def __init__(self, gym_env: gym.Env,
set_state: Callable[[gym.Env, D.T_memory[D.T_state]], None] = None,
get_state: Callable[[gym.Env], D.T_memory[D.T_state]] = None,
continuous_feature_fidelity: int = 1,
discretization_factor: int = 10,
branching_factor: int = None,
max_depth: int = None) -> None:
"""Initialize GymRIWDomain.
# Parameters
gym_env: The deterministic Gym environment (gym.env) to wrap.
set_state: Function to call to set the state of the gym environment.
If None, default behavior is to deepcopy the environment when changing state
get_state: Function to call to get the state of the gym environment.
If None, default behavior is to deepcopy the environment when changing state
continuous_feature_fidelity: Number of integers to represent a continuous feature
in the interval-based feature abstraction (higher is more precise)
discretization_factor: Number of discretized action variable values per continuous action variable
branching_factor: if not None, sample branching_factor actions from the resulting list of discretized actions
max_depth: maximum depth of states to explore from the initial state
"""
GymDomainHashable.__init__(self, gym_env=gym_env)
GymDiscreteActionDomain.__init__(self,
discretization_factor=discretization_factor,
branching_factor=branching_factor)
GymWidthDomain.__init__(self, continuous_feature_fidelity=continuous_feature_fidelity)
gym_env._max_episode_steps = max_depth
self._max_depth = max_depth
self._current_depth = 0
self._cumulated_reward = 0
self._continuous_feature_fidelity = continuous_feature_fidelity
self._map = None
self._path = None
self._must_reset_features = True
def __init__(
self,
gym_env: gym.Env,
discretization_factor: int = 10,
branching_factor: int = None,
max_depth: int = None,
) -> None:
"""Initialize GymRIWDomain.
# Parameters
gym_env: The deterministic Gym environment (gym.env) to wrap.
discretization_factor: Number of discretized action variable values per continuous action variable
branching_factor: if not None, sample branching_factor actions from the resulting list of discretized actions
max_depth: maximum depth of states to explore from the initial state
"""
GymDomain.__init__(self, gym_env=gym_env)
GymDiscreteActionDomain.__init__(
self,
discretization_factor=discretization_factor,
branching_factor=branching_factor,
)
gym_env._max_episode_steps = max_depth
self.current_outcome = None
self._map = None
self._path = None
def __init__(self,
gym_env: gym.Env,
set_state: Callable[[gym.Env, D.T_memory[D.T_state]],
None] = None,
get_state: Callable[[gym.Env], D.T_memory[D.T_state]] = None,
discretization_factor: int = 10,
branching_factor: int = None,
horizon: int = 1000) -> None:
"""Initialize GymRIWDomain.
# Parameters
gym_env: The deterministic Gym environment (gym.env) to wrap.
set_state: Function to call to set the state of the gym environment.
If None, default behavior is to deepcopy the environment when changing state
get_state: Function to call to get the state of the gym environment.
If None, default behavior is to deepcopy the environment when changing state
discretization_factor: Number of discretized action variable values per continuous action variable
branching_factor: if not None, sample branching_factor actions from the resulting list of discretized actions
horizon: maximum number of steps allowed for the gym environment
"""
DeterministicGymDomain.__init__(self,
gym_env=gym_env,
set_state=set_state,
get_state=get_state)
GymDiscreteActionDomain.__init__(
self,
discretization_factor=discretization_factor,
branching_factor=branching_factor)
gym_env._max_episode_steps = horizon
self._map = None
self._path = None
def __init__(self, gym_env: gym.Env,
set_state: Callable[[gym.Env, D.T_memory[D.T_state]], None] = None,
get_state: Callable[[gym.Env], D.T_memory[D.T_state]] = None,
termination_is_goal: bool = True,
continuous_feature_fidelity: int = 1,
discretization_factor: int = 3,
branching_factor: int = None,
max_depth: int = 50) -> None:
GymPlanningDomain.__init__(self,
gym_env=gym_env,
set_state=set_state,
get_state=get_state,
termination_is_goal=termination_is_goal,
max_depth=max_depth)
GymDiscreteActionDomain.__init__(self,
discretization_factor=discretization_factor,
branching_factor=branching_factor)
GymWidthDomain.__init__(self, continuous_feature_fidelity=continuous_feature_fidelity)
gym_env._max_episode_steps = max_depth
def __init__(self,
gym_env: gym.Env,
set_state: Callable[[gym.Env, D.T_memory[D.T_state]],
None] = None,
get_state: Callable[[gym.Env], D.T_memory[D.T_state]] = None,
termination_is_goal: bool = True,
continuous_feature_fidelity: int = 1,
discretization_factor: int = 3,
branching_factor: int = None,
max_depth: int = 50) -> None:
"""Initialize GymIWDomain.
# Parameters
gym_env: The deterministic Gym environment (gym.env) to wrap.
set_state: Function to call to set the state of the gym environment.
If None, default behavior is to deepcopy the environment when changing state
get_state: Function to call to get the state of the gym environment.
If None, default behavior is to deepcopy the environment when changing state
termination_is_goal: True if the termination condition is a goal (and not a dead-end)
continuous_feature_fidelity: Number of integers to represent a continuous feature
in the interval-based feature abstraction (higher is more precise)
discretization_factor: Number of discretized action variable values per continuous action variable
branching_factor: if not None, sample branching_factor actions from the resulting list of discretized actions
max_depth: maximum depth of states to explore from the initial state
"""
GymPlanningDomain.__init__(self,
gym_env=gym_env,
set_state=set_state,
get_state=get_state,
termination_is_goal=termination_is_goal,
max_depth=max_depth)
GymDiscreteActionDomain.__init__(
self,
discretization_factor=discretization_factor,
branching_factor=branching_factor)
GymWidthDomain.__init__(
self, continuous_feature_fidelity=continuous_feature_fidelity)
gym_env._max_episode_steps = max_depth