/
wumpus.py
160 lines (121 loc) · 4.69 KB
/
wumpus.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
# -*- coding: utf-8 -*-
from simpleai.machine_learning.reinforcement_learning import TDQLearner, \
make_exponential_temperature, \
PerformanceCounter, \
RLProblem
from simpleai.environments import RLEnvironment
from Tkinter import *
import threading
class PerpetualTimer(threading._Timer):
def __init__(self, interval, function, name=None, daemon=False,
args=(), kwargs={}):
super(PerpetualTimer, self).__init__(interval, function, args, kwargs)
self.setName(name)
self.setDaemon(daemon)
def run(self):
while True:
self.finished.wait(self.interval)
if self.finished.isSet():
return
self.function(*self.args, **self.kwargs)
def stop(self, timeout=None):
self.cancel()
self.join(timeout)
def callback(egg):
egg.cook()
class WumpusViewer(object):
def __init__(self, environment):
self.environment = environment
root = Tk()
w = Canvas(root, width=400, height=400)
self.w = w
w.pack()
w.create_polygon(0, 0, 0, 400, 400, 400, 400, 0, fill='white')
for x in range(4):
w.create_line(100 * x, 0, 100 * x, 400)
w.create_line(0, 100 * x, 400, 100 * x)
#make holes
for x, y in environment.holes:
w.create_oval(*self._coord(x, y), fill='red')
#make wumpus
x, y = environment.wumpus
w.create_oval(*self._coord(x, y), fill='blue')
#make agent
x, y, have_gold = environment.state
self.agent = w.create_oval(*self._coord(x, y), fill='green')
#make gold
x, y = environment.gold
self.gold = w.create_oval(*self._coord(x, y), fill='yellow')
frame = Frame(root)
frame.pack()
self.start = Button(frame, text="Start", command=self.start)
self.start.pack(side=LEFT)
self.stop = Button(frame, text="Stop", command=self.stop)
self.stop.pack(side=LEFT)
root.mainloop()
def _coord(self, x, y):
y = 5 - y
cx = 100 * (x - 1) + 50
cy = 100 * (y - 1) + 50
r = 30
return (cx - r, cy - r, cx + r, cy + r)
def step(self):
if self.environment.is_completed(self.environment.state):
self.timer.cancel()
else:
self.environment.step(viewer=self)
def start(self):
self.timer = PerpetualTimer(0.5, self.step)
self.timer.start()
self.environment.state = self.environment.initial_state
def stop(self):
self.timer.cancel()
def event(self, state1, action, state2, agent):
print 'action: %s state: %s' % (action, str(state2))
x, y, have_gold = state2
if have_gold and self.gold:
self.w.delete(self.gold)
self.w.itemconfig(self.agent, outline="yellow", width=6.0)
self.w.coords(self.agent, self._coord(x, y))
class WumpusEnvironment(RLEnvironment):
def __init__(self, agent):
super(RLEnvironment, self).__init__([agent], (1, 1, False))
self.action_dict = {'up': (0, 1), 'down': (0, -1), 'left': (-1, 0), 'rigth': (1, 0)}
self.wumpus = (1, 3)
self.holes = [(3, 1), (3, 3), (4, 4)]
self.gold = (3, 4)
self.threats = [self.wumpus] + self.holes
self.rewards = {(1, 1, True): 10}
for c, r in self.threats:
self.rewards[(c, r, True)] = -10
self.rewards[(c, r, False)] = -10
def do_action(self, state, action, agent):
c1, r1, have_gold = state
c2, r2 = self.action_dict[action]
rn = r1 + r2
cn = c1 + c2
if not have_gold and (cn, rn) == self.gold:
have_gold = True
_next = (cn, rn, have_gold)
if (1 <= rn <= 4) and (1 <= cn <= 4):
return _next
return state
def is_completed(self, state):
return state in self.rewards.keys()
def reward(self, state, agent):
return self.rewards.get(state, -0.08)
class WumpusProblem(RLProblem):
def actions(self, state):
actions = ['up', 'down', 'left', 'rigth']
return actions
if __name__ == '__main__':
agent = TDQLearner(WumpusProblem(),
temperature_function=make_exponential_temperature(1000, 0.01),
discount_factor=0.8)
game = WumpusEnvironment(agent)
p = PerformanceCounter([agent], ['Q-learner Agent'])
print 'Training...'
for i in range(10000):
game.run()
p.show_statistics()
game.run(viewer=WumpusViewer(game))