class Agent(object):
def __init__(self, position):
self.encoder = OneDDepthEncoder(positions=positions,
radius=5,
wrapAround=True,
nPerPosition=28,
wPerPosition=3,
minVal=0,
maxVal=1)
self.plotter = Plotter(self.encoder)
self.learner = QLearner(ACTIIONS, n=1008)
self.lastState = None
self.lastAction = None
def sync(self, outputData):
if not ("ForwardsSweepSensor" in outputData and
"steer" in outputData):
print "Warning: Missing data:", outputData
return
if outputData.get("reset"):
print "Reset."
sensor = outputData["ForwardsSweepSensor"]
steer = outputData["steer"]
reward = outputData.get("reward") or 0
encoding = self.encoder.encode(numpy.array(sensor))
if self.lastState is not None:
self.learner.update(self.lastState, str(self.lastAction),
encoding, str(steer), reward)
value = self.learner.value(encoding)
qValues = {}
for action in ACTIIONS:
qValues[action] = self.learner.qValue(encoding, action)
inputData = {}
inputData["qValues"] = qValues
inputData["bestAction"] = self.learner.bestAction(encoding)
self.plotter.update(sensor, encoding, steer, reward, value, qValues)
if outputData.get("reset"):
self.plotter.render()
self.lastState = encoding
self.lastAction = steer
return inputData
class Agent(object):
def __init__(self, position):
self.encoder = OneDDepthEncoder(positions=positions,
radius=5,
wrapAround=True,
nPerPosition=28,
wPerPosition=3,
minVal=0,
maxVal=1)
self.plotter = Plotter(self.encoder)
self.learner = QLearner(ACTIIONS, n=1008)
self.lastState = None
self.lastAction = None
def sync(self, outputData):
if not ("ForwardsSweepSensor" in outputData and "steer" in outputData):
print "Warning: Missing data:", outputData
return
if outputData.get("reset"):
print "Reset."
sensor = outputData["ForwardsSweepSensor"]
steer = outputData["steer"]
reward = outputData.get("reward") or 0
encoding = self.encoder.encode(numpy.array(sensor))
if self.lastState is not None:
self.learner.update(self.lastState, str(self.lastAction), encoding,
str(steer), reward)
value = self.learner.value(encoding)
qValues = {}
for action in ACTIIONS:
qValues[action] = self.learner.qValue(encoding, action)
inputData = {}
inputData["qValues"] = qValues
inputData["bestAction"] = self.learner.bestAction(encoding)
self.plotter.update(sensor, encoding, steer, reward, value, qValues)
if outputData.get("reset"):
self.plotter.render()
self.lastState = encoding
self.lastAction = steer
return inputData
def __init__(self, position):
self.encoder = OneDDepthEncoder(positions=positions,
radius=5,
wrapAround=True,
nPerPosition=28,
wPerPosition=3,
minVal=0,
maxVal=1)
self.plotter = Plotter(self.encoder)
self.learner = QLearner(ACTIIONS, n=1008)
self.lastState = None
self.lastAction = None
def run(positions, plotEvery=1):
encoder = OneDDepthEncoder(positions=positions,
radius=5,
wrapAround=True,
nPerPosition=28,
wPerPosition=3,
minVal=0,
maxVal=1)
fetcher = Fetcher()
plotter = Plotter(encoder)
learner = QLearner(ACTIIONS, n=1008)
lastState = None
lastAction = None
while True:
outputData = fetcher.sync()
if outputData is None:
continue
if fetcher.skippedTimesteps > 0:
print ("Warning: skipped {0} timesteps, "
"now at {1}").format(fetcher.skippedTimesteps, fetcher.timestep)
if not ("reset" in outputData and
"ForwardsSweepSensor" in outputData and
"steer" in outputData):
print ("Warning: Missing data on timestep {0}: {1}".format(
fetcher.timestep, outputData))
continue
if outputData["reset"]:
print "Reset."
sensor = outputData["ForwardsSweepSensor"]
steer = outputData["steer"]
reward = outputData.get("reward") or 0
encoding = encoder.encode(numpy.array(sensor))
if lastState is not None:
learner.update(lastState, str(lastAction), encoding, str(steer), reward)
value = learner.value(encoding)
qValues = {}
for action in ACTIIONS:
qValues[action] = learner.qValue(encoding, action)
fetcher.inputData["qValues"] = qValues
fetcher.inputData["bestAction"] = learner.bestAction(encoding)
plotter.update(sensor, encoding, steer, reward, value, qValues)
# if fetcher.timestep % plotEvery == 0:
# plotter.render()
lastState = encoding
lastAction = steer
def __init__(self, position):
self.encoder = OneDDepthEncoder(positions=positions,
radius=5,
wrapAround=True,
nPerPosition=28,
wPerPosition=3,
minVal=0,
maxVal=1)
self.plotter = Plotter(self.encoder)
self.learner = QLearner(ACTIIONS, n=1008)
self.lastState = None
self.lastAction = None
