def getReadingDistributionGivenObservations(self, observations,
newLocation):
""" Given a current set of observations and a location, it computes the
probability of each possibly sensor reading at this position given other sensor readings.
Thus the return value is a mapping from sensor readings to probabilities. """
# QUESTION 3
if (observations.has_key(newLocation) > 0):
#print "WEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE"
#print newLocation, observations
dist = util.listToDistribution([observations[newLocation]])
#print dist
#print "WEE"
return dist
readingDistribution = util.Counter()
from ghostbusters import Readings
currentSetProbability = self.getReadingSetProbability(observations)
for reading in Readings.getReadings():
newObservations = observations.copy()
newObservations[newLocation] = reading
readingDistribution[reading] = self.getReadingSetProbability(
newObservations) / currentSetProbability
return readingDistribution
def observe(self, observation, emissionFn):
# update beliefs
# P(X_t | e_1:t, e') ~ P(e'|X_t) * P(X_t | e_1:t)
for particle in self.particles.keys():
# reweight each particle by observations
self.particles[particle] *= emissionFn(observation, particle)
# resample observation
self.particles.normalize()
samples = util.sampleMultiple(self.particles, self.numParticles)
self.particles = util.listToDistribution(samples)
def observe(self, observation, emissionFn):
# update beliefs
# P(X_t | e_1:t, e') ~ P(e'|X_t) * P(X_t | e_1:t)
for particle in self.particles.keys():
# reweight each particle by observations
self.particles[particle] *= emissionFn(observation, particle)
# resample observation
self.particles.normalize()
samples = util.sampleMultiple(self.particles, self.numParticles)
self.particles = util.listToDistribution(samples)
def getBeliefDistribution(self):
"""
Return the agent's current belief (approximation) as a distribution
over ghost tuples. Note that this distribution can and should be
sparse in the sense that many tuples may not be represented in the
distribution if there are more tuples than particles. The probability
over these missing tuples will be treated as zero by the GUI.
"""
"*** YOUR CODE HERE ***"
return util.listToDistribution(self.particles)
def getBeliefDistribution(self):
"""
Return the agent's current belief (approximation) as a distribution
over ghost tuples. Note that this distribution can and should be
sparse in the sense that many tuples may not be represented in the
distribution if there are more tuples than particles. The probability
over these missing tuples will be treated as zero by the GUI.
"""
"*** YOUR CODE HERE ***"
return util.listToDistribution(self.particles)
def getReadingDistributionGivenObservations(self, observations, newLocation):
""" Given a current set of observations and a location, it computes the
probability of each possibly sensor reading at this position given other sensor readings.
Thus the return value is a mapping from sensor readings to probabilities. """
# QUESTION 3
if (observations.has_key(newLocation) > 0):
#print "WEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE"
#print newLocation, observations
dist = util.listToDistribution([observations[newLocation]])
#print dist
#print "WEE"
return dist
readingDistribution = util.Counter()
from ghostbusters import Readings
currentSetProbability = self.getReadingSetProbability(observations)
for reading in Readings.getReadings():
newObservations = observations.copy()
newObservations[newLocation] = reading
readingDistribution[reading] = self.getReadingSetProbability(newObservations) / currentSetProbability
return readingDistribution
