def _update(self):
""" integrate self.action and set new self.state and self.reward.
self.state is coded as follows:
index 0: if in start state, this indicates goal up
index 1: if in start state, this indicates goal down
index 2: this indicates robot is neither at start state nor at the junction
index 3: this indicates robot is at the junction.
"""
Maze._update(self)
# set new state
self.state = zeros(4)
if self.perseus == self.initPos[0]:
if self.goUp:
self.state[0] = 1
else:
self.state[1] = 1
elif self.perseus[1] == self.length + 1:
self.state[2] = 1
else:
self.state[3] = 1
# set new reward
if self.perseus[1] == self.length + 1:
if abs(self.perseus[0] - self.goal[0]) == 2:
# bad choice taken
self.perseus = self.goal
self.reward = self.bangReward
def _update(self):
Maze._update(self)
# set reward
if self.goal == self.perseus:
self.reward = 1.
# self.reset()
else:
self.reward = 0
# set state
self.state = array([self.perseus[0] * self.mazeTable.shape[0] + self.perseus[1]])