[A3C]:Tensorflow代码实现详解

强化学习：A3C算法Tensorflow实现

最近在看A3C，理论知识很容易理解，代码还是有一定难度，先分享本人学习莫烦大佬A3C代码的注释，理论知识后补！！！

具体的算法伪代码如下：

tensorflow代码如下：

"""
Asynchronous Advantage Actor Critic (A3C) with continuous action space, Reinforcement Learning.

The Pendulum example.

View more on my tutorial page: https://morvanzhou.github.io/tutorials/

Using:
tensorflow 1.8.0
gym 0.10.5
"""

import multiprocessing  # 多线程模块
import threading  # 线程模块
import tensorflow as tf
import numpy as np
import gym
import os
import shutil  # 拷贝文件用
import matplotlib.pyplot as plt

GAME = 'Pendulum-v0'
OUTPUT_GRAPH = True
LOG_DIR = './log'
N_WORKERS = multiprocessing.cpu_count()  # 独立玩家个体数为cpu数
MAX_EP_STEP = 200
MAX_GLOBAL_EP = 2000  # 中央大脑最大回合数
GLOBAL_NET_SCOPE = 'Global_Net'  # 中央大脑的名字
UPDATE_GLOBAL_ITER = 10  # 中央大脑每N次更新一次
GAMMA = 0.9  # 衰减度
ENTROPY_BETA = 0.01  # β项熵
LR_A = 0.0001    # learning rate for actor
LR_C = 0.001    # learning rate for critic
GLOBAL_RUNNING_R = []  # 存储总的reward
GLOBAL_EP = 0  # 中央大脑步数

env = gym.make(GAME)  # 定义游戏环境

N_S = env.observation_space.shape[0]  # 观测值个数
N_A = env.action_space.shape[0]  # 动作值个数
A_BOUND = [env.action_space.low, env.action_space.high]  # 动作界限

# 这个 class 可以被调用生成一个 global net.
# 也能被调用生成一个 worker 的 net, 因为他们的结构是一样的,
# 所以这个 class 可以被重复利用.
class ACNet(object):
    def __init__(self, scope, globalAC=None):

        if scope == GLOBAL_NET_SCOPE:   # get global network
            with tf.variable_scope(scope):
                self.s = tf.placeholder(tf.float32, [None, N_S], 'S')  # [None, N_S]数据形状，None代表batch，N_S是每个state的观测值个数
                self.a_params, self.c_params = self._build_net(scope)[-2:]  # 定义中央大脑actor和critic的参数
        else:   # local net, calculate losses
            with tf.variable_scope(scope):
                self.s = tf.placeholder(tf.float32, [None, N_S], 'S')
                self.a_his = tf.placeholder(tf.float32, [None, N_A], 'A')
                self.v_target = tf.placeholder(tf.float32, [None, 1], 'Vtarget')

                mu, sigma, self.v, self.a_params, self.c_params = self._build_net(scope)  # 均值μ，方差σ，

                td = tf.subtract(self.v_target, self.v, name='TD_error')  # TD_error=v_target-v
                with tf.name_scope('c_loss'):
                    self.c_loss = tf.reduce_mean(tf.square(td))  # TD加平方避免负数

                with tf.name_scope('wrap_a_out'):
                    mu, sigma = mu * A_BOUND[1], sigma + 1e-4

                normal_dist = tf.distributions.Normal(mu, sigma)  # tf.distributions.normal可以生成一个均值为μ，方差为σ的正态分布。

                with tf.name_scope('a_loss'):
                    log_prob = normal_dist.log_prob(self.a_his)  # 正态分布中概率的log值
                    exp_v = log_prob * tf.stop_gradient(td)
                    entropy = normal_dist.entropy()  # 最大熵
                    self.exp_v = ENTROPY_BETA * entropy + exp_v  # 完整的目标函数
                    self.a_loss = tf.reduce_mean(-self.exp_v)

                with tf.name_scope('choose_a'):  # use local params to choose action
                    self.A = tf.clip_by_value(tf.squeeze(normal_dist.sample(1), axis=[0, 1]), A_BOUND[0], A_BOUND[1])
                    # tf.clip_by_value将正态分布输出值压缩在min~max之间得到action输出
                with tf.name_scope('local_grad'):
                    self.a_grads = tf.gradients(self.a_loss, self.a_params)
                    # 实现a_loss对a_params每一个参数的求导，返回一个list
                    self.c_grads = tf.gradients(self.c_loss, self.c_params)
                    # 实现c_loss对c_params每一个参数的求导，返回一个list

            with tf.name_scope('sync'):  # worker和global的同步过程
                with tf.name_scope('pull'):  # 获取global参数,复制到local—net
                    self.pull_a_params_op = [l_p.assign(g_p) for l_p, g_p in zip(self.a_params, globalAC.a_params)]
                    self.pull_c_params_op = [l_p.assign(g_p) for l_p, g_p in zip(self.c_params, globalAC.c_params)]
                with tf.name_scope('push'):  # 将参数传送到gloabl中去
                    self.update_a_op = OPT_A.apply_gradients(zip(self.a_grads, globalAC.a_params))
                    self.update_c_op = OPT_C.apply_gradients(zip(self.c_grads, globalAC.c_params))
                    # 其中传送的是local—net的actor和critic的参数梯度grads,具体计算在上面定义
                    # apply_gradients是tf.train.Optimizer中自带的功能函数，将求得的梯度参数更新到global中

    def _build_net(self, scope):
        w_init = tf.random_normal_initializer(0., .1)  # 返回一个生成具有正态分布的张量的初始化器
        with tf.variable_scope('actor'):
            l_a = tf.layers.dense(self.s, 200, tf.nn.relu6, kernel_initializer=w_init, name='la')
            mu = tf.layers.dense(l_a, N_A, tf.nn.tanh, kernel_initializer=w_init, name='mu')
            sigma = tf.layers.dense(l_a, N_A, tf.nn.softplus, kernel_initializer=w_init, name='sigma')
            # actor 输出动作的均值和方差
        with tf.variable_scope('critic'):
            l_c = tf.layers.dense(self.s, 100, tf.nn.relu6, kernel_initializer=w_init, name='lc')
            v = tf.layers.dense(l_c, 1, kernel_initializer=w_init, name='v')
            # critic 输出state value用于计算td
        a_params = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope=scope + '/actor')
        c_params = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope=scope + '/critic')
        return mu, sigma, v, a_params, c_params   # return 均值, 方差, state_value

    def update_global(self, feed_dict):  # push
        SESS.run([self.update_a_op, self.update_c_op], feed_dict)  # 进行 push 操作

    def pull_global(self):
        SESS.run([self.pull_a_params_op, self.pull_c_params_op])  # 进行 pull 操作

    def choose_action(self, s):
        s = s[np.newaxis, :]
        return SESS.run(self.A, {self.s: s})  # 根据 s 选动作


class Worker(object):
    def __init__(self, name, globalAC):
        self.env = gym.make(GAME).unwrapped  # 创建自己的环境
        self.name = name  # 自己的名字
        self.AC = ACNet(name, globalAC)   # 自己的 local net, 并绑定上 globalAC

    def work(self):
        global GLOBAL_RUNNING_R, GLOBAL_EP  # R是所有worker的总reward，ep是所有worker的总episode
        total_step = 1  # 本worker的总步数
        buffer_s, buffer_a, buffer_r = [], [], []   # s, a, r 的缓存, 用于 n_steps 更新
        while not COORD.should_stop() and GLOBAL_EP < MAX_GLOBAL_EP:  # worker运行的条件
            s = self.env.reset()  # 重置环境
            ep_r = 0  # 统计ep的总reward
            for ep_t in range(MAX_EP_STEP):
                # if self.name == 'W_0':  # 只有worker0才将动画图像显示
                #     self.env.render()
                a = self.AC.choose_action(s)  # 将当前状态state传入AC网络选择动作action
                s_, r, done, info = self.env.step(a)  # 行动并获得s_和r等信息
                done = True if ep_t == MAX_EP_STEP - 1 else False  #

                ep_r += r  # 记录本回合总体reward
                buffer_s.append(s)  # 将当前s，a和r加入缓存
                buffer_a.append(a)
                buffer_r.append((r+8)/8)    # normalize
                # TD(n)的架构
                if total_step % UPDATE_GLOBAL_ITER == 0 or done:   # 每 UPDATE_GLOBAL_ITER 步 或者回合完了, 进行 sync 操作
                    # 获得用于计算 TD error 的 下一 state 的 value
                    if done:
                        v_s_ = 0   # terminal
                    else:
                        v_s_ = SESS.run(self.AC.v, {self.AC.s: s_[np.newaxis, :]})[0, 0]  # reduce dim from 2 to 0
                    buffer_v_target = []   # 下 state value 的缓存, 用于算 TD

                    for r in buffer_r[::-1]:     # 进行 n_steps forward view
                        v_s_ = r + GAMMA * v_s_
                        buffer_v_target.append(v_s_)  # 将每一步的v现实都加入缓存中
                    buffer_v_target.reverse()

                    buffer_s, buffer_a, buffer_v_target = np.vstack(buffer_s), np.vstack(buffer_a), np.vstack(buffer_v_target)
                    feed_dict = {
                        self.AC.s: buffer_s,  # 本次走过的所有状态，用于计算v估计
                        self.AC.a_his: buffer_a,  # 本次进行过的所有操作，用于计算a—loss
                        self.AC.v_target: buffer_v_target,  # 走过的每一个state的v现实值，用于计算td
                    }
                    # 更新全局网络的参数
                    self.AC.update_global(feed_dict)    # update gradients on global network
                    buffer_s, buffer_a, buffer_r = [], [], []   # 清空缓存
                    self.AC.pull_global()  # update local network from global network

                s = s_
                total_step += 1  # 本回合总步数加1
                if done:
                    if len(GLOBAL_RUNNING_R) == 0:  # record running episode reward
                        GLOBAL_RUNNING_R.append(ep_r)
                    else:
                        GLOBAL_RUNNING_R.append(0.9 * GLOBAL_RUNNING_R[-1] + 0.1 * ep_r)
                    print(
                        self.name,
                        "Ep:", GLOBAL_EP,
                        "| Ep_r: %i" % GLOBAL_RUNNING_R[-1],
                          )
                    GLOBAL_EP += 1   # 加一回合
                    break  # 结束这回合

if __name__ == "__main__":
    SESS = tf.Session()

    with tf.device("/cpu:0"):  # 指定在cpu：0进行以下代码（CPU不区分设备号，统一使用 /cpu:0）
        OPT_A = tf.train.RMSPropOptimizer(LR_A, name='RMSPropA')   # 创建Actor的优化器
        OPT_C = tf.train.RMSPropOptimizer(LR_C, name='RMSPropC')   # 创建Critic的优化器
        GLOBAL_AC = ACNet(GLOBAL_NET_SCOPE)   # 创建全局网络GLOBAL_AC
        workers = []  # workers列表

        # 创建 worker
        for i in range(N_WORKERS):
            # 创建n个worker，worker的数量最好和cpu的核一致，因为每个线程都是在一个单独的cpu进行
            i_name = 'W_%i' % i   # worker name
            workers.append(Worker(i_name, GLOBAL_AC))  # 创建worker，并放在workers列表中，方便统一管理
            # 把每个worker对象都存放在一个workers列表中，方便使用

    COORD = tf.train.Coordinator()   # Tensorflow 用于并行的工具
    SESS.run(tf.global_variables_initializer())  # global变量初始化

    if OUTPUT_GRAPH:
        if os.path.exists(LOG_DIR):
            shutil.rmtree(LOG_DIR)
        tf.summary.FileWriter(LOG_DIR, SESS.graph)  # 生成 tensorboard

    worker_threads = []
    for worker in workers:  # 执行每一个worker
        # t = threading.Thread(target=worker.work)
        job = lambda: worker.work()   # worker要执行的工作
        t = threading.Thread(target=job)  # threading.Thread(target=job)创建线程，其中target要执行的函数
        t.start()  # 开始线程，并执行
        worker_threads.append(t)  # 把线程加入worker_threads中
    COORD.join(worker_threads)  # 线程由COORD统一管理即可


    plt.plot(np.arange(len(GLOBAL_RUNNING_R)), GLOBAL_RUNNING_R)
    plt.xlabel('step')
    plt.ylabel('Total moving reward')
    plt.show()

源码github：源代码