def makeObservation(self): returnObs = Observation() returnObs.doubleArray = self.pos.tolist() if self.fuel_loc is not None: returnObs.doubleArray += [self.fuel] returnObs.intArray = [self.pass_loc, self.pass_dest] return returnObs
def env_start(self):
self.reset()
returnObs = Observation()
returnObs.doubleArray = [
self.cart_location, self.cart_velocity
] + self.pole_angle.tolist() + self.pole_velocity.tolist()
return returnObs
def makeObservation(self):
returnObs = Observation()
returnObs.doubleArray = self.pos.tolist()
if self.fuel_loc is not None:
returnObs.doubleArray += [self.fuel]
returnObs.intArray = [self.pass_loc, self.pass_dest]
return returnObs
def env_start(self):
self.reset()
returnObs = Observation()
returnObs.doubleArray = (
[self.cart_location, self.cart_velocity] + self.pole_angle.tolist() + self.pole_velocity.tolist()
)
return returnObs
def env_step(self, thisAction):
log = logging.getLogger('pyrl.environments.gridworld.env_step')
episodeOver = 0
intAction = thisAction.intArray[0]
log.debug("Action to take: %d", intAction)
theReward = self.takeAction(intAction)
if self.isAtGoal():
log.info("Episode completed!!")
episodeOver = 1
if self.reward_noise > 0:
theReward += numpy.random.normal(scale=self.reward_noise)
theObs = Observation()
theObs.doubleArray = self.getState()
returnRO = Reward_observation_terminal()
returnRO.r = theReward
returnRO.o = theObs
returnRO.terminal = episodeOver
log.info("(Action - State - Reward): (%d - %s - %f)", intAction,
pformat(theObs), theReward)
return returnRO
def createObservation(self):
obs = Observation(numDoubles = self.obsSize**2)
tmp = self.vmem[
int(self.obsOrg[0]):int(self.obsOrg[0]+self.obsSize),
int(self.obsOrg[1]):int(self.obsOrg[1]+self.obsSize)
]
obs.doubleArray = list(tmp.flatten())
return obs
def env_start(self):
log = logging.getLogger('pyrl.environments.gridworld.env_start')
self.reset()
log.info("Environment started")
returnObs = Observation()
returnObs.doubleArray = self.getState()
log.debug("Observation to return: %s", pformat(returnObs))
return returnObs
def getObservation(self):
returnObs = Observation()
features = [1.]
if self.original_features:
features += mdptetris.features_original()
if self.dellacherie_features:
features += mdptetris.features_dellacherie()
returnObs.intArray = [mdptetris.current_piece()]
returnObs.doubleArray = features
return returnObs
def env_start(self):
""" Instantiate a new :class:`PinballModel` environment
:returns: The initial state
:rtype: :class:`Observation`
"""
self.pinball = PinballModel(self.configuration)
obs = Observation()
obs.doubleArray = self.pinball.get_state()
return obs
def getObservation(self):
returnObs = Observation()
features = [1.]
if self.original_features:
features += mdptetris.features_original()
if self.dellacherie_features:
features += mdptetris.features_dellacherie()
returnObs.intArray = [mdptetris.current_piece()]
returnObs.doubleArray = features
return returnObs
def env_start(self):
""" Instantiate a new :class:`PinballModel` environment
:returns: The initial state
:rtype: :class:`Observation`
"""
self.pinball = PinballModel(self.configuration)
obs = Observation()
obs.doubleArray = self.pinball.get_state()
return obs
def env_step(self,thisAction):
intAction = thisAction.intArray[0]
theReward, episodeOver = self.takeAction(intAction)
theObs = Observation()
theObs.doubleArray = self.state.tolist()
returnRO = Reward_observation_terminal()
returnRO.r = theReward
returnRO.o = theObs
returnRO.terminal = int(episodeOver)
return returnRO
def env_step(self, thisAction):
# print self.agentRow, self.agentCol
hitBoundary = self.updatePosition(thisAction.doubleArray[0])
theObs = Observation()
theObs.doubleArray = [self.agentRow, self.agentCol]
returnRO = Reward_observation_terminal()
returnRO.r = self.calculateReward(hitBoundary)
returnRO.o = theObs
returnRO.terminal = self.checkCurrentTerminal()
return returnRO
def env_step(self,thisAction): intAction = thisAction.intArray[0] obs, reward = self.takeAction(intAction) theObs = Observation() theObs.doubleArray = [obs] returnRO = Reward_observation_terminal() returnRO.r = reward returnRO.o = theObs returnRO.terminal = 0 return returnRO
def env_step(self, thisAction):
intAction = thisAction.intArray[0]
obs, reward = self.takeAction(intAction)
theObs = Observation()
theObs.doubleArray = [obs]
returnRO = Reward_observation_terminal()
returnRO.r = reward
returnRO.o = theObs
returnRO.terminal = 0
return returnRO
def env_step(self,thisAction):
# validate the action
assert len(thisAction.doubleArray)==2,"Expected 4 double actions."
self.takeAction(thisAction.doubleArray)
theObs = Observation()
theObs.doubleArray = self.getState().tolist()
theReward,terminate = self.getReward()
returnRO = Reward_observation_terminal()
returnRO.r = theReward
returnRO.o = theObs
returnRO.terminal = int(terminate)
return returnRO
def env_step(self, thisAction):
# validate the action
assert len(thisAction.doubleArray) == 2, "Expected 4 double actions."
self.takeAction(thisAction.doubleArray)
theObs = Observation()
theObs.doubleArray = self.getState().tolist()
theReward, terminate = self.getReward()
returnRO = Reward_observation_terminal()
returnRO.r = theReward
returnRO.o = theObs
returnRO.terminal = int(terminate)
return returnRO
def env_step(self, action):
""" Take a step in the environment
:param action: The action that the agent wants to take
:returns: The next state, reward and whether the current state is terminal
:rtype: :class:`Reward_observation_terminal`
"""
returnRO = Reward_observation_terminal()
returnRO.r = self.pinball.take_action(action.intArray[0])
obs = Observation()
obs.doubleArray = self.pinball.get_state()
returnRO.o = obs
returnRO.terminal = self.pinball.episode_ended()
return returnRO
def env_step(self, action):
""" Take a step in the environment
:param action: The action that the agent wants to take
:returns: The next state, reward and whether the current state is terminal
:rtype: :class:`Reward_observation_terminal`
"""
returnRO = Reward_observation_terminal()
returnRO.r = self.pinball.take_action(action.intArray[0])
obs = Observation()
obs.doubleArray = self.pinball.get_state()
returnRO.o = obs
returnRO.terminal = self.pinball.episode_ended()
return returnRO
def env_step(self, thisAction):
intAction = thisAction.intArray[0]
theReward = self.takeAction(intAction)
episodeOver = int(self.terminate())
if self.reward_noise > 0:
theReward += numpy.random.normal(scale=self.reward_noise)
theObs = Observation()
theObs.doubleArray = (
[self.cart_location, self.cart_velocity] + self.pole_angle.tolist() + self.pole_velocity.tolist()
)
returnRO = Reward_observation_terminal()
returnRO.r = theReward
returnRO.o = theObs
returnRO.terminal = episodeOver
return returnRO
def env_step(self, thisAction):
intAction = thisAction.intArray[0]
theReward = self.takeAction(intAction)
episodeOver = int(self.terminate())
if self.reward_noise > 0:
theReward += numpy.random.normal(scale=self.reward_noise)
theObs = Observation()
theObs.doubleArray = [
self.cart_location, self.cart_velocity
] + self.pole_angle.tolist() + self.pole_velocity.tolist()
returnRO = Reward_observation_terminal()
returnRO.r = theReward
returnRO.o = theObs
returnRO.terminal = episodeOver
return returnRO
def env_step(self,thisAction):
episodeOver = 0
theReward = -1.0
intAction = thisAction.intArray[0]
self.step(intAction, self.noise)
seized = 0
theReward = self.stim_penalty if intAction == 1 else 0.0
if self.getLabel(self.current_neighbor) == self.seiz_label:
theReward += self.seizure_penalty
theObs = Observation()
theObs.doubleArray = self.state.tolist()
returnRO = Reward_observation_terminal()
returnRO.r = theReward
returnRO.o = theObs
returnRO.terminal = 0
return returnRO
def env_step(self,thisAction):
episodeOver = 0
intAction = thisAction.intArray[0]
theReward = self.takeAction(intAction)
if self.isAtGoal():
episodeOver = 1
if self.reward_noise > 0:
theReward += numpy.random.normal(scale=self.reward_noise)
theObs = Observation()
theObs.doubleArray = self.getState()
returnRO = Reward_observation_terminal()
returnRO.r = theReward
returnRO.o = theObs
returnRO.terminal = episodeOver
return returnRO
def env_start(self):
self.reset()
returnObs = Observation()
returnObs.doubleArray = self.state.tolist()
return returnObs
def createObservation(self):
obs = Observation(numDoubles=self.obsSize**2)
tmp = self.vmem[int(self.obsOrg[0]):int(self.obsOrg[0] + self.obsSize),
int(self.obsOrg[1]):int(self.obsOrg[1] + self.obsSize)]
obs.doubleArray = list(tmp.flatten())
return obs
def env_start(self):
self.restart_simulation()
returnObs=Observation()
returnObs.doubleArray=self.getState().tolist()
return returnObs
def env_start(self):
self.reset()
returnObs = Observation()
returnObs.doubleArray = self.pos.tolist() + [self.fuel]
return returnObs
def env_start(self): self.reset() returnObs = Observation() returnObs.doubleArray = self.pos.tolist() + [self.fuel] return returnObs
def env_start(self):
self.reset()
returnObs = Observation()
returnObs.doubleArray = self.state.tolist()
return returnObs
def env_start(self):
self.reset()
returnObs = Observation()
returnObs.doubleArray = self.getState()
return returnObs
def env_start(self): self.reset() returnObs = Observation() returnObs.doubleArray = [self.sampleObservation(0)] return returnObs
def env_start(self):
self.restart_simulation()
returnObs = Observation()
returnObs.doubleArray = self.getState().tolist()
return returnObs
def env_start(self):
self.setStartState()
returnObs = Observation()
returnObs.doubleArray = [self.agentRow, self.agentCol]
return returnObs
def env_start(self):
self.reset()
returnObs = Observation()
returnObs.doubleArray = [self.sampleObservation(0)]
return returnObs
