最近再改一个代码，需要改成PPO方式的，由于之前没有接触过此类算法，因此进行了简单学习，论文没有看的很详细，重点看了实现部分，这里只做简单记录。
  这里附上论文链接，需要的可以详细看一下。
   Proximal Policy Optimization Algorithms.

一、PPO算法

  PPO算法本质上是一个On-Policy的算法，它可以对采样到的样本进行多次利用，在一定程度上解决样本利用率低的问题，收到较好的效果。论文里有两种实现方式，一种是结合KL的penalty的，另一种是clip裁断的方法。大部分都是采用的后者，本文记录的也主要是后者的实现。

二、伪代码

  在网上找了一下伪代码，大概两类，前者是Open AI的，比较精炼，后者是Deepmind的，写的比较详细，在这里同时附上.

三、相关的简单理论

1.ratio

  这里的比例ratio，是两种策略下动作的概率比，而在程序实现中，用的是对动作分布取对数，而后使用e指数相减的方法，具体实现如下所示：

action_logprobs = dist.log_prob(action)
ratios = torch.exp(logprobs - old_logprobs.detach())

2.裁断

  其中，裁断对应的部分如下图所示：

  上述公式代表的含义如下：
  clip公式含义.

  这里我是这样理解的：
  (1)如果A>0,说明现阶段的(st,at)相对较好，那么我们希望该二元组出现的概率越高越好，即ratio中的分子越大越好，但是分母分子不能差太多，因此需要加一个上限;
  (2)如果A<0,说明现阶段的(st,at)相对较差，那么我们希望该二元组出现的概率越低越好，即ratio中的分子越小越好，但是分母分子不能差太多，因此需要加一个下限.

3.Advantage的计算

  论文里计算At的方式如下，在一些情况下可以令lamda为1；还有一种更常用的计算方式是VAE，这里不进行描述.。

  对应的代码块如下：

 def update(self, memory):# Monte Carlo estimate of rewards:rewards = []discounted_reward = 0for reward, is_terminal in zip(reversed(memory.rewards), reversed(memory.is_terminals)):if is_terminal:discounted_reward = 0discounted_reward = reward + (self.gamma * discounted_reward)rewards.insert(0, discounted_reward)

4.loss的计算

  这里的第一项，对应裁断项,需要计算ratio和Advantage，之后进行裁断；
  这里的第二项，对应的为对应的值的均方误差;
  这里的第三项，为交叉熵
  程序的实现如下所示：

surr1 = ratios * advantages
surr2 = torch.clamp(ratios, 1 - self.eps_clip, 1 + self.eps_clip) * advantages
loss = -torch.min(surr1, surr2) + 0.5 * self.MseLoss(state_values, rewards) - 0.01 * dist_entropy

四、算法实现

  这里算法的实现参考了一位博主
  PPO代码.

#!/usr/bin/python3
# -*-coding:utf-8 -*-# @Time    : 2022/6/18 15:53
# @Author  : Wang xiangyu
# @File    : PPO.py
import torch
import torch.nn as nn
from torch.distributions import MultivariateNormal
import gym
import numpy as npdevice = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")class Memory:def __init__(self):self.actions = []self.states = []self.logprobs = []self.rewards = []self.is_terminals = []def clear_memory(self):# del语句作用在变量上，而不是数据对象上。删除的是变量，而不是数据。del self.actions[:]del self.states[:]del self.logprobs[:]del self.rewards[:]del self.is_terminals[:]class ActorCritic(nn.Module):def __init__(self, state_dim, action_dim, action_std):super(ActorCritic, self).__init__()# action mean range -1 to 1self.actor = nn.Sequential(nn.Linear(state_dim, 64),nn.Tanh(),nn.Linear(64, 32),nn.Tanh(),nn.Linear(32, action_dim),nn.Tanh())# criticself.critic = nn.Sequential(nn.Linear(state_dim, 64),nn.Tanh(),nn.Linear(64, 32),nn.Tanh(),nn.Linear(32, 1))# 方差self.action_var = torch.full((action_dim,), action_std * action_std).to(device)def forward(self):# 手动设置异常raise NotImplementedErrordef act(self, state, memory):action_mean = self.actor(state)cov_mat = torch.diag(self.action_var).to(device)dist = MultivariateNormal(action_mean, cov_mat)action = dist.sample()action_logprob = dist.log_prob(action)memory.states.append(state)memory.actions.append(action)memory.logprobs.append(action_logprob)return action.detach()def evaluate(self, state, action):action_mean = self.actor(state)action_var = self.action_var.expand_as(action_mean)# torch.diag_embed(input, offset=0, dim1=-2, dim2=-1) → Tensor# Creates a tensor whose diagonals of certain 2D planes (specified by dim1 and dim2) are filled by inputcov_mat = torch.diag_embed(action_var).to(device)# 生成一个多元高斯分布矩阵dist = MultivariateNormal(action_mean, cov_mat)# 我们的目的是要用这个随机的去逼近真正的选择动作action的高斯分布action_logprobs = dist.log_prob(action)# log_prob 是action在前面那个正太分布的概率的log ，我们相信action是对的 ，# 那么我们要求的正态分布曲线中点应该在action这里，所以最大化正太分布的概率的log， 改变mu,sigma得出一条中心点更加在a的正太分布。dist_entropy = dist.entropy()state_value = self.critic(state)return action_logprobs, torch.squeeze(state_value), dist_entropyclass PPO:def __init__(self, state_dim, action_dim, action_std, lr, betas, gamma, K_epochs, eps_clip):self.lr = lrself.betas = betasself.gamma = gammaself.eps_clip = eps_clipself.K_epochs = K_epochsself.policy = ActorCritic(state_dim, action_dim, action_std).to(device)self.optimizer = torch.optim.Adam(self.policy.parameters(), lr=lr, betas=betas)self.policy_old = ActorCritic(state_dim, action_dim, action_std).to(device)self.policy_old.load_state_dict(self.policy.state_dict())self.MseLoss = nn.MSELoss()def select_action(self, state, memory):state = torch.FloatTensor(state.reshape(1, -1)).to(device)return self.policy_old.act(state, memory).cpu().data.numpy().flatten()def update(self, memory):# Monte Carlo estimate of rewards:rewards = []discounted_reward = 0for reward, is_terminal in zip(reversed(memory.rewards), reversed(memory.is_terminals)):if is_terminal:discounted_reward = 0discounted_reward = reward + (self.gamma * discounted_reward)rewards.insert(0, discounted_reward)# Normalizing the rewards:rewards = torch.tensor(rewards, dtype=torch.float32).to(device)rewards = (rewards - rewards.mean()) / (rewards.std() + 1e-5)# convert list to tensor# 使用stack可以保留两个信息：[1. 序列] 和 [2. 张量矩阵] 信息，属于【扩张再拼接】的函数；old_states = torch.squeeze(torch.stack(memory.states).to(device), 1).detach()old_actions = torch.squeeze(torch.stack(memory.actions).to(device), 1).detach()old_logprobs = torch.squeeze(torch.stack(memory.logprobs), 1).to(device).detach()#这里即可以对样本进行多次利用，提高利用率# Optimize policy for K epochs:for _ in range(self.K_epochs):# Evaluating old actions and values :logprobs, state_values, dist_entropy = self.policy.evaluate(old_states, old_actions)# Finding the ratio (pi_theta / pi_theta__old):ratios = torch.exp(logprobs - old_logprobs.detach())# Finding Surrogate Loss:advantages = rewards - state_values.detach()surr1 = ratios * advantagessurr2 = torch.clamp(ratios, 1 - self.eps_clip, 1 + self.eps_clip) * advantagesloss = -torch.min(surr1, surr2) + 0.5 * self.MseLoss(state_values, rewards) - 0.01 * dist_entropy# take gradient stepself.optimizer.zero_grad()loss.mean().backward()self.optimizer.step()# Copy new weights into old policy:self.policy_old.load_state_dict(self.policy.state_dict())def main():############## Hyperparameters ##############env_name = "Pendulum-v1"render = Falsesolved_reward = 300  # stop training if avg_reward > solved_rewardlog_interval = 20  # print avg reward in the intervalmax_episodes = 10000  # max training episodesmax_timesteps = 1500  # max timesteps in one episodeupdate_timestep = 4000  # update policy every n timestepsaction_std = 0.5  # constant std for action distribution (Multivariate Normal)K_epochs = 80  # update policy for K epochseps_clip = 0.2  # clip parameter for PPOgamma = 0.99  # discount factorlr = 0.0003  # parameters for Adam optimizerbetas = (0.9, 0.999)############################################## creating environmentenv = gym.make(env_name)state_dim = env.observation_space.shape[0]action_dim = env.action_space.shape[0]memory = Memory()ppo = PPO(state_dim, action_dim, action_std, lr, betas, gamma, K_epochs, eps_clip)print(lr, betas)# logging variablesrunning_reward = 0avg_length = 0time_step = 0# training loopfor i_episode in range(1, max_episodes + 1):state = env.reset()for t in range(max_timesteps):time_step += 1# Running policy_old:action = ppo.select_action(state, memory)state, reward, done, _ = env.step(action)# Saving reward and is_terminals:memory.rewards.append(reward)memory.is_terminals.append(done)# update if its timeif time_step % update_timestep == 0:ppo.update(memory)memory.clear_memory()time_step = 0running_reward += rewardif render:env.render()if done:breakavg_length += t+1# stop training if avg_reward > solved_rewardif running_reward > (log_interval * solved_reward):print("########## Solved! ##########")torch.save(ppo.policy.state_dict(), './PPO_continuous_solved_{}.pth'.format(env_name))break# save every 500 episodesif i_episode % 500 == 0:torch.save(ppo.policy.state_dict(), './PPO_continuous_{}.pth'.format(env_name))# loggingif i_episode % log_interval == 0:avg_length = int(avg_length / log_interval)running_reward = int((running_reward / log_interval))print('Episode {} \t Avg length: {} \t Avg reward: {}'.format(i_episode, avg_length, running_reward))running_reward = 0avg_length = 0if __name__ == '__main__':main()

五、效果

  可以看到经过一段时间的训练，奖励有了一定升高.

六、感悟

  感悟是对改的项目的总结，和本文没有什么关系。
  这次改的项目参考了PPO的代码，架子基本也是搭好的，所以改起来也没有想象的那么困难。但应该是我第一次改代码，之前只是看代码，从来没有尝试改过那么多，可以感觉到看代码和改代码这两个能力间差的真的很多，写代码就更困难了emm，可以说经过这一次，可以更好的看到和别人的差距，不过对自己也有很大提高。在以后的学习中，还是需要多看多写，逐步提高。