def valueIteration(defaultReward):
discountedValue = 0.9
from World import World
instance = World()
instance.default_Reward = defaultReward
# print instance.isWalls(3,2)
# old actions ={'right':[0.8,0.2],'left':[1.0],'up':[0.8,0.2],'down':[1.0]}
actions = {
"right": {"right": 0.8, "down": 0.2},
"left": {"left": 1.0},
"up": {"up": 0.8, "left": 0.2},
"down": {"down": 1.0},
}
# initialize the value
valueGrid = [[0 for x in range(instance.world_Column)] for x in range(instance.world_Row)]
previousValueGrid = [[0 for x in range(instance.world_Column)] for x in range(instance.world_Row)]
iterations = 0
stop = False
while not stop:
iterations += 1
previousValueGrid = copyMatrix(valueGrid, instance.world_Row, instance.world_Column)
for row in range(instance.world_Row):
for col in range(instance.world_Column):
# for all states
# for all actions
valueActions = [0, 0, 0, 0]
count = 0
if not instance.isWalls(row, col):
for key, pairs in actions.iteritems():
total = 0.0
for action, value in pairs.iteritems():
if instance.isWithinWorld(action, row, col):
newCoordinates = instance.newPosition(action, row, col)
total += value * valueGrid[newCoordinates[0]][newCoordinates[1]]
valueActions[count] = instance.getRewards(row, col) + (discountedValue * total)
count += 1
valueGrid[row][col] = max(valueActions)
# print valueGrid
stop = convergence(valueGrid, previousValueGrid, instance.world_Row, instance.world_Column)
print valueGrid
print "The number of iterations is " + str(iterations)
def valueIteration(defaultReward):
discountedValue = 0.9
from World import World
instance = World()
instance.default_Reward = defaultReward
#print instance.isWalls(3,2)
# old actions ={'right':[0.8,0.2],'left':[1.0],'up':[0.8,0.2],'down':[1.0]}
actions = {'right':{'right':0.8,'down':0.2},'left':{'left':1.0},'up':{'up':0.8,'left':0.2},'down':{'down':1.0}}
#initialize the value
valueGrid =[[0 for x in range(instance.world_Column)] for x in range(instance.world_Row)]
previousValueGrid =[[0 for x in range(instance.world_Column)] for x in range(instance.world_Row)]
iterations = 0
stop = False
while not stop :
iterations +=1
previousValueGrid = copyMatrix(valueGrid,instance.world_Row,instance.world_Column)
for row in range(instance.world_Row):
for col in range(instance.world_Column):
#for all states
#for all actions
valueActions=[0,0,0,0]
count = 0
if not instance.isWalls(row,col):
for key,pairs in actions.iteritems():
total =0.0
for action,value in pairs.iteritems():
if instance.isWithinWorld(action,row,col):
newCoordinates = instance.newPosition(action,row,col)
total += (value*valueGrid[newCoordinates[0]][newCoordinates[1]])
valueActions[count] = instance.getRewards(row,col) + (discountedValue * total)
count +=1
valueGrid[row][col] = max(valueActions)
#print valueGrid
stop = convergence(valueGrid,previousValueGrid,instance.world_Row,instance.world_Column)
print valueGrid
print "The number of iterations is "+str(iterations)
