-
Notifications
You must be signed in to change notification settings - Fork 27
/
ddpg02.py
185 lines (163 loc) · 6.09 KB
/
ddpg02.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
182
183
184
185
# -*- coding: utf-8 -*-
"""
Created on Sun Oct 28 10:33:19 2018
DDGP CU action-critic / gradient
self.cost = - self.q
self.cost = tf.nn.l2_loss(self.q_tar - self.q)
Baseline 2 feature: 1.525
Baseline 3 feature: 1.751
@author: mengxiaomao
"""
import time
import numpy as np
import tensorflow as tf
from Environment_CU import Env_cellular
from DDPG_2 import DDPG, Actor, Critic
fd = 10
Ts = 20e-3
n_x = 5
n_y = 5
L = 2
C = 16
maxM = 4 # user number in one BS
min_dis = 0.01 #km
max_dis = 1. #km
max_p = 38. #dBm
p_n = -114. #dBm
power_num = 1
def Train(sess, env, weight_file):
max_reward = 0
max_episode = 5000
Ns = 11
env.set_Ns(Ns)
dnn = DDPG(env, weight_file)
actor = Actor(sess, dnn)
critic = Critic(sess, dnn)
tf.global_variables_initializer().run()
interval = 100
st = time.time()
reward_hist = list()
loss_hist = list()
for k in range(1, max_episode+1):
reward_list = list()
loss_list = list()
s_actor, s_critic = env.reset()
for i in range(int(Ns)-1):
p, p_exp = actor.get_random_action(s_actor, k)
s_actor_next, s_critic_next, q_tar, r = env.step(p_exp[:,0])
loss = critic.train(s_critic, p_exp, q_tar)
grads = critic.get_gradient(s_critic, p)
actor.train(s_actor, grads[0])
s_actor, s_critic = s_actor_next, s_critic_next
reward_list.append(r)
loss_list.append(loss)
reward_hist.append(np.mean(reward_list))
loss_hist.append(np.mean(loss_list))
if k % interval == 0:
reward = np.mean(reward_hist[-interval:])
if reward > max_reward:
dnn.save_params()
max_reward = reward
print("Episode(train):%d DDPG: %.3f Loss: %.3f Time cost: %.2fs"
%(k, reward, np.mean(loss_hist[-interval:]), time.time()-st))
st = time.time()
return reward_hist
def Test(sess, env, weight_file):
max_episode = 100
Ns = 5e2+1
env.set_Ns(Ns)
dnn = DDPG(env, weight_file)
actor = Actor(sess, dnn)
critic = Critic(sess, dnn)
tf.global_variables_initializer().run()
actor.load_params()
critic.load_params()
st = time.time()
reward_hist = list()
for k in range(1, max_episode+1):
reward_list = list()
s_actor, _ = env.reset()
for i in range(int(Ns)-1):
p = actor.predict_p(s_actor)
s_actor_next, _, _, r = env.step(p[:,0])
s_actor = s_actor_next
reward_list.append(r)
reward_hist.append(np.mean(reward_list))
print("Episode(test):%d DDPG: %.3f Time cost: %.2fs"
%(k, reward_hist[-1], time.time()-st))
st = time.time()
print("Test average rate: %.3f" %(np.mean(reward_hist)))
return reward_hist
def Train_ddpg(weight_file, fd, max_dis, maxM):
env = Env_cellular(fd, Ts, n_x, n_y, L, C, maxM, min_dis, max_dis, max_p, p_n, power_num)
tf.reset_default_graph()
with tf.Session() as sess:
Train(sess, env, weight_file)
def Test_ddpg(weight_file, max_episode, Ns, fd, max_dis, maxM):
env = Env_cellular(fd, Ts, n_x, n_y, L, C, maxM, min_dis, max_dis, max_p, p_n, power_num)
tf.reset_default_graph()
with tf.Session() as sess:
env.set_Ns(Ns)
dnn = DDPG(env, weight_file)
actor = Actor(sess, dnn)
tf.global_variables_initializer().run()
actor.load_params()
reward_hist = list()
for k in range(1, max_episode+1):
reward_list = list()
s_actor, _ = env.reset()
for i in range(int(Ns)-1):
p = actor.predict_p(s_actor)
s_actor_next, _, _, r = env.step(p[:,0])
s_actor = s_actor_next
reward_list.append(r)
reward_hist.append(np.mean(reward_list))
return np.mean(reward_hist)
def Test_ddpg_all(weight_file, max_episode, Ns, fd, max_dis, maxM):
env = Env_cellular(fd, Ts, n_x, n_y, L, C, maxM, min_dis, max_dis, max_p, p_n, power_num)
tf.reset_default_graph()
with tf.Session() as sess:
env.set_Ns(Ns)
dnn = DDPG(env, weight_file)
actor = Actor(sess, dnn)
tf.global_variables_initializer().run()
actor.load_params()
reward_hist = list()
for k in range(1, max_episode+1):
reward_list = list()
s_actor, _ = env.reset()
for i in range(int(Ns)-1):
p = actor.predict_p(s_actor)
s_actor_next, _, _, r = env.step(p[:,0])
s_actor = s_actor_next
reward_list.append(r)
reward_hist.append(np.mean(reward_list))
return np.mean(reward_hist)
def Test_ddpg_time(weight_file, max_episode, Ns, fd, max_dis, maxM):
env = Env_cellular(fd, Ts, n_x, n_y, L, C, maxM, min_dis, max_dis, max_p, p_n, power_num)
tf.reset_default_graph()
with tf.Session() as sess:
env.set_Ns(Ns)
dnn = DDPG(env, weight_file)
actor = Actor(sess, dnn)
tf.global_variables_initializer().run()
actor.load_params()
time_cost = 0
for k in range(1, max_episode+1):
s_actor, _ = env.reset()
for i in range(int(Ns)-1):
st = time.time()
p = actor.predict_p(s_actor)
time_cost = time_cost + time.time() - st
s_actor_next, _, _, _ = env.step(p[:,0])
s_actor = s_actor_next
return time_cost
if __name__ == '__main__':
env = Env_cellular(fd, Ts, n_x, n_y, L, C, maxM, min_dis, max_dis, max_p, p_n, power_num)
weight_file = 'C:/Software/workshop/python/ddpg_2.mat'
tf.reset_default_graph()
with tf.Session() as sess:
reward_train = Train(sess, env, weight_file)
tf.reset_default_graph()
with tf.Session() as sess:
reward_test = Test(sess, env, weight_file)