# print("Cart velocity bins:", cartVelocityGroup)
observationDigitizer = ArrayDigitizer([cartPositionGroup, cartVelocityGroup])
transformation = EnvTransformation(observationDigitizer)
task = GymTask.createTask(gymRawEnv)
env = task.env
env.setTransformation(transformation)
## env.setCumulativeRewardMode()
# create value table and initialize with ones
table = ActionValueTable(observationDigitizer.states, env.numActions)
if state_load_file is not None:
loadedParams = np.fromfile(state_load_file)
table._setParameters(loadedParams)
# print( "raw data:", loadedParams )
else:
# table.initialize(0.0)
rand_arr = np.random.rand(table.paramdim)
table.initialize(rand_arr)
# create agent with controller and learner - use SARSA(), Q() or QLambda() here
## alpha -- learning rate (preference of new information)
## gamma -- discount factor (importance of future reward)
# learner = Q(0.5, 0.99)
learner = SARSA(0.5, 0.99)
# learner = QLambda(0.5, 0.99, 0.9)
explorer = learner.explorer
