def draw_state(self, state, treward, creward):
# declarations
apos = [None] * NUM_AGENTS
# clear cells
for i in range(len(self.cells)):
self.cells[i].removeAll()
# get agents' positions
for i in range(NUM_AGENTS):
apos[i] = ROWS * COLS - state[i]
# last element in the array is wumpus' position
wpos = ROWS * COLS - state[NUM_AGENTS]
# add agents
for i in range(NUM_AGENTS):
self.cells[apos[i]].add(
JLabel(self.scale(self.aIcon.getImage(), 0.6)))
# add the wumpus
self.cells[wpos].add(JLabel(self.scale(self.wIcon.getImage(), 0.6)))
# gray-out illegal positions and update all cells
for i in range(len(self.cells)):
if gp.isLegalPosition(i) == False:
self.cells[i].setBackground(Color.BLACK)
self.cells[i].updateUI()
# display total and cum rew on ctrl panel
self.rewLbl.setText(Double.toString(treward))
self.crewLbl.setText(Double.toString(creward))
def update(self, ss, when):
strLine = Long.toString(when.getTimeInMillis()).encode('utf-8')
for s in bricSymbols:
if s in ss:
strLine = strLine + ',' \
+ Double.toString(market.getLastPrice(0,s+'-OPEN'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,s+'-HIGH'))\
.encode('utf-8') + ',' \
+ Double.toString(market.getLastPrice(0,s+'-LOW'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,s+'-CLOSE'))\
.encode('utf-8') + ','\
+ Long.toString(market.getLastVolume(0,s))\
.encode('utf-8')
else:
strLine = strLine + ',-,-,-,-,-'
if 'XAG' in ss:
strLine = strLine + ',' \
+ Double.toString(market.getLastPrice(0,'XAG-OPEN'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,'XAG-HIGH'))\
.encode('utf-8') + ',' \
+ Double.toString(market.getLastPrice(0,'XAG-LOW'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,'XAG-CLOSE'))\
.encode('utf-8') + ','\
+ Long.toString(market.getLastVolume(0,'XAG'))\
.encode('utf-8')
else:
strLine = strLine + ',-,-,-,-,-'
print strLine
def update(self, ss, when):
strLine = Long.toString(when.getTimeInMillis()).encode('utf-8')
strLine = strLine + ',' \
+ Double.toString(market.getLastPrice(0,'XAG-OPEN'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,'XAG-HIGH'))\
.encode('utf-8') + ',' \
+ Double.toString(market.getLastPrice(0,'XAG-LOW'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,'XAG-CLOSE'))\
.encode('utf-8') + ','\
+ Long.toString(market.getLastVolume(0,'XAG'))\
.encode('utf-8')
print strLine
def update(self, ss, when):
strLine = Long.toString(when.getTimeInMillis()).encode('utf-8')
if 'X' in ss:
strLine = strLine + ',' \
+ Double.toString(market.getLastPrice(0,'X-OPEN'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,'X-HIGH'))\
.encode('utf-8') + ',' \
+ Double.toString(market.getLastPrice(0,'X-LOW'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,'X-CLOSE'))\
.encode('utf-8') + ','\
+ Long.toString(market.getLastVolume(0,'X'))\
.encode('utf-8')
else:
strLine = strLine + ',-,-,-,-,-'
if 'Y' in ss:
strLine = strLine + ',' \
+ Double.toString(market.getLastPrice(0,'Y-OPEN'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,'Y-HIGH'))\
.encode('utf-8') + ',' \
+ Double.toString(market.getLastPrice(0,'Y-LOW'))\
.encode('utf-8') + ','\
+ Double.toString(market.getLastPrice(0,'Y-CLOSE'))\
.encode('utf-8') + ','\
+ Long.toString(market.getLastVolume(0,'Y'))\
.encode('utf-8')
else:
strLine = strLine + ',-,-,-,-,-'
print strLine
def draw_state(self, state, treward, creward):
# declarations
apos = [None] * NUM_AGENTS
# clear cells
for i in range(len(self.cells)):
self.cells[i].removeAll()
# get agents' positions
for i in range(NUM_AGENTS):
apos[i] = ROWS * COLS - state[i]
# last element in the array is wumpus' position
wpos = ROWS * COLS - state[NUM_AGENTS]
# add agents
for i in range(NUM_AGENTS):
self.cells[apos[i]].add(JLabel(self.scale(self.aIcon.getImage(), 0.6)))
# add the wumpus
self.cells[wpos].add(JLabel(self.scale(self.wIcon.getImage(), 0.6)))
# gray-out illegal positions and update all cells
for i in range(len(self.cells)):
if gp.isLegalPosition(i) == False:
self.cells[i].setBackground(Color.BLACK)
self.cells[i].updateUI()
# display total and cum rew on ctrl panel
self.rewLbl.setText(Double.toString(treward))
self.crewLbl.setText(Double.toString(creward))
def __str__(self): output = "" for i in range(self.theMatrix.getRowDimension()): if i == 0: output = output +"[" for j in range(self.theMatrix.getColumnDimension()): if j == 0: output = output +"[" output = output + Double.toString(self.theMatrix.get(i,j)) if (j == (self.theMatrix.getColumnDimension() -1)): output = output +"]" else: output = output + ", " if (i == (self.theMatrix.getRowDimension() -1)): output = output +"]" else: output = output + ", " return output
def __str__(self):
output = ""
for i in range(self.theMatrix.getRowDimension()):
if i == 0:
output = output + "["
for j in range(self.theMatrix.getColumnDimension()):
if j == 0:
output = output + "["
output = output + Double.toString(self.theMatrix.get(i, j))
if (j == (self.theMatrix.getColumnDimension() - 1)):
output = output + "]"
else:
output = output + ", "
if (i == (self.theMatrix.getRowDimension() - 1)):
output = output + "]"
else:
output = output + ", "
return output
def run_sim(self):
episoderunning = True
# starting belief state
currbelief = initBelief
factoredS = initState
instance = 1
cumrew = 0
run = 1
# execution loop
# while(episoderunning):
while (run <= 100):
# draw current state
# self.draw_state(factoredS)
# extract action from direct controller - ie., 0-step LA
exreward = valueFunction.V(currbelief)
action = valueFunction.directControl(currbelief)
print "value of b %f and selected action %d" % (exreward, action)
# show action that we are about to execute
self.nactLbl.setText(pomdpProblem.getactStr(action))
time.sleep(1)
# sample new state, observation, and calculate reward
print factoredS
factoredS1 = pomdpProblem.sampleNextState(factoredS, action)
print " sampled state is"
print factoredS1
factoredO = pomdpProblem.sampleObservation(factoredS, factoredS1,
action)
print " sampled o is"
print factoredO
reward = pomdpProblem.getReward(factoredS, action)
print reward
# draw new state, and display what's happening
self.draw_state(factoredS1)
self.lactLbl.setText(pomdpProblem.getactStr(action))
self.nactLbl.setText("nil")
self.rewLbl.setText(Double.toString(reward))
cumrew = cumrew + reward * gamma**instance
self.crewLbl.setText(Double.toString(cumrew))
self.obsLbl.setText(
pomdpProblem.getobsStr(
pomdpProblem.sencode(factoredO, pomdpProblem.getnrObsV(),
pomdpProblem.getobsArity()) - 1))
# iterate
nextbelief = pomdpProblem.tao(
currbelief, action,
pomdpProblem.sencode(factoredO, pomdpProblem.getnrObsV(),
pomdpProblem.getobsArity()) - 1)
currbelief = nextbelief
factoredS = factoredS1
instance = instance + 1
# step button
if self.stepChk.isSelected() == True:
self.stepBtn.setEnabled(True)
while self.stepBtn.isEnabled() == True:
time.sleep(1)
# check whether this episode has ended
#if reward == CATCH_REWARD - 1:
# print "Episode ended!"
# episoderunning = false
# smooth the sim a little in case we're not stepping
time.sleep(1)
# stop after 100 iterations
if instance == 100:
run = run + 1
stats.append(cumrew)
currbelief = initBelief
factoredS = initState
cumrew = 0
instance = 1
print stats
def run_sim(self):
episoderunning = True
# starting belief state
currbelief = initBelief
factoredS = initState
instance = 1
cumrew = 0
run = 1
# execution loop
# while(episoderunning):
while run <= 100:
# draw current state
# self.draw_state(factoredS)
# extract action from direct controller - ie., 0-step LA
exreward = valueFunction.V(currbelief)
action = valueFunction.directControl(currbelief)
print "value of b %f and selected action %d" % (exreward, action)
# show action that we are about to execute
self.nactLbl.setText(pomdpProblem.getactStr(action))
time.sleep(1)
# sample new state, observation, and calculate reward
print factoredS
factoredS1 = pomdpProblem.sampleNextState(factoredS, action)
print " sampled state is"
print factoredS1
factoredO = pomdpProblem.sampleObservation(factoredS, factoredS1, action)
print " sampled o is"
print factoredO
reward = pomdpProblem.getReward(factoredS, action)
print reward
# draw new state, and display what's happening
self.draw_state(factoredS1)
self.lactLbl.setText(pomdpProblem.getactStr(action))
self.nactLbl.setText("nil")
self.rewLbl.setText(Double.toString(reward))
cumrew = cumrew + reward * gamma ** instance
self.crewLbl.setText(Double.toString(cumrew))
self.obsLbl.setText(
pomdpProblem.getobsStr(
pomdpProblem.sencode(factoredO, pomdpProblem.getnrObsV(), pomdpProblem.getobsArity()) - 1
)
)
# iterate
nextbelief = pomdpProblem.tao(
currbelief,
action,
pomdpProblem.sencode(factoredO, pomdpProblem.getnrObsV(), pomdpProblem.getobsArity()) - 1,
)
currbelief = nextbelief
factoredS = factoredS1
instance = instance + 1
# step button
if self.stepChk.isSelected() == True:
self.stepBtn.setEnabled(True)
while self.stepBtn.isEnabled() == True:
time.sleep(1)
# check whether this episode has ended
# if reward == CATCH_REWARD - 1:
# print "Episode ended!"
# episoderunning = false
# smooth the sim a little in case we're not stepping
time.sleep(1)
# stop after 100 iterations
if instance == 100:
run = run + 1
stats.append(cumrew)
currbelief = initBelief
factoredS = initState
cumrew = 0
instance = 1
print stats
