffer.sample(batch_size)
        
        # 计算目标Q值
        target_Q = self.critic_target(next_state, self.actor_target(next_state))
        target_Q = reward + (1 - done) * self.gamma * target_Q.detach()
        
        # 更新Critic
        current_Q = self.critic(state, action)
        critic_loss = nn.MSELoss()(current_Q, target_Q)
        self.critic_optimizer.zero_grad()
        critic_loss.backward()
        self.critic_optimizer.step()
        
        # 更新Actor
        actor_loss = -self.critic(state, self.actor(state)).mean()
        self.actor_optimizer.zero_grad()
        actor_loss.backward()
        self.actor_optimizer.step()
        
        # 软更新目标网络
        for param, target_param in zip(self.critic.parameters(), self.critic_target.parameters()):
            target_param.data.copy_(self.tau * param.data + (1 - self.tau) * target_param.data)
        
        for param, target_param in zip(self.actor.parameters(), self.actor_target.parameters()):
            target_param.data.copy_(self.tau * param.data + (1 - self.tau) * target_param.data)

TD3 (Twin Delayed DDPG)

TD3是DDPG的改进版本,通过引入双Q学习、延迟策略更新和目标策略平滑等技巧来提高性能和稳定性。

import torch
import torch.nn as nn
import torch.optim as optim

class Actor(nn.Module):
    def __init__(self, state_dim, action_dim, max_action):
        super(Actor, self).__init__()
        self.fc1 = nn.Linear(state_dim, 400)
        self.fc2 = nn.Linear(400, 300)
        self.fc3 = nn.Linear(300, action_dim)
        self.max_action = max_action
        
    def forward(self, state):
        a = torch.relu(self.fc1(state))
        a = torch.relu(self.fc2(a))
        return self.max_action * torch.tanh(self.fc3(a))

class Critic(nn.Module):
    def __init__(self, state_dim, action_dim):
        super(Critic, self).__init__()
        self.fc1 = nn.Linear(state_dim + action_dim, 400)
        self.fc2 = nn.Linear(400, 300)
        self.fc3 = nn.Linear(300, 1)
        
    def forward(self, state, action):
        sa = torch.cat([state, action], 1)
        q = torch.relu(self.fc1(sa))
        q = torch.relu(self.fc2(q))
        return self.fc3(q)

class TD3Agent:
    def __init__(self, state_dim, action_dim, max_action, lr=1e-3, gamma=0.99, tau=0.005, policy_noise=0.2, noise_clip=0.5, policy_freq=2):
        self.actor = Actor(state_dim, action_dim, max_action)
        self.actor_target = Actor(state_dim, action_dim, max_action)
        self.actor_target.load_state_dict(self.actor.state_dict())
        self.actor_optimizer = optim.Adam(self.actor.parameters(), lr=lr)
        
        self.critic1 = Critic(state_dim, action_dim)
        self.critic2 = Critic(state_dim, action_dim)
        self.critic1_target = Critic(state_dim, action_dim)
        self.critic2_target = Critic(state_dim, action_dim)
        self.critic1_target.load_state_dict(self.critic1.state_dict())
        self.critic2_target.load_state_dict(self.critic2.state_dict())
        self.critic1_optimizer = optim.Adam(self.critic1.parameters(), lr=lr)
        self.critic2_optimizer = optim.Adam(self.critic2.parameters(), lr=lr)
        
        self.max_action = max_action
        self.gamma = gamma
        self.tau = tau
        self.policy_noise = policy_noise
        self.noise_clip = noise_clip
        self.policy_freq = policy_freq