-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathPPO.py
261 lines (242 loc) · 10.7 KB
/
PPO.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
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
import gym
from new_network_state import make_network
import numpy as np
import random
from gym import spaces
from stable_baselines.common.policies import MlpPolicy
from stable_baselines.common.vec_env import DummyVecEnv
from stable_baselines import PPO2
import time
random.seed(32184939)
class NetworkEnv(gym.Env):
metadata={'render.modes': ['human']}
def __init__(self,state,ans,chk) -> None:
super().__init__()
self.action_space=spaces.Discrete(2)
if chk==3:
self.observation_space=spaces.Box(low=1,high=3,shape=(1,1,10),dtype=np.int)
else:
self.observation_space=spaces.Box(low=1,high=5,shape=(1,1,10),dtype=np.int)
self.reward=0
self.next_state=state[0]
self.state=state
self.answer=ans
self.idx=0
self.terminal=False
self.chk=chk
def reward_cal(self,future_state,predict_Action):
# 1 idle 2 zigbee occupy 3 wifi occupy 4 wifi only idle 5 transmission
if chk==3:
if future_state==3 and predict_Action==1:
reward=0.8
return reward
elif future_state==3 and predict_Action==0:
reward=-0.8
return reward
elif (future_state==1 or future_state==2 ) and predict_Action==1:
reward=-0.3
return reward
elif (future_state==1 or future_state==2 ) and predict_Action==0:
reward=0.1
return reward
else:
if future_state==5 and predict_Action==1:
reward=0.8
return reward
elif future_state==5 and predict_Action==0:
reward=-0.8
return reward
elif (future_state==1 or future_state==2 or future_state==3 or future_state==4) and predict_Action==1:
reward=-0.3
return reward
elif (future_state==1 or future_state==2 or future_state==3 or future_state==4) and predict_Action==0:
reward=0.1
return reward
def step(self, action):
self.idx+=1
self.terminal=False
if self.idx>=10000:
self.reset()
return self.next_state,self.reward,self.terminal,{}
elif (self.idx+1)%5==0:
self.next_state=self.state[self.idx]
self.reward=self.reward_cal(self.answer[self.idx-1],action)
return self.next_state, self.reward, self.terminal,{}
else:
self.next_state=self.state[self.idx]
self.reward=self.reward_cal(self.answer[self.idx-1],action)
return self.next_state, self.reward, self.terminal,{}
def reset(self):
self.action_space=spaces.Discrete(2)
if self.chk==3:
self.observation_space=spaces.Box(low=1,high=3,shape=(1,1,10),dtype=np.int)
else:
self.observation_space=spaces.Box(low=1,high=5,shape=(1,1,10),dtype=np.int)
self.reward=0
self.next_state=self.state[0]
self.idx=0
return self.next_state
def render(self, mode="human") :
return super().render(mode)
chk=3
for a in range(2):
jam_trans_sum1=0
wait_trans_sum1=0
jam_idle_sum1=0
jam_occupy_sum1=0
react_success_sum1=0
react_fail_sum1=0
idle_num_sum1=0
occupy_num_sum1=0
trans_num_sum1=0
jam_trans_sum2=0
wait_trans_sum2=0
jam_idle_sum2=0
jam_occupy_sum2=0
react_success_sum2=0
react_fail_sum2=0
idle_num_sum2=0
occupy_num_sum2=0
trans_num_sum2=0
jam_trans_sum3=0
wait_trans_sum3=0
jam_idle_sum3=0
jam_occupy_sum3=0
react_success_sum3=0
react_fail_sum3=0
idle_num_sum3=0
occupy_num_sum3=0
trans_num_sum3=0
if a==1:
chk=5
for _ in range(1):
for i in range(1,4):
num=i
for_train,slot_state=make_network(num,chk)
for_train=np.array(for_train,dtype=list)
env=NetworkEnv(for_train,slot_state,chk)
env=DummyVecEnv([lambda: env])
start=time.time()
model=PPO2(MlpPolicy,env,verbose=0)
model.learn(total_timesteps=100000)
answer=[]
jam_trans=0
jam_idle=0
jam_occupy=0
wait_idle=0
wait_occupy=0
wait_trans=0
react_success=0
react_fail=0
idle_num=0
occupy_num=0
trans_num=0
if chk==3:
for _ in range(10000):
test=for_train[_].reshape(1,1,10)
pred=model.predict(test)[0]
if pred==1 and slot_state[_]==3:
jam_trans+=1
elif pred==1 and (slot_state[_]==1):
jam_idle+=1
elif pred==1 and (slot_state[_]==2 ):
jam_occupy+=1
elif pred==0 and (slot_state[_]==1 ):
wait_idle+=1
elif pred==0 and (slot_state[_]==2 ):
wait_occupy+=1
elif pred==0 and slot_state[_]==3:
wait_trans+=1
if (slot_state[_]==1) :
idle_num+=1
elif (slot_state[_]==2 ):
react_fail+=1
occupy_num+=1
elif slot_state[_]==3 :
trans_num+=1
react_success+=1
else:
for _ in range(10000):
test=for_train[_].reshape(1,1,10)
pred=model.predict(test)[0]
if pred==1 and slot_state[_]==5 :
jam_trans+=1
elif pred==1 and (slot_state[_]==1 or slot_state[_]==4):
jam_idle+=1
elif pred==1 and (slot_state[_]==2 or slot_state[_]==3):
jam_occupy+=1
elif pred==0 and (slot_state[_]==1 or slot_state[_]==4):
wait_idle+=1
elif pred==0 and (slot_state[_]==2 or slot_state[_]==3):
wait_occupy+=1
elif pred==0 and slot_state[_]==3:
wait_trans+=1
if (slot_state[_]==1 or slot_state[_]==4) :
idle_num+=1
elif (slot_state[_]==2 or slot_state[_]==3):
react_fail+=1
occupy_num+=1
elif slot_state[_]==5 :
trans_num+=1
react_success+=1
if i==1:
jam_trans_sum1+=jam_trans
wait_trans_sum1+=wait_trans
jam_idle_sum1+=jam_idle
jam_occupy_sum1+=jam_occupy
react_success_sum1+=react_success
react_fail_sum1+=react_fail
idle_num_sum1+=idle_num
occupy_num_sum1+=occupy_num
trans_num_sum1+=trans_num
print(f'\nstate{chk}-1')
print(f'\njam success {jam_trans} jam idle {jam_idle} jam occupy {jam_occupy} wait idle {wait_idle} wait occupy {wait_occupy} wait trans {wait_trans}')
print(f'\nreact success {react_success_sum1/1} react fail {react_fail_sum1/1}')
print(f'\njam rate {jam_trans_sum1/(jam_trans_sum1+wait_trans_sum1)}')
print(f'\nHit rate deep {jam_trans_sum1/(jam_trans_sum1+jam_idle_sum1+jam_occupy_sum1)}')
print(f'\nHit rate react {react_success_sum1/(react_success_sum1+react_fail_sum1)}')
print(f'\nidle num {idle_num_sum1/1} occupy num {occupy_num_sum1/1} trans num {trans_num_sum1/1}')
elif i==2:
jam_trans_sum2+=jam_trans
wait_trans_sum2+=wait_trans
jam_idle_sum2+=jam_idle
jam_occupy_sum2+=jam_occupy
react_success_sum2+=react_success
react_fail_sum2+=react_fail
idle_num_sum2+=idle_num
occupy_num_sum2+=occupy_num
trans_num_sum2+=trans_num
print(f'\nstate{chk}-2')
print(f'\njam success {jam_trans} jam idle {jam_idle} jam occupy {jam_occupy} wait idle {wait_idle} wait occupy {wait_occupy} wait trans {wait_trans}')
print(f'\nreact success {react_success_sum2/1} react fail {react_fail_sum2/1}')
print(f'\njam rate {jam_trans_sum2/(jam_trans_sum2+wait_trans_sum2)}')
print(f'\nHit rate deep {jam_trans_sum2/(jam_trans_sum2+jam_idle_sum2+jam_occupy_sum2)}')
print(f'\nHit rate react {react_success_sum2/(react_success_sum2+react_fail_sum2)}')
print(f'\nidle num {idle_num_sum2/1} occupy num {occupy_num_sum2/1} trans num {trans_num_sum2/1}')
elif i==3:
jam_trans_sum3+=jam_trans
wait_trans_sum3+=wait_trans
jam_idle_sum3+=jam_idle
jam_occupy_sum3+=jam_occupy
react_success_sum3+=react_success
react_fail_sum3+=react_fail
idle_num_sum3+=idle_num
occupy_num_sum3+=occupy_num
trans_num_sum3+=trans_num
print(f'\nstate{chk}-3')
print(f'\njam success {jam_trans} jam idle {jam_idle} jam occupy {jam_occupy} wait idle {wait_idle} wait occupy {wait_occupy} wait trans {wait_trans}')
print(f'\nreact success {react_success_sum3/1} react fail {react_fail_sum3/1}')
print(f'\njam rate {jam_trans_sum3/(jam_trans_sum3+wait_trans_sum3)}')
print(f'\nHit rate deep {jam_trans_sum3/(jam_trans_sum3+jam_idle_sum3+jam_occupy_sum3)}')
print(f'\nHit rate react {react_success_sum3/(react_success_sum3+react_fail_sum3)}')
print(f'\nidle num {idle_num_sum3/1} occupy num {occupy_num_sum3/1} trans num {trans_num_sum3/1}')
'''print(f"\n-------------------zigbee {num}-------------------------")
print(f'\njam success {jam_trans} jam idle {jam_idle} jam occupy {jam_occupy} wait idle {wait_idle} wait occupy {wait_occupy} wait trans {wait_trans}')
print(f'\nreact success {react_success} react fail {react_fail}')
print(f'\njam rate {jam_trans/(jam_trans+wait_trans)}')
print(f'\nHit rate deep {jam_trans/(jam_trans+jam_idle+jam_occupy)}')
print(f'\nHit rate react {react_success/(react_success+react_fail)}')
print(f'\nidle num {idle_num} occupy num {occupy_num} trans num {trans_num}')
print("\nstate3")
print("\n------------------------------------------------------")'''
print(time.time()-start)