-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy path02_iterativ_policy_evaluation.py
184 lines (168 loc) · 6.77 KB
/
02_iterativ_policy_evaluation.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
import numpy as np
import sys
import math
sys.path.insert(0, './simplistic_world')
from SimpleGridEnvironment import *
from SimpleGridAction import OnlyMoveAction
#variation defined in Satton/Barto Book
def iterativ_policy_eval_satton_barto(grid, thr=0.001, prob=0.25, iter=1000):
'''
V(s) of start/end states are 0.0
For performing invalid states we take V(s) and Reward defined in rewards
'''
V = {}
states = grid.get_states()
for s in states:
V[s] = 0.0
for k in range(0, iter):
print('-------------- Iteration = %s ---------------' % k)
print('Old Value function for this iteration = %s' %V )
delta = 0
V_old = V.copy()
for s in states:
if s == grid.start_cell or s == grid.finish_cell:
continue
possible_actions = grid.actions[s]
v_new = 0
for a in possible_actions:
s_prime = grid.next_state(s,a)
v_new += V_old[s_prime] - 1.0
for a in range(4 - len(possible_actions)):
v_new +=(V_old[s]-1.0)
V[s] = v_new*prob
delta = max(delta, abs(V[s]- V_old[s]))
if delta < thr:
print('Delta = ', delta)
break
print('--Final value function --')
for s in states:
print('On state = %s, value func is= %s' % (s, V[s]))
def iterativ_policy_eval(grid, thr=0.001, prob=0.25, gamma=1.0, iter=1000):
'''
V(s) of start/end states are 0.0
For performing invalid states we take V(s) and Reward defined in rewards
'''
V = {}
states = grid.get_states()
for s in states:
V[s] = 0.0
for k in range(0, iter):
print('-------------- Iteration = %s ---------------' % k)
print('Old Value function for this iteration = %s' %V )
delta = 0
#V_old = V.copy()
for s in states:
v_old = V[s]
if s == grid.start_cell or s == grid.finish_cell:
continue
v_new = 0
possible_actions = grid.actions[s]
for a in possible_actions:
print('Current action = ', a)
grid.set_state(s)
reward = grid.perform_action(a)
print('I will get reward = ', reward)
print('Current state=', (grid.posX, grid.posY))
#s_prime = grid.next_state(s,a)
current_state = (grid.posX, grid.posY)
v_new += prob * (reward + gamma*V[current_state])
V[s] = v_new
delta = max(delta, abs(V[s]- v_old))
if delta < thr:
print('Delta = ', delta)
break
print('--Final value function --')
for s in states:
print('On state = %s, value func is= %s' % (s, V[s]))
def iterativ_policy_eval_with_policy(grid, policy, thr=0.001, gamma=1.0, iter=1000):
'''
V(s) of start/end states are 0.0
'''
V = {}
states = grid.get_states()
for s in states:
V[s] = 0.0
for k in range(0, iter):
print('-------------- Iteration = %s ---------------' % k)
print('Old Value function for this iteration = %s' %V )
delta = 0
for s in states:
if s in policy:
v_old = V[s]
if s == grid.start_cell or s == grid.finish_cell:
continue
grid.set_state(s)
a = policy[s]
print('Current action = ', a)
reward = grid.perform_action(a)
current_state = (grid.posX, grid.posY)
V[s] = reward + gamma*V[current_state]
delta = max(delta, abs(V[s]- v_old))
if delta < thr:
print('Delta = ', delta)
break
print('--Final value function --')
for s in states:
print('On state = %s, value func is= %s' % (s, V[s]))
if __name__ == '__main__':
print('Creating a grid environment first')
env = SimpleGridEnvironment([0,0], [3,3], 4)
actions = {(0,0): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT],
(0,1): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP],
(0,2): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP],
(0,3): [OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP],
(1,0): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_LEFT],
(1,1): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP, OnlyMoveAction.MOVE_LEFT],
(1,2): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP, OnlyMoveAction.MOVE_LEFT],
(1,3): [OnlyMoveAction.MOVE_LEFT, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP],
(2,0): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_LEFT],
(2,1): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP, OnlyMoveAction.MOVE_LEFT],
(2,2): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP, OnlyMoveAction.MOVE_LEFT],
(2,3): [OnlyMoveAction.MOVE_LEFT, OnlyMoveAction.MOVE_RIGHT, OnlyMoveAction.MOVE_UP],
(3,0): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_LEFT],
(3,1): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_LEFT, OnlyMoveAction.MOVE_UP],
(3,2): [OnlyMoveAction.MOVE_DOWN, OnlyMoveAction.MOVE_LEFT, OnlyMoveAction.MOVE_UP],
(3,3): [OnlyMoveAction.MOVE_LEFT, OnlyMoveAction.MOVE_UP]
}
rewards = {(0,0):0.0,
(0,1):-1.0,
(0,2):-1.0,
(0,3):-1.0,
(1,0):-1.0,
(1,1):-1.0,
(1,2):-1.0,
(1,3):-1.0,
(2,0):-1.0,
(2,1):-1.0,
(2,2):-1.0,
(2,3):-1.0,
(3,0):-1.0,
(3,1):-1.0,
(3,2):-1.0,
(3,3): 0.0
}
env.set_rewards(rewards)
env.set_actions(actions)
env2 = SimpleGridEnvironment([2,0], [0,3], 4)
rewards = {(0,0): 0.0,
(0,1): 0.0,
(0,2): 0.0,
(0,3): 1.0,
(1,0): 0.0,
(1,1): 0.0,
(1,2): 0.0,
(1,3):-1.0,
(2,0): 0.0,
(2,1): 0.0,
(2,2): 0.0,
(2,3): 0.0,
(3,0): 0.0,
(3,1): 0.0,
(3,2): 0.0,
(3,3): 0.0
}
env2.set_rewards(rewards2)
env2.set_actions(actions)
print(env.get_states())
#iterativ_policy_eval(actions, rewards)
iterativ_policy_eval(env)