Repository navigation
Expand file tree
/
Copy pathenv.py
More file actions
81 lines (66 loc) · 2.47 KB
/
Copy pathenv.py
File metadata and controls
81 lines (66 loc) · 2.47 KB
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
import numpy as np
#########
# |--------> y
# |
# |
# |
# v
# x
#########
class GridWorld():
def __init__(self, size=5, forbidden_states=None, target_state=None,
gamma=0.9, boundary_penalty=-1, forbidden_penalty=-1, target_reward=1):
self.size = size
self.grid = np.zeros((self.size, self.size))
if forbidden_states is not None:
self.forbidden_states = forbidden_states
else:
print("No forbidden states set!")
if target_state is not None:
self.target_state = target_state
else:
print("No target state set!")
self.gamma = gamma
self.boundary_penalty = boundary_penalty
self.forbidden_penalty = forbidden_penalty
self.target_reward = target_reward
self.action_number = 5
self.action_value = np.zeros((self.size, self.size, self.action_number), dtype=np.float64)
self.state_value = np.zeros((self.size, self.size), dtype=np.float64)
# 策略
self.pi = np.zeros((self.size, self.size, self.action_number), dtype=np.float64)
# 最优策略
self.pi_star = np.zeros((self.size, self.size), dtype=np.float64)
self.reset()
def reset(self):
self.action_value = np.zeros((self.size, self.size, self.action_number), dtype=np.float64)
self.state_value = np.zeros((self.size, self.size), dtype=np.float64)
self.pi = np.ones((self.size, self.size, self.action_number), dtype=np.float64) / self.action_number
self.pi_star = np.zeros((self.size, self.size), dtype=np.float64)
def action_space(self):
return ['up', 'right', 'down', 'left', 'stay']
def not_in_boundary(self, s):
return 0 <= s[0] < self.size and 0 <= s[1] < self.size
def take_action(self, s1, a):
r = 0
if a == 0: # up
s2 = (s1[0] - 1, s1[1])
if a == 1: # right
s2 = (s1[0], s1[1] + 1)
if a == 2: # down
s2 = (s1[0] + 1, s1[1])
if a == 3: # left
s2 = (s1[0], s1[1] - 1)
if a == 4: # stay
s2 = s1
# 边界检查
if not self.not_in_boundary(s2):
s2 = s1
r += self.boundary_penalty
else:
# 障碍检查
if s2 in self.forbidden_states:
r += self.forbidden_penalty
if s2 == self.target_state:
r += self.target_reward
return s2, r