def get_all_states(self):
"""Returns all states.
"""
states = []
for x in range(1, self.width + 1):
for y in range(1, self.height + 1):
state = NavigationWorldState(x, y)
if self.is_goal(x, y) and self.is_goal_terminal:
state.set_terminal(True)
states.append(state)
return states
def _transition_func(self, state, action):
"""
Args:
state (State)
action (str)
Returns
(State)
"""
if state.is_terminal():
return state
r = random.random()
if self.slip_prob > r:
# Flip dir.
if action == "up":
action = random.choice(["left", "right"])
elif action == "down":
action = random.choice(["left", "right"])
elif action == "left":
action = random.choice(["up", "down"])
elif action == "right":
action = random.choice(["up", "down"])
if action == "up" and state.y < self.height and not self.is_wall(
state.x, state.y + 1):
next_state = NavigationWorldState(state.x, state.y + 1)
elif action == "down" and state.y > 1 and not self.is_wall(
state.x, state.y - 1):
next_state = NavigationWorldState(state.x, state.y - 1)
elif action == "right" and state.x < self.width and not self.is_wall(
state.x + 1, state.y):
next_state = NavigationWorldState(state.x + 1, state.y)
elif action == "left" and state.x > 1 and not self.is_wall(state.x - 1,
state.y):
next_state = NavigationWorldState(state.x - 1, state.y)
else:
next_state = NavigationWorldState(state.x, state.y)
if self.is_goal(next_state.x, next_state.y) and self.is_goal_terminal:
next_state.set_terminal(True)
return next_state