-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathddpg_update_v2.py
181 lines (148 loc) · 6.7 KB
/
ddpg_update_v2.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
import os.path
import sys
import gym
import numpy as np
import tensorflow as tf
import time
##################### hyper parameters ####################
MAX_EPISODES = 200
MAX_EP_STEPS = 200
LR_A = 0.001
LR_C = 0.002
GAMMA = 0.9
TAU = 0.01
MEMORY_CAPACITY = 10000
BATCH_SIZE = 32
RENDER = True
ENV_NAME = 'Pendulum-v0'
############################### DDPG ####################################
class DDPG(object):
def __init__(self, a_dim, s_dim, a_bound):
tf.compat.v1.disable_eager_execution()
self.memory = np.zeros(
(MEMORY_CAPACITY, s_dim * 2 + a_dim + 1), dtype=np.float32)
self.pointer = 0
self.sess = tf.compat.v1.Session()
self.a_dim, self.s_dim, self.a_bound = a_dim, s_dim, a_bound,
self.S = tf.compat.v1.placeholder(tf.float32, [None, s_dim], 's')
self.S_ = tf.compat.v1.placeholder(tf.float32, [None, s_dim], 's_')
self.R = tf.compat.v1.placeholder(tf.float32, [None, 1], 'r')
with tf.compat.v1.variable_scope('Actor'):
self.a = self._build_a(
self.S, scope='eval', trainable=True)
a_ = self._build_a(self.S_, scope='target', trainable=False)
with tf.compat.v1.variable_scope('Critic'):
# assign self.a = a in memory when calculating q for td_error,
# otherwise the self.a is from Actor when updating Actor
q = self._build_c(self.S, self.a, scope='eval', trainable=True)
q_ = self._build_c(self.S_, a_, scope='target', trainable=False)
# networks parameters
self.ae_params = tf.compat.v1.get_collection(
tf.compat.v1.GraphKeys.GLOBAL_VARIABLES, scope='Actor/eval')
self.at_params = tf.compat.v1.get_collection(
tf.compat.v1.GraphKeys.GLOBAL_VARIABLES, scope='Actor/target')
self.ce_params = tf.compat.v1.get_collection(
tf.compat.v1.GraphKeys.GLOBAL_VARIABLES, scope='Critic/eval')
self.ct_params = tf.compat.v1.get_collection(
tf.compat.v1.GraphKeys.GLOBAL_VARIABLES, scope='Critic/target')
# target net replacement
self.soft_replace = [tf.compat.v1.assign(t, (1 - TAU) * t + TAU * e)
for t, e in zip(self.at_params + self.ct_params, self.ae_params + self.ce_params)]
q_target = self.R + GAMMA * q_
# in the feed_dic for the td_error, the self.a should change to actions in memory
td_error = tf.compat.v1.losses.mean_squared_error(
labels=q_target, predictions=q)
self.ctrain = tf.compat.v1.train.AdamOptimizer(LR_C).minimize(
td_error, var_list=self.ce_params)
a_loss = - tf.reduce_mean(input_tensor=q) # maximize the q
self.atrain = tf.compat.v1.train.AdamOptimizer(
LR_A).minimize(a_loss, var_list=self.ae_params)
# self.sess.run(tf.compat.v1.global_variables_initializer(), feed_dict={
# self.S: s_init[np.newaxis, :], self.S_: s_init[np.newaxis, :]})
self.sess.run(tf.compat.v1.global_variables_initializer())
#
def choose_action(self, s):
return self.sess.run(self.a, {self.S: s[np.newaxis, :]})[0]
def learn(self):
# soft target replacement
self.sess.run(self.soft_replace)
indices = np.random.choice(MEMORY_CAPACITY, size=BATCH_SIZE)
bt = self.memory[indices, :]
bs = bt[:, :self.s_dim]
ba = bt[:, self.s_dim: self.s_dim + self.a_dim]
br = bt[:, -self.s_dim - 1: -self.s_dim]
bs_ = bt[:, -self.s_dim:]
self.sess.run(self.atrain, {self.S: bs})
self.sess.run(self.ctrain, {self.S: bs,
self.a: ba, self.R: br, self.S_: bs_})
def store_transition(self, s, a, r, s_):
transition = np.hstack((s, a, [r], s_))
# replace the old memory with new memory
index = self.pointer % MEMORY_CAPACITY
self.memory[index, :] = transition
self.pointer += 1
def _build_a(self, s, scope, trainable):
with tf.compat.v1.variable_scope(scope):
net = tf.compat.v1.layers.dense(
s, 30, activation=tf.nn.relu, name='l1', trainable=trainable)
a = tf.compat.v1.layers.dense(
net, self.a_dim, activation=tf.nn.tanh, name='a', trainable=trainable)
return tf.multiply(a, self.a_bound, name='scaled_a')
def _build_c(self, s, a, scope, trainable):
with tf.compat.v1.variable_scope(scope):
n_l1 = 30
w1_s = tf.compat.v1.get_variable(
'w1_s', [self.s_dim, n_l1], trainable=trainable)
w1_a = tf.compat.v1.get_variable(
'w1_a', [self.a_dim, n_l1], trainable=trainable)
b1 = tf.compat.v1.get_variable(
'b1', [1, n_l1], trainable=trainable)
net = tf.nn.relu(tf.matmul(s, w1_s) + tf.matmul(a, w1_a) + b1)
# Q(s,a)
return tf.compat.v1.layers.dense(net, 1, trainable=trainable)
############################### training ####################################
env = gym.make(ENV_NAME)
env = env.unwrapped
env.seed(1)
Rs = []
# 2*ZONE+1
# the first ZONE number is demand(i.e. how many bikes are taken away in this ZONE)
# the last ZONE number is the amount of resource on zone K (dS_) + time
s_dim = env.observation_space.shape[0]
# print(s_dim)
# equal to get_observe function in env
a_dim = env.action_space.shape[0]
# s = env.reset()
# print(a_dim,"YEEEEEEE")
# print(env.action_space.low,"low")
a_bound = env.action_space.high # higher bound, which is set in the .txt file
ddpg = DDPG(a_dim, s_dim, a_bound)
var = 3 # control exploration
t1 = time.time()
for i in range(MAX_EPISODES):
s = env.reset()
ep_reward = 0
for j in range(MAX_EP_STEPS):
if RENDER:
env.render()
# Add exploration noise
a = ddpg.choose_action(s)
# add randomness to action selection for exploration
a = np.clip(np.random.normal(a, var), -2, 2)
s_, r, done, info = env.step(a)
ddpg.store_transition(s, a, r / 10, s_)
if ddpg.pointer > MEMORY_CAPACITY:
var *= .9995 # decay the action randomness
ddpg.learn()
s = s_
ep_reward += r
if j == MAX_EP_STEPS-1:
print('Episode:', i, ' Reward: %i' %
int(ep_reward), 'Explore: %.2f' % var, )
# if ep_reward > -300:RENDER = True
break
# Rs.append(ep_reward)
print('Running time: ', time.time() - t1)
print('')
# print('---------------------------')
# print('Average reward per episode:', np.average(Rs))