-
Notifications
You must be signed in to change notification settings - Fork 0
/
valueIterationAgents.py
155 lines (122 loc) · 5.13 KB
/
valueIterationAgents.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
# valueIterationAgents.py
# -----------------------
# Licensing Information: Please do not distribute or publish solutions to this
# project. You are free to use and extend these projects for educational
# purposes. The Pacman AI projects were developed at UC Berkeley, primarily by
# John DeNero (denero@cs.berkeley.edu) and Dan Klein (klein@cs.berkeley.edu).
# For more info, see http://inst.eecs.berkeley.edu/~cs188/sp09/pacman.html
import mdp, util
from learningAgents import ValueEstimationAgent
class ValueIterationAgent(ValueEstimationAgent):
"""
* Please read learningAgents.py before reading this.*
A ValueIterationAgent takes a Markov decision process
(see mdp.py) on initialization and runs value iteration
for a given number of iterations using the supplied
discount factor.
"""
def __init__(self, mdp, discount = 0.9, iterations = 100):
"""
Your value iteration agent should take an mdp on
construction, run the indicated number of iterations
and then act according to the resulting policy.
Some useful mdp methods you will use:
mdp.getStates()
mdp.getPossibleActions(state)
mdp.getTransitionStatesAndProbs(state, action)
mdp.getReward(state, action, nextState)
"""
self.mdp = mdp
self.discount = discount
self.iterations = iterations
self.values = util.Counter() # A Counter is a dict with default 0
"*** YOUR CODE HERE ***"
allStates = mdp.getStates()
vPrimes = util.Counter() # A Counter is a dict with default 0
iteration = 0
while iteration < iterations:
for s in allStates:
if mdp.isTerminal(s):
vPrimes[s] = mdp.getReward(s, None, s);
else:
sreward = mdp.getReward(s, None, s)
vPrimes[s] = sreward + discount * self.utilOfBestAction(mdp, s )
for s in allStates:
self.values[s] = vPrimes[s]
iteration +=1
"""Returns the value of the best action to take given that we are
at state s.
s = current state
"""
def utilOfBestAction (self, mdp, s):
possibleActions = mdp.getPossibleActions(s)
import sys
maxUtility = -sys.maxint-1 # equivalent to "-infinity"
for action in possibleActions :
transitionStatesAndProbs = mdp.getTransitionStatesAndProbs(s, action)
utilityOfAction = self.getPUsumForGivenAction (mdp, transitionStatesAndProbs, s, action)
if utilityOfAction >= maxUtility:
maxUtility = utilityOfAction
return maxUtility
""" transitionStatesAndProbs a list of (nextState, prob) pairs for a given state
returns the sum of the probabilities and utilities of each action
"""
def getPUsumForGivenAction (self, mdp, transitionStatesAndProbs,s, action):
utilityOfAction = 0
for tsAndP in transitionStatesAndProbs:
nextState = tsAndP[0]
transitionProbability = tsAndP[1]
utilityOfAction += (transitionProbability * self.values[nextState]) #
return utilityOfAction
def getValue(self, state):
"""
Return the value of the state (computed in __init__).
"""
return self.values[state]
def getQValue(self, state, action):
"""
The q-value of the state action pair
(after the indicated number of value iteration
passes). Note that value iteration does not
necessarily create this quantity and you may have
to derive it on the fly.
"""
"*** YOUR CODE HERE ***"
qvalue = 0
for (next_state, probability) in self.mdp.getTransitionStatesAndProbs(state, action):
qvalue += probability * (self.mdp.getReward(state, action, next_state) +
self.discount * self.getValue(next_state))
return qvalue
#util.raiseNotDefined()
def getPolicy(self, state):
"""
The policy is the best action in the given state
according to the values computed by value iteration.
You may break ties any way you see fit. Note that if
there are no legal actions, which is the case at the
terminal state, you should return None.
"""
"*** YOUR CODE HERE ***"
#this method looks at all the neighboring states,
#then returns the action that attempts to goto the state with the highest utility
print "getPolicy ", state
legalActions = self.mdp.getPossibleActions(state)
if len(legalActions)<1:
print "-getPolicy No legal actions"
return None
import sys
maxValue= -sys.maxint-1 #-infinity
bestAction = None
for action in legalActions:
statesReachable=self.mdp.getTransitionStatesAndProbs(state,action)
expectedUtility=0
for newstate in statesReachable:
expectedUtility=expectedUtility+newstate[1]*self.getValue(newstate[0])
if expectedUtility >= maxValue:
maxValue=expectedUtility
bestAction=action
return bestAction
#util.raiseNotDefined()
def getAction(self, state):
"Returns the policy at the state (no exploration)."
return self.getPolicy(state)