-
Notifications
You must be signed in to change notification settings - Fork 1
/
qlearnning_maze_mofan.py
470 lines (408 loc) · 18.5 KB
/
qlearnning_maze_mofan.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
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
#### 搭建可视化环境 ####
import numpy as np
import pandas as pd
import time
import sys
import math
if sys.version_info.major == 2:
import Tkinter as tk
else:
import tkinter as tk
UNIT = 40 # pixels
MAZE_H = 9 # grid height
MAZE_W = 9 # grid width
class Maze(tk.Tk, object):
def __init__(self):
super(Maze, self).__init__()
self.action_space = ['u', 'd', 'l', 'r']
self.n_actions = len(self.action_space)
self.n_features = 4
self.title('maze')
self.geometry('{0}x{1}'.format(MAZE_H * UNIT, MAZE_H * UNIT))
self._build_maze()
def _build_maze(self):
self.canvas = tk.Canvas(self, bg='white',
height=MAZE_H * UNIT,
width=MAZE_W * UNIT)
# create grids
for c in range(0, MAZE_W * UNIT, UNIT):
x0, y0, x1, y1 = c, 0, c, MAZE_H * UNIT
self.canvas.create_line(x0, y0, x1, y1)
for r in range(0, MAZE_H * UNIT, UNIT):
x0, y0, x1, y1 = 0, r, MAZE_W * UNIT, r
self.canvas.create_line(x0, y0, x1, y1)
# create origin
origin = np.array([20, 20])
# hell
hell1_center = origin + np.array([UNIT , 0])
self.hell1 = self.canvas.create_rectangle(
hell1_center[0] - 15, hell1_center[1] - 15,
hell1_center[0] + 15, hell1_center[1] + 15,
fill='black')
# hell
hell2_center = origin + np.array([UNIT * 4, 0])
self.hell2 = self.canvas.create_rectangle(
hell2_center[0] - 15, hell2_center[1] - 15,
hell2_center[0] + 15, hell2_center[1] + 15,
fill='black')
# hell
hell3_center = origin + np.array([UNIT * 8, 0])
self.hell3 = self.canvas.create_rectangle(
hell3_center[0] - 15, hell3_center[1] - 15,
hell3_center[0] + 15, hell3_center[1] + 15,
fill='black')
#hell
hell4_center = origin + np.array([UNIT * 6, UNIT])
self.hell4 = self.canvas.create_rectangle(
hell4_center[0] - 15, hell4_center[1] - 15,
hell4_center[0] + 15, hell4_center[1] + 15,
fill='black')
# hell
hell5_center = origin + np.array([0, UNIT * 2])
self.hell5 = self.canvas.create_rectangle(
hell5_center[0] - 15, hell5_center[1] - 15,
hell5_center[0] + 15, hell5_center[1] + 15,
fill='black')
# hell
hell6_center = origin + np.array([UNIT * 3, UNIT * 2])
self.hell6 = self.canvas.create_rectangle(
hell6_center[0] - 15, hell6_center[1] - 15,
hell6_center[0] + 15, hell6_center[1] + 15,
fill='black')
# hell
hell7_center = origin + np.array([UNIT * 2, UNIT * 3])
self.hell7 = self.canvas.create_rectangle(
hell7_center[0] - 15, hell7_center[1] - 15,
hell7_center[0] + 15, hell7_center[1] + 15,
fill='black')
# hell
hell8_center = origin + np.array([UNIT * 5, UNIT * 3])
self.hell8 = self.canvas.create_rectangle(
hell8_center[0] - 15, hell8_center[1] - 15,
hell8_center[0] + 15, hell8_center[1] + 15,
fill='black')
# hell
hell9_center = origin + np.array([UNIT * 7, UNIT * 3])
self.hell9 = self.canvas.create_rectangle(
hell9_center[0] - 15, hell9_center[1] - 15,
hell9_center[0] + 15, hell9_center[1] + 15,
fill='black')
# hell
hell10_center = origin + np.array([UNIT * 8, UNIT * 3])
self.hell10 = self.canvas.create_rectangle(
hell10_center[0] - 15, hell10_center[1] - 15,
hell10_center[0] + 15, hell10_center[1] + 15,
fill='black')
# hell
hell11_center = origin + np.array([0, UNIT * 4])
self.hell11 = self.canvas.create_rectangle(
hell11_center[0] - 15, hell11_center[1] - 15,
hell11_center[0] + 15, hell11_center[1] + 15,
fill='black')
# hell
hell12_center = origin + np.array([UNIT * 3, UNIT * 4])
self.hell12 = self.canvas.create_rectangle(
hell12_center[0] - 15, hell12_center[1] - 15,
hell12_center[0] + 15, hell12_center[1] + 15,
fill='black')
# hell
hell13_center = origin + np.array([UNIT, UNIT * 6])
self.hell13 = self.canvas.create_rectangle(
hell13_center[0] - 15, hell13_center[1] - 15,
hell13_center[0] + 15, hell13_center[1] + 15,
fill='black')
# hell
hell14_center = origin + np.array([UNIT * 3, UNIT * 7])
self.hell14 = self.canvas.create_rectangle(
hell14_center[0] - 15, hell14_center[1] - 15,
hell14_center[0] + 15, hell14_center[1] + 15,
fill='black')
# hell
hell15_center = origin + np.array([0, UNIT * 8])
self.hell15 = self.canvas.create_rectangle(
hell15_center[0] - 15, hell15_center[1] - 15,
hell15_center[0] + 15, hell15_center[1] + 15,
fill='black')
# hell
hell16_center = origin + np.array([UNIT * 5, UNIT * 5])
self.hell16 = self.canvas.create_rectangle(
hell16_center[0] - 15, hell16_center[1] - 15,
hell16_center[0] + 15, hell16_center[1] + 15,
fill='black')
# hell
hell17_center = origin + np.array([UNIT * 5, UNIT * 6])
self.hell17 = self.canvas.create_rectangle(
hell17_center[0] - 15, hell17_center[1] - 15,
hell17_center[0] + 15, hell17_center[1] + 15,
fill='black')
# hell
hell18_center = origin + np.array([UNIT * 5, UNIT * 7])
self.hell18 = self.canvas.create_rectangle(
hell18_center[0] - 15, hell18_center[1] - 15,
hell18_center[0] + 15, hell18_center[1] + 15,
fill='black')
# hell
hell19_center = origin + np.array([UNIT * 6, UNIT * 5])
self.hell19 = self.canvas.create_rectangle(
hell19_center[0] - 15, hell19_center[1] - 15,
hell19_center[0] + 15, hell19_center[1] + 15,
fill='black')
# hell
hell20_center = origin + np.array([UNIT * 7, UNIT * 5])
self.hell20 = self.canvas.create_rectangle(
hell20_center[0] - 15, hell20_center[1] - 15,
hell20_center[0] + 15, hell20_center[1] + 15,
fill='black')
# hell
hell21_center = origin + np.array([UNIT * 7, UNIT * 6])
self.hell21 = self.canvas.create_rectangle(
hell21_center[0] - 15, hell21_center[1] - 15,
hell21_center[0] + 15, hell21_center[1] + 15 ,
fill='black')
# hell
hell22_center = origin + np.array([UNIT * 7, UNIT * 7])
self.hell22 = self.canvas.create_rectangle(
hell22_center[0] - 15, hell22_center[1] - 15,
hell22_center[0] + 15, hell22_center[1] + 15,
fill='black')
# create oval
oval_center = origin + UNIT * 6
self.oval = self.canvas.create_oval(
oval_center[0] - 15, oval_center[1] - 15,
oval_center[0] + 15, oval_center[1] + 15,
fill='yellow')
# create red rect
self.rect = self.canvas.create_rectangle(
origin[0] - 15, origin[1] - 15,
origin[0] + 15, origin[1] + 15,
fill='red')
# pack all
self.canvas.pack()
def reset(self):
self.update()
time.sleep(0.5)
self.canvas.delete(self.rect)
origin = np.array([20, 20])
self.rect = self.canvas.create_rectangle(
origin[0] - 15, origin[1] - 15,
origin[0] + 15, origin[1] + 15,
fill='red')
# return observation
return self.canvas.coords(self.rect) # list [起始点四个角点的坐标]
def step(self, action,episode):
s = self.canvas.coords(self.rect)
base_action = np.array([0, 0])
self.path_line = None
if action == 0: # up
if s[1] > UNIT:
base_action[1] -= UNIT
elif action == 1: # down
if s[1] < (MAZE_H - 1) * UNIT:
base_action[1] += UNIT
elif action == 2: # right
if s[0] < (MAZE_W - 1) * UNIT:
base_action[0] += UNIT
elif action == 3: # left
if s[0] > UNIT:
base_action[0] -= UNIT
self.canvas.move(self.rect, base_action[0], base_action[1]) # move agent
# 绘制路径线
if episode == 499:
x1, y1, x2, y2 = self.canvas.coords(self.rect)
if self.path_line is None:
self.path_line = self.canvas.create_line(x1, y1, x2, y2, fill='red')
else:
self.canvas.coords(self.path_line, x1, y1, x2, y2)
s_ = self.canvas.coords(self.rect) # next state
# reward function
if s_ == self.canvas.coords(self.oval):
reward = 2
done = True
s_ = [1,1,1,1] # 一个特殊的列表，代表Terminal,只是起到标志性的作用，所以该列表也可以任意设置，最好不要和正常状态列表一样
elif s_ in [self.canvas.coords(self.hell1), self.canvas.coords(self.hell2), self.canvas.coords(self.hell3),
self.canvas.coords(self.hell4),self.canvas.coords(self.hell5),self.canvas.coords(self.hell6),
self.canvas.coords(self.hell7),self.canvas.coords(self.hell8),self.canvas.coords(self.hell9),
self.canvas.coords(self.hell10),self.canvas.coords(self.hell11),self.canvas.coords(self.hell12),
self.canvas.coords(self.hell13),self.canvas.coords(self.hell14),self.canvas.coords(self.hell15),
self.canvas.coords(self.hell16),self.canvas.coords(self.hell17),self.canvas.coords(self.hell18),
self.canvas.coords(self.hell19),self.canvas.coords(self.hell20),self.canvas.coords(self.hell21),
self.canvas.coords(self.hell22)]:
reward = -1
done = True
s_ = [1,1,1,1] # 一个特殊的列表，代表Terminal
else:
reward = 0
done = False
return s_, reward, done # s_是一个list，里面包含四个角点的坐标，reward是int，done用以结束当前episode
def render(self):
time.sleep(0.1)
self.update()
def update():
for t in range(10):
s = env.reset()
while True:
env.render()
a = 1
s, r, done = env.step(a)
if done:
break
# if __name__ == '__main__':
# env = Maze()
# env.after(100, update)
# env.mainloop()
#### 强化学习主体代码 ####
class QLearningTable:
num = 0
epsilion_step = 0.05
def __init__(self, actions, n_features, learning_rate=0.1, e_greedy=0.5, memory_size=5000):
self.actions = actions # a list
self.n_features = n_features
self.lr = learning_rate
# self.gamma = reward_decay
# self.gamma_step = discount_rate_step
self.epsilon = e_greedy
self.memory_size = memory_size
self.q_table = pd.DataFrame(columns=self.actions, dtype=np.float64) # q_table初始化,q_table是逐步扩充的，每探索到一个新的状态，就会增加一行
# initialize zero memory [s, a, r, s_]
self.memory = np.zeros((self.memory_size, n_features * 2 + 3)) # memory是一个二维数组
def store_transition(self, s, a, r, s_, gamma): # 传进来的s和s_是列表
self.num += 1
if not hasattr(self, 'memory_counter'):
self.memory_counter = 0
error = self.error(str(s), a, r, str(s_), gamma)
transition = np.hstack((s, [a, r], s_,[error])) # 将多个列表组合成一个列表[s,a,r,s_,error] 共11个元素
# replace the old memory with new memory
if self.memory_counter < self.memory_size:
index = self.memory_counter % self.memory_size
self.memory[index, :] = transition # 当memory没有满时，正常顺序填充
else:
# index = self.memory_counter % self.memory_size
self.memory[self.memory_size-1, :] = transition # 当memory满时，每次固定替换最后的元素
#if (self.num > 200) and (self.num % 5 == 0): # 这样可以不必每次存储时都排序，浪费时间，只要在每次学习前排序就好。但是当memory满时，进行替换时，并不能保证memory里的顺序是排好的
self.memory = self.memory[np.lexsort(-self.memory.T)] #按照最后一个数字（误差）排序
self.memory_counter += 1
# print("----------memory--------")
# print(self.memory)
# print("----------memory--------")
def choose_action(self, observation, episode): # 传入的是字符串：'[....]'
self.check_state_exist(observation)
if episode % 50 ==0:
self.epsilon += self.epsilion_step
self.epsilon = min(self.epsilon, 0.99)
# action selection
if np.random.uniform() < self.epsilon:
# choose best action
state_action = self.q_table.loc[observation, :]
# some actions may have the same value, randomly choose on in these actions
action = np.random.choice(state_action[state_action == np.max(state_action)].index)
else:
# choose random action
action = np.random.choice(self.actions)
return action
# 返回误差用以排序
def error(self,s, a, r, s_, gamma): # s和s_是字符串 '[....]'
self.check_state_exist(s_)
q_predict = self.q_table.loc[s, a]
if s_ != [1,1,1,1]:
q_target = r + gamma * self.q_table.loc[s_, :].max() # next state is not terminal
else:
q_target = r # next state is terminal
# error = q_target - q_predict
error = abs(q_predict - q_target)
return error
# 根据元素顺序进行采样,越靠后的，概率越低
def sample_whole_row_from_2d_array(self,array):
rows, cols = np.array(array).shape
probabilities = 1 / np.arange(1, rows + 1)
probabilities /= np.sum(probabilities)
sampled_row = np.random.choice(np.arange(rows), p=probabilities)
sampled_whole_row = array[sampled_row].copy()
prob = probabilities[sampled_row].copy()
#print('****************',probabilities)
return sampled_whole_row,prob
def learn(self, gamma, lr_start, beta):
# sample batch memory from all memory
#if self.memory_counter > self.memory_size:
#sample_index = np.random.choice(self.memory_size, size=1)########self.batch_size
sampled_elements, prob = self.sample_whole_row_from_2d_array(self.memory)
sampled_elements = list(sampled_elements) # 数组变列表，是为了让从memory提取出来的元素与q_table中的状态列表类型一致，'[....]',避免计算Q值索引q_table时出现错误
#else:
#sample_index = np.random.choice(self.memory_counter, size=1)#######self.batch_size
#batch_memory = self.memory[sample_index, :]
self.lr = lr_start / math.pow((self.memory_size * prob), beta) # 随着采样概率的增加，学习率降低
s = str(sampled_elements[:4])
a = sampled_elements[4]
r = sampled_elements[5]
s_ = str(sampled_elements[6:10])
self.check_state_exist(s_)
q_predict = self.q_table.loc[s, a]
if s_ != [1,1,1,1]:
q_target = r + gamma * self.q_table.loc[s_, :].max() # next state is not terminal
else:
q_target = r # next state is terminal
self.q_table.loc[s, a] += self.lr * (q_target - q_predict) # update
#self.q_table.loc[s, a] += 1 - (q_target * self.lr) # update
def check_state_exist(self, state):
if state not in self.q_table.index:
# append new state to q table
self.q_table = self.q_table.append(
pd.Series(
[0]*len(self.actions),
index=self.q_table.columns,
name=state,
)
)
print(self.q_table)
#print(type(self.q_table))
#### 智能体学习实现 ####
def update():
step = 0
gamma_step = 0.05
gamma = 0
lr_start = 0.5
beta = 0.3
for episode in range(500):
# initial observation
observation = env.reset()
reward_sum = 0
while True:
# fresh env
env.render()
# RL choose action based on observation
action = RL.choose_action(str(observation), episode)
# RL take action and get next observation and reward
observation_, reward, done = env.step(action,episode)
reward_sum += reward
RL.store_transition(observation, action, reward, observation_, gamma)
#if (step > 50) and (step % 10 == 0):
if (step > 20):
# RL learn from this transition
#RL.learn(str(observation), action, reward, str(observation_), 1)
RL.learn(gamma, lr_start, beta)
# swap observation
observation = observation_
# break while loop when end of this episode
if done:
break
step += 1
# if step % 50 ==0:
# print("----------step-------------")
# print(step)
# print("-------------step---------")
# time.sleep(3)
# 增加折扣率
if episode % 25 ==0: # 最好随着episode的成倍增加而成倍增加
gamma += gamma_step
gamma = min(gamma, 0.9)
if episode % 11 ==0:
print("------------------")
print(episode,reward_sum)
print("------------------")
# end of game
print('game over')
#env.destroy()
if __name__ == "__main__":
env = Maze()
RL = QLearningTable(actions=list(range(env.n_actions)), n_features=env.n_features)
env.after(100, update)
env.mainloop()