-
Notifications
You must be signed in to change notification settings - Fork 0
/
qAgent.py
163 lines (119 loc) · 4.15 KB
/
qAgent.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
import qEnvironment
import deepQ
import time
import h5py
import os, os.path
import matplotlib.pyplot as plt
import pickle
import numpy as np
import random
'''
Agents : Takes Action in the environment,sets up simulation(Episodes, step size)
Environment : sets up the environment(v-rep simulation), returns rewards and next states to agent based on the actions it took
Algorithm : The Learning Algorithm
'''
class Agent:
def __init__(self):
self.epsilon = 0.25
self.inputs = 3 #state
self.outputs = 3 #action
self.dicountFactor = 0.8
self.learningRate = 0.25
self.savePath='pioneer_qlearn_deep/ep'
self.network_layers = [4]
self.graphPath = 'Images/'
self.rewardFile = 'Files/rewards.pickle'
self.stepFile = 'Files/steps.pickle'
self.dict = {}
def start(self):
file_count= len([name for name in os.listdir('pioneer_qlearn_deep') ])
# print("FILE",file_count)
weights_path = 'pioneer_qlearn_deep/ep'+str(file_count)+'.h5'
deepQLearn = deepQ.NNQ(self.inputs,self.outputs,self.dicountFactor,self.learningRate)
deepQLearn.initNetworks(self.network_layers)
#deepQLearn.plotModel('/home/kaizen/BTP/Python/NeuralNet/Images/')
if file_count != 0:
deepQLearn.loadWeights(weights_path)
print("Weights Loaded from path ",weights_path,"\n")
env = qEnvironment.Environment()
num_episodes = 10000
steps = 200
start_time = time.time()
# for plotting,data per episode
stepList = []
rewardList = []
index = 1
replay = [] # stores tuples of (S, A, R, S').
total_steps_in_simulation = 0
observe = 500
max_buffer_size = 10000 # average 30 steps, 10000/30 = 333 episodes
batch_size = 100
for episode in range(num_episodes):
state = env.reset()
while type(state) ==int:
state = env.reset()
cumulated_reward = 0
for step in range(steps):
print("state",state,end="" )
qValues = deepQLearn.getQValues(state)
action = deepQLearn.selectAction(qValues, self.epsilon )
self.dict[''.join(str(e) for e in state)] = action
nextState,reward,done,info = env.step(action)
cumulated_reward += reward
# LEARNING PART
replay.append((state,action,reward,nextState,done))
if total_steps_in_simulation > observe:
if len(replay) > max_buffer_size:
replay.pop(0)
training_set = random.sample(replay,observe)
deepQLearn.learn_on_minbatch(training_set,batch_size)
# deepQLearn.learn_on_one_example(state,action,reward,nextState,done,batch_size = batch_size)
if not(done):
state = nextState
else:
print('done')
break
print("Step = ",step)
# Average time per step = 0.004s
total_steps_in_simulation += 1
time.sleep(0.8)
stepList.append(step)
rewardList.append(cumulated_reward)
m, s = divmod(int(time.time() - start_time), 60)
h, m = divmod(m, 60)
print ("\n\n\EP "+str(episode+1)+" Reward: "+ str(cumulated_reward) +" Time: %d:%02d:%02d" % (h, m, s))
# time.sleep(0.5)
if (episode +1)%50 == 0:
print(replay)
rewardList = pickle.load(open(self.rewardFile, 'rb')) + rewardList
stepList = pickle.load(open(self.stepFile, 'rb')) + stepList
print(len(rewardList))
#print(stepList)
pickle.dump(rewardList, open(self.rewardFile, 'wb'))
pickle.dump(stepList, open(self.stepFile, 'wb'))
''' COMMMENT THESE IF TESTING ANYTHING TO PREVENT DATA DAMAGE'''
deepQLearn.saveModel(self.savePath+str(file_count + index)+'.h5')
index = index + 1
deepQLearn.saveQValues(episode)
deepQLearn.saveWeights(episode)
stepList = []
rewardList = []
print(self.dict)
# print values
f = open('Files/deep_q_table.txt','a')
print(episode + 1,file = f)
print_state = np.empty((0,3), int)
print_state = np.append(print_state,[0,0,0])
for i in range(4):
print_state[0] = i
for j in range(4):
print_state[1] = j
for k in range(4):
print_state[2] = k
qValues = deepQLearn.getQValues(print_state)
action = deepQLearn.selectAction(qValues, 0 )# no randomness
print(print_state , " ", action, file = f)
f.close()
if __name__ == '__main__':
agent = Agent()
agent.start()