def getAction(self, rawState, epsilon):
legalActions = rawState.getLegalActions()
legalActions.remove(Directions.STOP)
if util.flipCoin(epsilon):
return random.choice(legalActions)
else:
qValues = [(Directions.getIndex(action),
self.getQValue(rawState, action))
for action in legalActions]
qValues = sorted(qValues, key=lambda x: x[1], reverse=True)
for index, qValue in qValues:
action = Directions.fromIndex(index)
if action in legalActions:
return action
def getAction(self, rawState, epsilon):
legalActions = rawState.getLegalActions()
legalActions.remove(Directions.STOP)
qState = self.trainingRoom.featuresExtractor.getFeatures(
rawState, None)
if util.flipCoin(epsilon):
return random.choice(legalActions)
else:
qValues = list(
enumerate(self.model.model.predict(np.array([qState]))[0]))
qValues = sorted(qValues, key=lambda x: x[1], reverse=True)
for index, qValue in qValues:
action = Directions.fromIndex(index)
if action in legalActions:
return action
def chooseNextCrossroad(crossroads, ghostsPos, pacmanPos, distMap, walls):
sortedCrossroads = []
sorted = False
while not sorted:
if len(sortedCrossroads) < len(crossroads):
closest = random.choice(crossroads)
for crossroad in crossroads:
if distMap[crossroad, pacmanPos] < distMap[closest, pacmanPos] and crossroad not in sortedCrossroads:
closest = crossroad
sortedCrossroads.append(closest)
else:
sorted = True
for i in range(0, len(sortedCrossroads)):
reachable = util.accessibleAStar(pacmanPos, sortedCrossroads[i], ghostsPos, walls)
if reachable:
return sortedCrossroads[i]
return sortedCrossroads[len(sortedCrossroads) - 1]
def getAction(self, state):
return random.choice(state.getLegalActions(self.index))
def getAction(self, state):
import random, time
time.sleep(2.0)
return random.choice(state.getLegalActions(self.index))