def __init__(
self,
n_rooms: int,
fluents: AbstractFluents,
saved_automaton: str,
reward: float,
save_to: Optional[str] = None,
):
"""Initialize.
:param n_rooms: the number of rooms in this environment. From this
will follow the fluents and automaton states.
:param fluents: this object contains the fluents valuation function.
This method will check that the valuated fluents are the expected
ones.
:param saved_automaton: path to a saved automaton corresponding
to a temporal goal. Fluents must match.
:param reward: reward suplied when reward is reached.
:param save_to: path where the automaton should be exported.
"""
# Define the propositional symbols
expected_fluents = {"bip", "person", "closed"}
for i in range(1, n_rooms + 1):
at_room, in_room = f"at{i}", f"in{i}"
expected_fluents.add(at_room)
expected_fluents.add(in_room)
# Load automaton
with open(saved_automaton, "rb") as f:
automaton: SimpleDFA = pickle.load(f)
# Check same fluents in valuations
if not fluents.fluents == expected_fluents:
raise ValueError(
f"Symbols do not match: {fluents.fluents} != {expected_fluents}"
)
# NOTE: not checking also for automaton. Assuming correct
# Super
TemporalGoal.__init__(
self,
formula=None, # Provinding automaton directly
reward=reward,
automaton=automaton,
labels=fluents,
extract_fluents=fluents.evaluate,
reward_shaping=False,
zero_terminal_state=False,
)
# Maybe save
if save_to:
self.automaton.to_graphviz().render(save_to)
def __init__(
self,
colors: Sequence[str],
fluents: AbstractFluents,
reward: Optional[float] = 1.0,
save_to: Optional[str] = None,
):
"""Initialize.
:param colors: a sequence of colors, these are the positions that
the agent must reach with the correct order.
:param fluents: a fluents evaluator. All colors must be fluents, so
that we know when the agent is in each position.
:param reward: reward suplied when reward is reached.
:param save_to: path where the automaton should be exported.
"""
# Check
if not all((color in fluents.fluents for color in colors)):
raise ValueError(
"Some color has no associated fluent to evaluate it")
# Make automaton for this sequence
automaton = self._make_sapientino_automaton(colors)
# Super
TemporalGoal.__init__(
self,
formula=None, # Provinding automaton directly
reward=reward,
automaton=automaton,
labels=set(colors),
extract_fluents=fluents.evaluate,
reward_shaping=False,
zero_terminal_state=False,
)
# Maybe save
if save_to:
self.automaton.to_graphviz().render(save_to)